Manolis Kellis: Human Genome and Evolutionary Dynamics
生物与进化技术与编程音乐与艺术AI 与机器学习心理与人性
📋 章节目录
暂无章节信息
🔑 关键词
humangenomedonbrainhumanscellswholespeciesproteinceterabeautifulbettergenessingleevolutionaryvirusesvirusmeaninggeneevolution
💬 精彩语录
暂无语录
🎙️ 完整对话（3398 条）
Lex Fridman (00:00.000)
The following is a conversation with Manolis Kellis.
以下是与马诺利斯·凯利斯的对话。
Lex Fridman (00:03.080)
He's a professor at MIT and head
他是麻省理工学院的教授兼院长
Lex Fridman (00:05.440)
of the MIT Computational Biology Group.
麻省理工学院计算生物学组的成员。
Lex Fridman (00:08.400)
He's interested in understanding the human genome
他对了解人类基因组很感兴趣
Lex Fridman (00:11.280)
from a computational, evolutionary, biological,
从计算的、进化的、生物学的、
Lex Fridman (00:14.160)
and other cross disciplinary perspectives.
以及其他跨学科的观点。
Lex Fridman (00:17.000)
He has more big, impactful papers and awards
他有更多有影响力的大论文和奖项
Manolis Kellis (00:20.160)
than I can list, but most importantly,
我无法列出，但最重要的是，
Lex Fridman (00:22.480)
he's a kind, curious, brilliant human being,
他是一个善良、好奇、才华横溢的人，
Lex Fridman (00:26.240)
and just someone I really enjoy talking to.
我真的很喜欢与之交谈的人。
Lex Fridman (00:28.720)
His passion for science and life in general is contagious.
他对科学和生活的热情具有感染力。
Manolis Kellis (00:32.960)
The hours honestly flew by,
老实说，时间过得很快，
Lex Fridman (00:34.640)
and I'm sure we'll talk again on this podcast soon.
我相信我们很快就会在这个播客上再次讨论。
Manolis Kellis (00:37.720)
Quick summary of the ads.
广告的快速摘要。
Lex Fridman (00:39.120)
Three sponsors, Blinkist, Aidsleep, and Masterclass.
三个赞助商：Blinkist、Aidsleep 和 Masterclass。
Manolis Kellis (00:43.160)
Please consider supporting this podcast
请考虑支持此播客
Lex Fridman (00:44.960)
by going to blinkist.com slash lex,
通过访问blinkist.com斜杠lex，
Manolis Kellis (00:47.760)
aidsleep.com slash lex,
adsleep.com 斜杠 lex，
Lex Fridman (00:49.840)
and signing up at masterclass.com slash lex.
并在 masterclass.com 斜杠 lex 上注册。
Manolis Kellis (00:53.320)
Click the links, buy the stuff, get the discount.
点击链接，购买商品，获得折扣。
Lex Fridman (00:56.120)
It's the best way to support this podcast.
Manolis Kellis (00:58.880)
If you enjoy this thing, subscribe on YouTube,
Lex Fridman (01:01.160)
review it with five stars on Apple Podcast,
Manolis Kellis (01:03.400)
support it on Patreon,
Lex Fridman (01:04.680)
or connect with me on Twitter at Lex Friedman.
Manolis Kellis (01:08.160)
As usual, I'll do a few minutes of ads now,
Lex Fridman (01:10.320)
and never any ads in the middle
Manolis Kellis (01:11.840)
that can break the flow of the conversation.
Lex Fridman (01:14.520)
This episode is supported by Blinkist,
Manolis Kellis (01:17.200)
my favorite app for learning new things.
Lex Fridman (01:19.760)
Get it at blinkist.com slash lex
Manolis Kellis (01:22.040)
for a seven day free trial and 25% off afterwards.
Lex Fridman (01:26.960)
Blinkist takes the key ideas
Manolis Kellis (01:28.400)
from thousands of nonfiction books
Lex Fridman (01:30.360)
and condenses them down into just 15 minutes
Manolis Kellis (01:33.200)
that you can read or listen to.
Lex Fridman (01:35.400)
I'm a big believer in reading at least an hour every day.
Manolis Kellis (01:38.960)
As part of that, I use Blinkist every day
Lex Fridman (01:41.920)
to try out a book I may otherwise
Manolis Kellis (01:43.920)
never have a chance to read.
Lex Fridman (01:45.560)
And in general, it's a great way to broaden your view
Manolis Kellis (01:48.360)
of the ideal landscape out there,
Lex Fridman (01:50.600)
and find books that you may want to read more deeply.
Manolis Kellis (01:53.960)
With Blinkist, you get unlimited access
Lex Fridman (01:56.160)
to read or listen to a massive library
Manolis Kellis (01:58.640)
of condensed nonfiction books.
Lex Fridman (02:00.680)
Go to blinkist.com slash lex to try it free for seven days
Lex Fridman (02:05.280)
and save 25% off your new subscription.
Lex Fridman (02:08.320)
That's blinkist.com slash lex.
Manolis Kellis (02:10.800)
Blinkist, spelled B L I N K I S T.
Lex Fridman (02:15.960)
This show is also sponsored by Asleep
Lex Fridman (02:18.200)
and it's pod pro mattress.
Lex Fridman (02:19.800)
You can check out at Asleep.com slash lex
Manolis Kellis (02:22.480)
to get $200 off.
Lex Fridman (02:24.600)
It controls temperature with a nap
Lex Fridman (02:26.720)
and can cool down to as low as 55 degrees
Lex Fridman (02:29.200)
on each side of the bed separately.
Manolis Kellis (02:31.480)
Research shows that temperature has a big impact
Lex Fridman (02:33.840)
on the quality of our sleep.
Manolis Kellis (02:35.680)
Anecdotally, as been true for me,
Lex Fridman (02:37.840)
it's truly been a game changer.
Manolis Kellis (02:39.840)
I love it.
Lex Fridman (02:40.880)
The pod pro is packed with sensors
Manolis Kellis (02:42.400)
that track heart rate, heart rate variability
Lex Fridman (02:44.560)
and respiratory rate, showing it all in their app.
Manolis Kellis (02:48.120)
The app's health metrics are amazing
Lex Fridman (02:50.080)
but the cooling alone is honestly worth the money.
Manolis Kellis (02:52.880)
Check it out at Asleep.com slash lex to get $200 off.
Lex Fridman (02:57.440)
This show is also sponsored by Masterclass.
Manolis Kellis (03:00.280)
Sign up at Masterclass.com slash lex
Lex Fridman (03:02.560)
to get a discount and to support this podcast.
Manolis Kellis (03:05.440)
When I first heard about Masterclass,
Lex Fridman (03:07.040)
I thought it was too good to be true.
Manolis Kellis (03:09.000)
For 180 bucks a year, you get an all access pass
Lex Fridman (03:12.080)
to watch courses from, to list some of my favorites,
Manolis Kellis (03:15.440)
Chris Hadfield on Space Exploration,
Lex Fridman (03:17.680)
Neil deGrasse Tyson on Scientific Thinking
Lex Fridman (03:19.600)
and Communication, Will Wright,
Lex Fridman (03:21.640)
one of my favorite game designers,
Manolis Kellis (03:23.440)
Carlos Santana, one of my favorite guitar players,
Lex Fridman (03:26.480)
Garry Kasparov, of course,
Manolis Kellis (03:27.960)
the greatest chess player of all time, I'm not biased,
Lex Fridman (03:30.800)
Daniel Negrano on poker and many more.
Manolis Kellis (03:33.760)
Chris Hadfield explaining how rockets work
Lex Fridman (03:35.880)
and the experience of being launched into space alone
Manolis Kellis (03:38.720)
is worth the money.
Lex Fridman (03:40.040)
By the way, you can watch it on basically any device.
Manolis Kellis (03:43.120)
Once again, sign up at Masterclass.com slash lex
Lex Fridman (03:46.200)
to get a discount and to support this podcast.
Lex Fridman (03:49.600)
And now, here's my conversation with Manolis Kellis.
Lex Fridman (03:54.600)
What to you is the most beautiful aspect
Lex Fridman (03:56.880)
of the human genome?
Lex Fridman (03:58.640)
Don't get me started.
Manolis Kellis (04:00.040)
So. We've got time.
Lex Fridman (04:04.280)
The first answer is that the beauty of genomes
Manolis Kellis (04:06.360)
transcends humanity.
Lex Fridman (04:07.680)
So it's not just about the human genome.
Manolis Kellis (04:09.440)
Genomes in general are amazingly beautiful.
Lex Fridman (04:12.640)
And again, I'm obviously biased.
Lex Fridman (04:14.040)
So in my view, the way that I like to introduce
Lex Fridman (04:18.680)
the human genome and the way that I like to introduce
Manolis Kellis (04:20.840)
genomics to my class is by telling them,
Lex Fridman (04:22.920)
you know, we're not the inventors
Manolis Kellis (04:24.960)
of the first digital computer.
Lex Fridman (04:26.560)
We are the descendants of the first digital computer.
Manolis Kellis (04:30.600)
Basically, life is digital.
Lex Fridman (04:32.360)
And that's absolutely beautiful about life.
Manolis Kellis (04:34.600)
The fact that at every replication step,
Lex Fridman (04:37.080)
you don't lose any information
Manolis Kellis (04:38.480)
because that information is digital.
Lex Fridman (04:40.040)
If it was analog, if it was just sprouting concentrations,
Manolis Kellis (04:43.160)
you'd lose it after a few generations.
Lex Fridman (04:44.640)
It would just dissolve away.
Lex Fridman (04:46.360)
And that's what the ancients
Lex Fridman (04:48.240)
didn't understand about inheritance.
Manolis Kellis (04:50.080)
The first person to understand digital inheritance
Lex Fridman (04:52.120)
was Mendel, of course.
Lex Fridman (04:54.440)
And his theory, in fact, stayed in a bookshelf
Lex Fridman (04:57.600)
for like 50 years while Darwin was getting famous
Manolis Kellis (05:00.720)
about natural selection.
Lex Fridman (05:02.440)
But the missing component was this digital inheritance,
Manolis Kellis (05:05.440)
the mechanism of evolution that Mendel had discovered.
Lex Fridman (05:09.720)
So that aspect in my view is the most beautiful aspect
Lex Fridman (05:13.120)
but it transcends all of life.
Lex Fridman (05:14.880)
And can you elaborate maybe the inheritance part?
Lex Fridman (05:18.040)
What was the key thing that the ancients didn't understand?
Lex Fridman (05:22.480)
So the very theory of inheritance as discrete units,
Manolis Kellis (05:28.520)
throughout the life of Mendel and well after he's writing,
Lex Fridman (05:32.880)
people thought that his P experiments
Manolis Kellis (05:35.240)
were just a little fluke,
Lex Fridman (05:36.600)
that they were just a little exception
Manolis Kellis (05:38.760)
that would normally not even apply to humans,
Lex Fridman (05:41.840)
that basically what they saw
Manolis Kellis (05:44.200)
is this continuum of eye color,
Lex Fridman (05:48.080)
this continuum of skin color,
Manolis Kellis (05:49.760)
this continuum of hair color,
Lex Fridman (05:51.120)
this continuum of height.
Lex Fridman (05:52.520)
And all of these continuums did not fit
Lex Fridman (05:55.080)
with a discrete type of inheritance
Manolis Kellis (05:56.800)
that Mendel was describing.
Lex Fridman (05:58.920)
But what's unique about genomics
Lex Fridman (06:00.480)
and what's unique about the genome
Lex Fridman (06:01.640)
is really that there are two copies
Lex Fridman (06:03.840)
and that you get a combination of these.
Lex Fridman (06:06.240)
But for every trait,
Manolis Kellis (06:08.200)
there are dozens of contributing variables.
Lex Fridman (06:10.720)
And it was only Ronald Fisher in the 20th century
Manolis Kellis (06:14.000)
that basically recognized that even five Mendelian traits
Lex Fridman (06:20.120)
would add up to a continuum like inheritance pattern.
Lex Fridman (06:24.920)
And he wrote a series of papers
Lex Fridman (06:27.520)
that still are very relevant today
Manolis Kellis (06:30.360)
about sort of this Mendelian inheritance
Lex Fridman (06:32.960)
of continuum like traits.
Lex Fridman (06:35.200)
And I think that that was the missing step in inheritance.
Lex Fridman (06:38.520)
So well before the discovery of the structure of DNA,
Manolis Kellis (06:41.560)
which is again, another amazingly beautiful aspect,
Lex Fridman (06:44.400)
the double helix,
Lex Fridman (06:45.240)
what I like to call the most noble molecule of our time,
Lex Fridman (06:50.320)
holds within it the secret of that discrete inheritance,
Lex Fridman (06:54.080)
but the conceptualization of discrete elements
Lex Fridman (06:58.480)
is something that precedes that.
Lex Fridman (06:59.680)
So even though it's discrete,
Lex Fridman (07:01.440)
when it materializes itself into actual traits that we see,
Manolis Kellis (07:06.560)
it can be continuous.
Lex Fridman (07:08.000)
Basically arbitrarily rich and complex.
Lex Fridman (07:10.840)
So if you have five genes that contribute to human height,
Lex Fridman (07:15.000)
and there aren't five, there's a thousand.
Manolis Kellis (07:16.960)
If there's only five genes
Lex Fridman (07:18.440)
and you inherit some combination of them,
Lex Fridman (07:20.880)
and every one makes you two inches taller
Lex Fridman (07:23.080)
or two inches shorter,
Manolis Kellis (07:24.520)
it'll look like a continuous trait.
Lex Fridman (07:28.520)
But instead of five, there are thousands.
Lex Fridman (07:30.440)
And every one of them contributes to less than one millimeter.
Lex Fridman (07:33.880)
We change in height more during the day
Manolis Kellis (07:36.920)
than each of these genetic variants contributes.
Lex Fridman (07:40.160)
So by the evening, you're shorter than you walk up with.
Manolis Kellis (07:43.960)
Isn't that weird then
Lex Fridman (07:45.320)
that we're not more different than we are?
Lex Fridman (07:48.080)
Why are we all so similar
Lex Fridman (07:49.680)
if there's so much possibility to be different?
Manolis Kellis (07:52.600)
Yeah, so there are selective advantages to being medium.
Lex Fridman (07:57.480)
If you're extremely tall or extremely short,
Manolis Kellis (07:59.920)
you run into selective disadvantages.
Lex Fridman (08:02.280)
So you have trouble breathing, you have trouble running,
Manolis Kellis (08:04.160)
you have trouble sitting if you're too tall.
Lex Fridman (08:06.320)
If you're too short, you might, I don't know,
Manolis Kellis (08:08.240)
have other selective pressures are acting against that.
Lex Fridman (08:11.000)
If you look at natural history of human population,
Manolis Kellis (08:13.600)
there's actually selection for height in Northern Europe
Lex Fridman (08:17.000)
and selection against height in Southern Europe.
Lex Fridman (08:19.960)
So there might actually be advantages
Lex Fridman (08:21.800)
to actually being not super tall.
Lex Fridman (08:25.000)
And if you look across the entire human population,
Lex Fridman (08:28.720)
for many, many traits,
Manolis Kellis (08:29.840)
there's a lot of push towards the middle.
Lex Fridman (08:32.080)
Balancing selection is the usual term
Manolis Kellis (08:35.280)
for selection that sort of seeks to not be extreme
Lex Fridman (08:39.680)
and to sort of have a combination of alleles
Manolis Kellis (08:43.600)
that sort of keep recombining.
Lex Fridman (08:45.920)
And if you look at mate selection,
Manolis Kellis (08:48.640)
super, super tall people
Lex Fridman (08:50.320)
will not tend to sort of marry super, super tall people.
Manolis Kellis (08:53.360)
Very often you see these couples
Lex Fridman (08:55.200)
that are kind of compensating for each other.
Lex Fridman (08:57.960)
And the best predictor of the kid's age
Lex Fridman (09:00.440)
is very often just take the average of the two parents
Lex Fridman (09:03.520)
and then adjust for sex and boom, you get it.
Lex Fridman (09:07.080)
It's extremely heritable.
Manolis Kellis (09:08.520)
Let me ask, you kind of took a step back to the genome
Lex Fridman (09:12.680)
outside of just humans,
Lex Fridman (09:13.880)
but is there something that you find beautiful
Lex Fridman (09:15.880)
about the human genome specifically?
Lex Fridman (09:18.600)
So I think the genome,
Lex Fridman (09:21.360)
if more people understood the beauty of the human genome,
Manolis Kellis (09:24.440)
there would be so many fewer wars,
Lex Fridman (09:26.600)
so much less anger in the world.
Manolis Kellis (09:28.960)
I mean, what's really beautiful about the human genome
Lex Fridman (09:31.400)
is really the variation
Manolis Kellis (09:33.840)
that teaches us both about individuality
Lex Fridman (09:36.520)
and about similarity.
Lex Fridman (09:38.200)
So any two people on the planet are 99.9% identical.
Lex Fridman (09:43.440)
How can you fight with someone who's 99.9% identical to you?
Manolis Kellis (09:47.280)
It's just counterintuitive.
Lex Fridman (09:49.680)
And yet any two siblings of the same parents
Manolis Kellis (09:53.880)
differ in millions of locations.
Lex Fridman (09:57.080)
So every one of them is basically two to the million unique
Manolis Kellis (10:01.000)
from any pair of parents,
Lex Fridman (10:03.120)
let alone any two random parents on the planet.
Lex Fridman (10:05.760)
So that's, I think, something that teaches us
Lex Fridman (10:08.520)
about sort of the nature of humanity in many ways,
Manolis Kellis (10:11.240)
that every one of us is as unique as any star
Lex Fridman (10:14.640)
and way more unique in actually many ways.
Lex Fridman (10:17.160)
And yet we're all brothers and sisters.
Lex Fridman (10:22.160)
Yeah, just like stars, most of it is just fusion reactions.
Manolis Kellis (10:26.160)
Yeah, you only have a few parameters to describe stars.
Lex Fridman (10:29.280)
Mass, size, initial size, and stage of life.
Manolis Kellis (10:33.120)
Whereas for humans, it's thousands of parameters
Lex Fridman (10:36.360)
scattered across our genome.
Lex Fridman (10:38.080)
So the other thing that makes humans unique,
Lex Fridman (10:41.360)
the other things that makes inheritance unique in humans
Manolis Kellis (10:45.240)
is that most species inherit things vertically.
Lex Fridman (10:50.360)
Basically instinct is a huge part of their behavior.
Manolis Kellis (10:54.480)
The way that, I mean, with my kids,
Lex Fridman (10:57.560)
we've been watching this nest of birds
Manolis Kellis (11:00.960)
with two little eggs outside our window
Lex Fridman (11:03.560)
for the last few months,
Manolis Kellis (11:05.200)
for the last few weeks as they've been growing.
Lex Fridman (11:07.440)
And there's so much behavior that's hard coded.
Manolis Kellis (11:12.240)
Birds don't just learn as they grow.
Lex Fridman (11:16.000)
There's no culture.
Manolis Kellis (11:17.160)
Like a bird that's born in Boston
Lex Fridman (11:19.480)
will be the same as a bird that's born in California.
Lex Fridman (11:22.040)
So there's not as much inheritance of ideas, of customs.
Lex Fridman (11:27.920)
A lot of it is hard coding in their genome.
Manolis Kellis (11:30.520)
What's really beautiful about the human genome
Lex Fridman (11:32.280)
is that if you take a person from today
Lex Fridman (11:35.000)
and you place them back in ancient Egypt,
Lex Fridman (11:37.080)
or if you take a person from ancient Egypt
Lex Fridman (11:39.120)
and you place them here today,
Lex Fridman (11:41.160)
they will grow up to be completely normal.
Manolis Kellis (11:44.920)
That is not genetics.
Lex Fridman (11:47.680)
This is the other type of inheritance in humans.
Lex Fridman (11:51.840)
So on one hand, we have the genetic inheritance,
Lex Fridman (11:53.880)
which is vertical from your parents down.
Manolis Kellis (11:56.160)
On the other hand, we have horizontal inheritance,
Lex Fridman (11:58.400)
which is the ideas that are built up at every generation
Manolis Kellis (12:02.400)
are horizontally transmitted.
Lex Fridman (12:04.600)
And the huge amount of time
Manolis Kellis (12:06.560)
that we spend in educating ourselves,
Lex Fridman (12:09.480)
a concept known as neoteny,
Manolis Kellis (12:11.920)
neo for newborn and then teny for holding.
Lex Fridman (12:15.200)
So if you look at humans,
Manolis Kellis (12:17.320)
I mean, the little birds that were eggs two weeks ago,
Lex Fridman (12:20.400)
and now one of them has already flown off.
Manolis Kellis (12:22.680)
The other one's ready to fly off.
Lex Fridman (12:24.520)
In two weeks, they're ready to just fend for themselves.
Manolis Kellis (12:27.280)
Humans, 16 years, 18 years, 24, getting out of college.
Lex Fridman (12:33.040)
I'm still learning.
Lex Fridman (12:34.440)
So that's so fascinating,
Lex Fridman (12:36.000)
this picture of a vertical and the horizontal.
Manolis Kellis (12:38.720)
When you talk about the horizontal,
Lex Fridman (12:40.080)
is it in the realm of ideas?
Manolis Kellis (12:41.960)
Exactly.
Lex Fridman (12:42.800)
Okay, so it's the actual social interactions.
Manolis Kellis (12:45.320)
That's exactly right.
Lex Fridman (12:46.200)
That's exactly right.
Lex Fridman (12:47.080)
So basically the concept of neoteny
Lex Fridman (12:49.200)
is that you spend acquiring characteristics
Manolis Kellis (12:52.760)
from your environment
Lex Fridman (12:54.160)
in an extremely malleable state of your brain
Lex Fridman (12:56.960)
and the wiring of your brain for a long period of your life.
Lex Fridman (13:00.640)
Compared to primates, we are useless.
Manolis Kellis (13:03.480)
You take any primate at seven weeks
Lex Fridman (13:05.320)
and any human at seven weeks, we lose the battle.
Lex Fridman (13:08.440)
But at 18 years, you know, all bets are off.
Lex Fridman (13:11.640)
Like we basically, our brain continues to develop
Manolis Kellis (13:14.840)
in an extremely malleable form till very late.
Lex Fridman (13:17.760)
And this is what allows education.
Manolis Kellis (13:20.360)
This is what allows the person from Egypt
Lex Fridman (13:22.480)
to do extremely well now.
Lex Fridman (13:24.680)
And the reason for that is that the wiring of our brain
Lex Fridman (13:31.400)
and the development of that wiring is actually delayed.
Manolis Kellis (13:34.960)
So, you know, the longer you delay that,
Lex Fridman (13:37.480)
the more opportunity you have to pass on knowledge,
Manolis Kellis (13:40.840)
to pass on concepts, ideals, ideas
Lex Fridman (13:44.240)
from the parents to the child.
Lex Fridman (13:46.080)
And what's really absolutely beautiful about humans today
Lex Fridman (13:49.080)
is that that lateral transfer of ideas and culture
Manolis Kellis (13:52.080)
is not just from uncles and aunts and teachers at school,
Lex Fridman (13:55.680)
but it's from Wikipedia and review articles on the web
Lex Fridman (14:00.280)
and thousands of journals
Lex Fridman (14:02.600)
that are sort of putting out information for free
Lex Fridman (14:05.480)
and podcasts and videocasts and all of that stuff
Lex Fridman (14:08.840)
where you can basically learn about any topic,
Manolis Kellis (14:12.840)
pretty much everything that would be in any
Lex Fridman (14:16.360)
super advanced textbook in a matter of days,
Manolis Kellis (14:19.520)
instead of having to go to the library of Alexandria
Lex Fridman (14:22.920)
and sail there to read three books
Lex Fridman (14:24.720)
and then sail for another few days to get to Athens
Lex Fridman (14:27.000)
and et cetera, et cetera, et cetera.
Lex Fridman (14:28.640)
So the democratization of knowledge
Lex Fridman (14:31.360)
and the spread, the speed of spread of knowledge
Manolis Kellis (14:34.280)
is what defines, I think, the human inheritance pattern.
Lex Fridman (14:38.920)
So you sound excited about it, are you also a little bit
Manolis Kellis (14:43.200)
afraid or are you more excited by the power
Lex Fridman (14:46.440)
of this kind of distributed spread of information?
Lex Fridman (14:49.880)
So you put it very kindly that most people
Lex Fridman (14:52.040)
are kind of using the internet and looking Wikipedia,
Manolis Kellis (14:55.280)
reading articles, reading papers and so on,
Lex Fridman (14:58.120)
but if we're honest, most people online,
Manolis Kellis (15:02.000)
especially when they're younger,
Lex Fridman (15:03.200)
probably looking at five second clips on TikTok
Manolis Kellis (15:05.840)
or whatever the new social network is,
Lex Fridman (15:08.520)
are you, given this power of horizontal inheritance,
Manolis Kellis (15:12.520)
are you optimistic or a little bit pessimistic
Lex Fridman (15:16.520)
about this new effect of the internet
Lex Fridman (15:22.520)
and democratization of knowledge on our,
Lex Fridman (15:26.960)
what would you call this, this genome,
Lex Fridman (15:29.400)
would you use the term genome, by the way, for this?
Lex Fridman (15:31.880)
Yeah, I think we use the genome to talk about DNA,
Lex Fridman (15:36.200)
but very often we say, I'm Greek,
Lex Fridman (15:38.960)
so people ask me, hey, what's in the Greek genome?
Lex Fridman (15:40.760)
And I'm like, well, yeah, what's in the Greek genome
Lex Fridman (15:42.800)
is both our genes and also our ideas
Lex Fridman (15:44.760)
and our ideals and our culture.
Lex Fridman (15:46.640)
So the poetic meaning of the word.
Manolis Kellis (15:48.240)
Exactly, exactly, yeah.
Lex Fridman (15:50.080)
So I think that there's a beauty
Manolis Kellis (15:55.960)
to the democratization of knowledge,
Lex Fridman (15:57.720)
the fact that you can reach as many people
Manolis Kellis (16:00.200)
as any other person on the planet
Lex Fridman (16:02.800)
and it's not who you are,
Manolis Kellis (16:04.280)
it's really your ideas that matter,
Lex Fridman (16:06.640)
is a beautiful aspect of the internet.
Manolis Kellis (16:11.880)
I think there's, of course, a danger of my ignorance
Lex Fridman (16:15.560)
is as important as your expertise.
Manolis Kellis (16:18.240)
The fact that with this democratization
Lex Fridman (16:21.360)
comes the abolishment of respecting expertise.
Manolis Kellis (16:25.120)
Just because you've spent 10,000 hours of your life
Lex Fridman (16:28.880)
studying, I don't know, human brain circuitry,
Lex Fridman (16:33.320)
why should I trust you?
Lex Fridman (16:34.160)
I'm just gonna make up my own theories
Lex Fridman (16:35.640)
and they'll be just as good as yours,
Lex Fridman (16:37.240)
is an attitude that sort of counteracts
Manolis Kellis (16:39.640)
the beauty of the democratization.
Lex Fridman (16:42.480)
And I think that within our educational system
Lex Fridman (16:47.400)
and within the upbringing of our children,
Lex Fridman (16:49.720)
we have to not only teach them knowledge,
Lex Fridman (16:52.320)
but we have to teach them the means to get to knowledge.
Lex Fridman (16:55.760)
And that, it's very similar to sort of you fish,
Manolis Kellis (16:59.320)
you catch a fish for a man for one day,
Lex Fridman (17:01.400)
you fed them for one day, you teach them how to fish,
Manolis Kellis (17:03.880)
you fed them for the rest of their life.
Lex Fridman (17:05.560)
So instead of just gathering the knowledge
Manolis Kellis (17:07.640)
they need for any one task,
Lex Fridman (17:09.480)
we can just tell them, all right,
Manolis Kellis (17:11.120)
here's how you Google it,
Lex Fridman (17:12.520)
here's how you figure out what's real and what's not,
Manolis Kellis (17:14.640)
here's how you check the sources,
Lex Fridman (17:16.440)
here's how you form a basic opinion for yourself.
Lex Fridman (17:19.200)
And I think that inquisitive nature
Lex Fridman (17:22.880)
is paramount to being able to sort through
Manolis Kellis (17:26.760)
this huge wealth of knowledge.
Lex Fridman (17:29.320)
So you need a basic educational foundation
Manolis Kellis (17:32.560)
based on which you can then add on
Lex Fridman (17:35.520)
the sort of domain specific knowledge,
Lex Fridman (17:38.200)
but that basic educational foundation
Lex Fridman (17:39.720)
should just not just be knowledge,
Lex Fridman (17:42.400)
but it should also be epistemology,
Lex Fridman (17:45.240)
the way to acquire knowledge.
Manolis Kellis (17:47.240)
I'm not sure any of us know how to do that
Lex Fridman (17:49.720)
in this modern day, we're actually learning.
Manolis Kellis (17:51.680)
One of the big surprising thing to me
Lex Fridman (17:53.600)
about the coronavirus, for example,
Manolis Kellis (17:57.280)
is that Twitter has been
Lex Fridman (17:59.680)
one of the best sources of information.
Manolis Kellis (18:02.800)
Basically like building your own network of experts,
Lex Fridman (18:07.960)
as opposed to the traditional centralized expertise
Manolis Kellis (18:11.040)
of the WHO and the CDC,
Lex Fridman (18:13.680)
or maybe any one particular respectable person
Manolis Kellis (18:19.280)
at the top of a department in some kind of institution,
Lex Fridman (18:21.760)
you instead look at 10, 20, hundreds of people,
Manolis Kellis (18:26.520)
some of whom are young kids that are incredibly good
Lex Fridman (18:32.600)
at aggregating data and plotting and visualizing that data.
Manolis Kellis (18:35.800)
That's been really surprising to me.
Lex Fridman (18:37.240)
I don't know what to make of it.
Manolis Kellis (18:39.880)
I don't know how that matures into something stable.
Lex Fridman (18:45.480)
I don't know if you have ideas.
Manolis Kellis (18:47.000)
If you were to just try to explain to your kids
Lex Fridman (18:49.960)
of where should you go to learn about coronavirus,
Lex Fridman (18:54.960)
what would you say?
Lex Fridman (18:56.800)
It's such a beautiful example.
Lex Fridman (18:58.080)
And I think the current pandemic
Lex Fridman (18:59.920)
and the speed at which the scientific community has moved
Manolis Kellis (19:03.960)
in the current pandemic,
Lex Fridman (19:04.780)
I think exemplifies this horizontal transfer
Lex Fridman (19:08.060)
and the speed of horizontal transfer of information.
Lex Fridman (19:10.820)
The fact that the genome was first sequenced
Manolis Kellis (19:15.320)
in early January,
Lex Fridman (19:16.380)
the first sample was obtained December 29, 2019,
Manolis Kellis (19:20.360)
a week after the publication of the first genome sequence,
Lex Fridman (19:23.560)
Moderna had already finalized its vaccine design
Lex Fridman (19:27.800)
and was moving to production.
Lex Fridman (19:29.500)
I mean, this is phenomenal.
Manolis Kellis (19:31.820)
The fact that we go from not knowing
Lex Fridman (19:34.980)
what the heck is killing people in Wuhan
Manolis Kellis (19:36.740)
to wow, it's SARS CoV2 and here's the set of genes,
Lex Fridman (19:41.900)
here's the genome, here's the sequence,
Manolis Kellis (19:43.580)
here are the polymorphisms, et cetera,
Lex Fridman (19:45.660)
in the matter of weeks is phenomenal.
Manolis Kellis (19:48.220)
In that incredible pace of transfer of knowledge,
Lex Fridman (19:52.760)
there have been many mistakes.
Manolis Kellis (19:54.380)
So, some of those mistakes
Lex Fridman (19:56.620)
may have been politically motivated
Manolis Kellis (19:57.940)
or other mistakes may have just been innocuous errors.
Lex Fridman (1:00:01.100)
Can we talk about the COVID 19 a little bit more?
Manolis Kellis (1:00:08.700)
What's your sense about the genome, the proteins,
Lex Fridman (1:00:13.180)
the functions that we understand about COVID 19?
Lex Fridman (1:00:16.340)
Where do we stand in your sense?
Lex Fridman (1:00:18.900)
What are the big open problems?
Lex Fridman (1:00:21.420)
And also, you kind of said it's important to understand
Lex Fridman (1:00:25.380)
what are the important proteins
Lex Fridman (1:00:29.860)
and why is that important?
Lex Fridman (1:00:34.140)
So what else does the comparison of these species tell us?
Lex Fridman (1:00:39.420)
What it tells us is how fast are things evolving?
Lex Fridman (1:00:43.020)
It tells us about at what level is the acceleration
Manolis Kellis (1:00:46.620)
or deceleration pedal set for every one of these proteins.
Lex Fridman (1:00:50.820)
So the genome has 30 some genes.
Manolis Kellis (1:00:54.100)
Some genes evolve super, super fast.
Lex Fridman (1:00:56.580)
Others evolve super, super slow.
Manolis Kellis (1:00:59.020)
If you look at the polymerase gene
Lex Fridman (1:01:00.460)
that basically replicates the genome,
Manolis Kellis (1:01:01.980)
that's a super slow evolving one.
Lex Fridman (1:01:04.220)
If you look at the nucleocapsid protein,
Manolis Kellis (1:01:06.340)
that's also super slow evolving.
Lex Fridman (1:01:09.420)
If you look at the spike one protein,
Manolis Kellis (1:01:11.460)
this is the part of the spike protein
Lex Fridman (1:01:13.380)
that actually touches the ACE2 receptor
Lex Fridman (1:01:15.740)
and then enables the virus to attach to your cells.
Lex Fridman (1:01:21.300)
That's the thing that gives it that visual...
Manolis Kellis (1:01:23.820)
Yeah, the corona look basically.
Lex Fridman (1:01:24.860)
The corona look, yeah.
Lex Fridman (1:01:26.020)
So basically the spike protein sticks out of the virus
Lex Fridman (1:01:28.540)
and there's a first part of the protein S1
Manolis Kellis (1:01:31.180)
which basically attaches to the ACE2 receptor.
Lex Fridman (1:01:34.540)
And then S2 is the latch that sort of pushes and channels
Manolis Kellis (1:01:39.500)
the fusion of the membranes
Lex Fridman (1:01:41.060)
and then the incorporation of the viral RNA inside our cells
Manolis Kellis (1:01:47.060)
which then gets translated into all of these 30 proteins.
Lex Fridman (1:01:50.460)
So the S1 protein is evolving ridiculously fast.
Lex Fridman (1:01:55.460)
So if you look at the stop versus gas pedal,
Lex Fridman (1:01:59.460)
the gas pedal is all the way down.
Manolis Kellis (1:02:02.660)
ORF8 is also evolving super fast
Lex Fridman (1:02:05.140)
and ORF6 is evolving super fast.
Manolis Kellis (1:02:06.700)
We have no idea what they do.
Lex Fridman (1:02:08.020)
We have some idea but nowhere near what S1 is.
Lex Fridman (1:02:11.340)
So what the...
Lex Fridman (1:02:12.180)
Isn't that terrifying that S1 is evolving?
Manolis Kellis (1:02:14.220)
That means that's a really useful function
Lex Fridman (1:02:16.900)
and if it's evolving fast,
Manolis Kellis (1:02:18.780)
doesn't that mean new strains could be created
Lex Fridman (1:02:20.700)
or it does something?
Manolis Kellis (1:02:21.540)
That means that it's searching for how to match,
Lex Fridman (1:02:24.220)
how to best match the host.
Lex Fridman (1:02:26.700)
So basically anything in general in evolution,
Lex Fridman (1:02:29.260)
if you look at genomes,
Manolis Kellis (1:02:30.100)
anything that's contacting the environment
Lex Fridman (1:02:32.300)
is evolving much faster than anything that's internal.
Lex Fridman (1:02:34.980)
And the reason is that the environment changes.
Lex Fridman (1:02:37.140)
So if you look at the evolution of the cervical viruses,
Manolis Kellis (1:02:42.180)
the S1 protein has evolved very rapidly
Lex Fridman (1:02:44.420)
because it's attaching to different hosts each time.
Manolis Kellis (1:02:47.420)
We think of them as bats,
Lex Fridman (1:02:48.500)
but there's thousands of species of bats
Lex Fridman (1:02:50.540)
and to go from one species of bat to another species of bat,
Lex Fridman (1:02:52.900)
you have to adjust S1 to the new ACE2 receptor
Manolis Kellis (1:02:55.940)
that you're gonna be facing in that new species.
Lex Fridman (1:02:58.100)
Sorry, quick tangent.
Lex Fridman (1:02:59.740)
Is it fascinating to you that viruses are doing this?
Lex Fridman (1:03:03.540)
I mean, it feels like they're this intelligent organism.
Manolis Kellis (1:03:06.900)
I mean, does it give you pause how incredible it is
Lex Fridman (1:03:12.380)
that the evolutionary dynamics that you're describing
Manolis Kellis (1:03:16.500)
is actually happening and they're freaking out,
Lex Fridman (1:03:19.260)
figuring out how to jump from bats to humans
Lex Fridman (1:03:22.180)
all in this distributed fashion?
Lex Fridman (1:03:24.220)
And then most of us don't even say
Manolis Kellis (1:03:25.620)
they're alive or intelligent or whatever.
Lex Fridman (1:03:27.660)
So intelligence is in the eye of the beholder.
Manolis Kellis (1:03:31.340)
Stupid is as stupid does, as Forrest Gump would say,
Lex Fridman (1:03:34.940)
and intelligent is as intelligent does.
Lex Fridman (1:03:36.580)
So basically if the virus is finding solutions
Lex Fridman (1:03:39.220)
that we think of as intelligent,
Manolis Kellis (1:03:40.900)
yeah, it's probably intelligent,
Lex Fridman (1:03:42.260)
but that's again in the eye of the beholder.
Lex Fridman (1:03:43.940)
Do you think viruses are intelligent?
Lex Fridman (1:03:45.700)
Oh, of course not.
Lex Fridman (1:03:47.300)
Really?
Lex Fridman (1:03:48.140)
No.
Manolis Kellis (1:03:49.060)
It's so incredible.
Lex Fridman (1:03:50.180)
So remember when I was talking about the two components
Manolis Kellis (1:03:52.740)
of evolution, one is the stupid mutation,
Lex Fridman (1:03:55.900)
which is completely blind,
Lex Fridman (1:03:57.100)
and the other one is the super smart selection,
Lex Fridman (1:04:00.260)
which is ruthless.
Lex Fridman (1:04:01.820)
So it's not viruses who are smart.
Lex Fridman (1:04:04.780)
It's this component of evolution that's smart.
Lex Fridman (1:04:06.860)
So it's evolution that sort of appears smart.
Lex Fridman (1:04:10.380)
And how is that happening?
Manolis Kellis (1:04:12.020)
By huge parallel search across thousands of parallel
Lex Fridman (1:04:17.020)
of parallel infections throughout the world right now.
Manolis Kellis (1:04:21.700)
Yes, but so to push back on that,
Lex Fridman (1:04:23.980)
so yes, so then the intelligence is in the mechanism,
Lex Fridman (1:04:27.980)
but then by that argument,
Lex Fridman (1:04:31.380)
viruses would be more intelligent
Manolis Kellis (1:04:32.860)
because there's just more of them.
Lex Fridman (1:04:34.700)
So the search, they're basically the brute force search
Manolis Kellis (1:04:38.740)
that's happening with viruses
Lex Fridman (1:04:40.340)
because there's so many more of them than humans,
Manolis Kellis (1:04:43.220)
then they're taken as a whole are more intelligent.
Lex Fridman (1:04:47.540)
I mean, so you don't think it's possible that,
Manolis Kellis (1:04:51.260)
I mean, who runs, would we even be here if viruses weren't,
Lex Fridman (1:04:55.580)
I mean, who runs this thing?
Lex Fridman (1:04:58.380)
So humans or viruses?
Lex Fridman (1:04:59.700)
So let me answer, yeah, let me answer your question.
Lex Fridman (1:05:03.060)
So we would not be here if it wasn't for viruses.
Lex Fridman (1:05:10.460)
And part of the reason is that
Manolis Kellis (1:05:11.820)
if you look at mammalian evolution early on
Lex Fridman (1:05:14.340)
in this mammalian radiation
Manolis Kellis (1:05:16.100)
that basically happened after the death of the dinosaurs
Lex Fridman (1:05:18.580)
is that some of the viruses that we had in our genome
Manolis Kellis (1:05:22.740)
spread throughout our genome
Lex Fridman (1:05:24.580)
and created binding sites
Manolis Kellis (1:05:27.260)
for new classes of regulatory proteins.
Lex Fridman (1:05:30.340)
And these binding sites that landed all over our genome
Manolis Kellis (1:05:33.340)
are now control elements that basically control our genes
Lex Fridman (1:05:36.860)
and sort of help the complexity of the circuitry
Manolis Kellis (1:05:40.420)
of mammalian genomes.
Lex Fridman (1:05:42.220)
So, you know, everything's coevolution.
Manolis Kellis (1:05:45.100)
That's fascinating, we're working together.
Lex Fridman (1:05:47.780)
And yet you say they're dumb.
Manolis Kellis (1:05:48.620)
We've coopted them.
Lex Fridman (1:05:49.620)
No, I never said they're dumb.
Manolis Kellis (1:05:51.660)
They just don't care.
Lex Fridman (1:05:53.620)
They don't care.
Lex Fridman (1:05:54.980)
Another thing, oh, is the virus trying to kill us?
Lex Fridman (1:05:56.980)
No, it's not.
Manolis Kellis (1:05:58.020)
The virus is not trying to kill you.
Lex Fridman (1:05:59.980)
It's actually actively trying to not kill you.
Lex Fridman (1:06:02.820)
So when you get infected, if you die,
Lex Fridman (1:06:05.980)
bomber, I killed him,
Manolis Kellis (1:06:07.340)
is what the reaction of the virus will be.
Lex Fridman (1:06:09.140)
Why? Because that virus won't spread.
Manolis Kellis (1:06:12.060)
Many people have a misconception of,
Lex Fridman (1:06:13.780)
oh, viruses are smart or oh, viruses are mean.
Manolis Kellis (1:06:16.780)
They don't care.
Lex Fridman (1:06:18.620)
It's like, you have to clean yourself
Manolis Kellis (1:06:20.780)
of any kind of anthropomorphism out there.
Lex Fridman (1:06:23.300)
I don't know.
Manolis Kellis (1:06:24.140)
Oh, yes.
Lex Fridman (1:06:24.980)
So there's a sense when taken as a whole that there's...
Manolis Kellis (1:06:31.780)
It's in the eye of the beholder.
Lex Fridman (1:06:32.980)
Stupid is as stupid does.
Manolis Kellis (1:06:34.180)
Intelligent is as intelligent does.
Lex Fridman (1:06:35.940)
So if you want to call them intelligent, that's fine.
Manolis Kellis (1:06:38.460)
Because the end result is that
Lex Fridman (1:06:40.420)
they're finding amazing solutions.
Manolis Kellis (1:06:42.700)
I mean, I am in awe.
Lex Fridman (1:06:44.300)
They're so dumb about it.
Manolis Kellis (1:06:45.580)
They're just doing dumb.
Lex Fridman (1:06:46.420)
They don't care.
Manolis Kellis (1:06:47.260)
They're not dumb and they're just don't care.
Lex Fridman (1:06:48.580)
They don't care.
Manolis Kellis (1:06:50.060)
The care word is really interesting.
Lex Fridman (1:06:51.780)
I mean, there could be an argument that they're conscious.
Manolis Kellis (1:06:54.380)
They're just dividing.
Lex Fridman (1:06:55.460)
They're not.
Manolis Kellis (1:06:56.300)
They're just dividing.
Lex Fridman (1:06:57.620)
They're just a little entity
Manolis Kellis (1:07:00.020)
which happens to be dividing and spreading.
Lex Fridman (1:07:02.660)
It just doesn't want to kill us.
Manolis Kellis (1:07:04.420)
In fact, it prefers not to kill us.
Lex Fridman (1:07:06.340)
It just wants to spread.
Lex Fridman (1:07:07.580)
And when I say wants, again, I'm anthropomorphizing,
Lex Fridman (1:07:11.020)
but it's just that if you have two versions of a virus,
Manolis Kellis (1:07:15.060)
one acquires a mutation that spreads more,
Lex Fridman (1:07:17.420)
that's going to spread more.
Manolis Kellis (1:07:18.620)
One acquires a mutation that spreads less,
Lex Fridman (1:07:20.260)
that's going to be lost.
Manolis Kellis (1:07:21.740)
One acquires a mutation that enters faster,
Lex Fridman (1:07:24.020)
that's going to be kept.
Manolis Kellis (1:07:25.100)
One acquires a mutation that kills you right away,
Lex Fridman (1:07:27.020)
it's going to be lost.
Lex Fridman (1:07:28.420)
So over evolutionary time,
Lex Fridman (1:07:30.500)
the viruses that spread super well
Lex Fridman (1:07:32.700)
but don't kill the host
Lex Fridman (1:07:33.940)
are the ones that are going to survive.
Manolis Kellis (1:07:36.260)
Yeah, but so you brilliantly described
Lex Fridman (1:07:39.060)
the basic mechanisms of how it all happens,
Lex Fridman (1:07:41.060)
but when you zoom out and you see the entirety of viruses,
Lex Fridman (1:07:46.500)
maybe across different strains of viruses,
Manolis Kellis (1:07:49.980)
it seems like a living organism.
Lex Fridman (1:07:52.380)
I am in awe of biology.
Manolis Kellis (1:07:55.020)
I find biology amazingly beautiful.
Lex Fridman (1:07:58.340)
I find the design of the current coronavirus,
Manolis Kellis (1:08:01.100)
however lethal it is, amazingly beautiful.
Lex Fridman (1:08:04.260)
The way that it is encoded,
Manolis Kellis (1:08:06.340)
the way that it tricks your cells
Lex Fridman (1:08:08.980)
into making 30 proteins from a single RNA.
Manolis Kellis (1:08:12.340)
Human cells don't do that.
Lex Fridman (1:08:14.340)
Human cells make one protein from each RNA molecule.
Manolis Kellis (1:08:18.180)
They don't make two, they make one.
Lex Fridman (1:08:20.220)
We are hardwired to make only one protein
Manolis Kellis (1:08:22.340)
from every RNA molecule.
Lex Fridman (1:08:23.780)
And yet this virus goes in,
Manolis Kellis (1:08:25.700)
throws in a single messenger RNA.
Lex Fridman (1:08:28.700)
Just like any messenger RNA,
Manolis Kellis (1:08:29.980)
we have tens of thousands of messenger RNAs
Lex Fridman (1:08:32.140)
in our cells in any one time.
Manolis Kellis (1:08:34.140)
In every one of our cells.
Lex Fridman (1:08:35.820)
It throws in one RNA and that RNA is so,
Manolis Kellis (1:08:40.620)
I'm gonna use your word here, not my word, intelligent.
Lex Fridman (1:08:44.100)
That it hijacks the entire machinery of your human cell.
Manolis Kellis (1:08:49.300)
It basically has at the beginning,
Lex Fridman (1:08:52.500)
a giant open reading frame.
Manolis Kellis (1:08:54.460)
That's a giant protein that gets translated.
Lex Fridman (1:08:57.140)
Two thirds of that RNA make a single giant protein.
Manolis Kellis (1:09:01.540)
That single protein is basically
Lex Fridman (1:09:03.340)
what a human cell would make.
Manolis Kellis (1:09:04.540)
It's like, oh, here's a start code.
Lex Fridman (1:09:06.340)
I'm gonna start translating here.
Manolis Kellis (1:09:07.700)
Human cells are kind of dumb.
Lex Fridman (1:09:08.620)
I'm sorry.
Manolis Kellis (1:09:09.460)
Again, this is not the words I would normally use.
Lex Fridman (1:09:12.420)
But the human cell basically says,
Manolis Kellis (1:09:13.420)
oh, this is an RNA, must be mine.
Lex Fridman (1:09:15.020)
Let me translate.
Lex Fridman (1:09:15.860)
And it starts translating it.
Lex Fridman (1:09:17.020)
And then you're in trouble.
Lex Fridman (1:09:18.460)
Why?
Lex Fridman (1:09:19.300)
Because that one protein as it's growing,
Manolis Kellis (1:09:22.380)
gets cleaved into about 20 different peptides.
Lex Fridman (1:09:26.980)
The first peptide and the second peptide start interacting
Lex Fridman (1:09:30.940)
and the third one and the fourth one.
Lex Fridman (1:09:32.460)
And they shut off the ribosome of the whole cell
Manolis Kellis (1:09:37.820)
to not translate human RNAs anymore.
Lex Fridman (1:09:42.820)
So the virus basically hijacks your cells
Lex Fridman (1:09:46.580)
and it cuts, it cleaves every one of your human RNAs
Lex Fridman (1:09:50.820)
to basically say to the ribosome,
Manolis Kellis (1:09:52.100)
don't translate this one, junk.
Lex Fridman (1:09:53.500)
Don't look at this one, junk.
Lex Fridman (1:09:55.140)
And it only spares its own RNAs
Lex Fridman (1:09:58.740)
because they have a particular mark that it spares.
Manolis Kellis (1:10:01.300)
Then all of the ribosomes that normally make protein
Lex Fridman (1:10:04.380)
in your human cells are now only able
Manolis Kellis (1:10:06.860)
to translate viral RNAs.
Lex Fridman (1:10:09.220)
And then more and more and more and more of them.
Manolis Kellis (1:10:11.460)
That's the first 20 proteins.
Lex Fridman (1:10:13.100)
In fact, halfway down about protein 11,
Manolis Kellis (1:10:16.300)
between 11 and 12,
Lex Fridman (1:10:17.940)
you basically have a translational slippage
Manolis Kellis (1:10:21.020)
where the ribosome skips reading frame.
Lex Fridman (1:10:23.420)
And it translates from one reading frame
Manolis Kellis (1:10:24.980)
to another reading frame.
Lex Fridman (1:10:25.820)
That means that about half of them
Manolis Kellis (1:10:27.100)
are gonna be translated from one to 11.
Lex Fridman (1:10:29.260)
And the other half are gonna be translated
Manolis Kellis (1:10:30.900)
from 12 to 16.
Lex Fridman (1:10:32.700)
It's gorgeous.
Lex Fridman (1:10:34.260)
And then you're done.
Lex Fridman (1:10:37.380)
Then that mRNA will never translate the last 10 proteins
Lex Fridman (1:10:40.420)
but spike is the one right after that one.
Lex Fridman (1:10:42.540)
So how does spike even get translated?
Manolis Kellis (1:10:45.140)
This positive strand RNA virus has a reverse transcriptase
Lex Fridman (1:10:50.020)
which is an RNA based reverse transcriptase.
Lex Fridman (1:10:52.340)
So from the RNA on the positive strand,
Lex Fridman (1:10:54.460)
it makes an RNA on the negative strand.
Lex Fridman (1:10:56.940)
And in between every single one of these genes,
Lex Fridman (1:10:59.620)
these open reading frames,
Manolis Kellis (1:11:01.060)
there's a little signal AACGCA or something like that,
Lex Fridman (1:11:05.580)
that basically loops over to the beginning of the RNA.
Lex Fridman (1:11:09.940)
And basically instead of sort of having
Lex Fridman (1:11:11.700)
a single full negative strand RNA,
Manolis Kellis (1:11:14.500)
it basically has a partial negative strand RNA
Lex Fridman (1:11:16.860)
that ends right before the beginning of that gene.
Lex Fridman (1:11:19.700)
And another one that ends right before
Lex Fridman (1:11:20.860)
the beginning of that gene.
Manolis Kellis (1:11:21.980)
These negative strand RNAs now make positive strand RNAs
Lex Fridman (1:11:25.340)
that then look to the human whole cell
Manolis Kellis (1:11:27.460)
just like any other human mRNA.
Lex Fridman (1:11:29.780)
It's like, ooh, great, I'm gonna translate that one
Manolis Kellis (1:11:31.780)
because it doesn't have the cleaving
Lex Fridman (1:11:32.980)
that the virus has now put on all your human genes.
Lex Fridman (1:11:36.340)
And then you've lost the battle.
Lex Fridman (1:11:38.300)
That cell is now only making proteins for the virus
Manolis Kellis (1:11:42.500)
that will then create the spike protein,
Lex Fridman (1:11:45.500)
the envelope protein, the membrane protein,
Manolis Kellis (1:11:47.620)
the nucleocapsid protein that will package up the RNA
Lex Fridman (1:11:50.380)
and then sort of create new viral envelopes.
Lex Fridman (1:11:53.820)
And these will then be secreted out of that cell
Lex Fridman (1:11:57.780)
in new little packages
Manolis Kellis (1:11:59.260)
that will then infect the rest of the cells.
Lex Fridman (1:12:00.580)
Repeat the whole process again.
Lex Fridman (1:12:01.980)
It's beautiful, right?
Lex Fridman (1:12:03.300)
It's mind boggling.
Manolis Kellis (1:12:04.140)
It's hard not to anthropomorphize it.
Lex Fridman (1:12:05.620)
I know, but it's so gorgeous.
Lex Fridman (1:12:08.100)
So there is a beauty to it.
Lex Fridman (1:12:09.900)
Of course.
Lex Fridman (1:12:12.140)
Is it terrifying to you?
Lex Fridman (1:12:13.980)
So this is something that has happened throughout history.
Manolis Kellis (1:12:16.940)
Humans have been nearly wiped out
Lex Fridman (1:12:19.700)
over and over and over again,
Lex Fridman (1:12:21.140)
and yet never fully wiped out.
Lex Fridman (1:12:23.260)
So yeah, I'm not concerned about the human race.
Manolis Kellis (1:12:25.940)
I'm not even concerned about the impact
Lex Fridman (1:12:29.340)
on sort of our survival as a species.
Manolis Kellis (1:12:33.780)
This is absolutely something,
Lex Fridman (1:12:35.620)
I mean, human life is so invaluable
Lex Fridman (1:12:38.660)
and every one of us is so invaluable,
Lex Fridman (1:12:40.060)
but if you think of it as sort of,
Lex Fridman (1:12:42.260)
is this the end of our species?
Lex Fridman (1:12:44.100)
By no means, basically.
Lex Fridman (1:12:46.420)
So let me explain.
Lex Fridman (1:12:48.100)
The Black Death killed what, 30% of Europe?
Manolis Kellis (1:12:51.260)
That has left a tremendous imprint,
Lex Fridman (1:12:55.260)
a huge hole, a horrendous hole
Manolis Kellis (1:12:59.260)
in the genetic makeup of humans.
Lex Fridman (1:13:03.620)
There's been series of wiping out of huge fractions
Manolis Kellis (1:13:08.620)
of entire species or just entire species altogether.
Lex Fridman (1:13:12.180)
And that has a consequence on the human immune repertoire.
Manolis Kellis (1:13:17.180)
If you look at how Europe was shaped
Lex Fridman (1:13:20.860)
and how Africa was shaped by malaria, for example,
Manolis Kellis (1:13:24.700)
all the individuals that carry a mutation
Lex Fridman (1:13:26.940)
that protects you from malaria
Manolis Kellis (1:13:29.220)
were able to survive much more.
Lex Fridman (1:13:31.100)
And if you look at the frequency of sickle cell disease
Lex Fridman (1:13:33.820)
and the frequency of malaria,
Lex Fridman (1:13:35.580)
the maps are actually showing the same pattern,
Manolis Kellis (1:13:38.340)
the same imprint on Africa.
Lex Fridman (1:13:40.420)
And that basically led people to hypothesize
Manolis Kellis (1:13:42.340)
that the reason why sickle cell disease
Lex Fridman (1:13:43.860)
is so much more frequent is because
Manolis Kellis (1:13:45.660)
sickle cell disease is so much more frequent
Lex Fridman (1:13:47.620)
in Americans of African descent
Manolis Kellis (1:13:50.060)
is because there was selection in Africa against malaria
Lex Fridman (1:13:55.260)
leading to sickle cell, because when the cells sickle,
Manolis Kellis (1:13:57.780)
malaria is not able to replicate inside your cells as well.
Lex Fridman (1:14:01.380)
And therefore you protect against that.
Lex Fridman (1:14:03.060)
So if you look at human disease,
Lex Fridman (1:14:05.380)
all of the genetic associations that we do
Manolis Kellis (1:14:07.540)
with human disease,
Lex Fridman (1:14:09.220)
you basically see the imprint
Manolis Kellis (1:14:13.620)
of these waves of selection killing off
Lex Fridman (1:14:16.620)
gazillions of humans.
Lex Fridman (1:14:18.300)
And there's so many immune processes that are coming up
Lex Fridman (1:14:23.180)
as associated with so many different diseases.
Manolis Kellis (1:14:25.900)
The reason for that is similar
Lex Fridman (1:14:27.500)
to what I was describing earlier,
Manolis Kellis (1:14:28.620)
where the outward facing proteins evolve much more rapidly
Lex Fridman (1:14:33.620)
because the environment is always changing.
Lex Fridman (1:14:35.860)
But what's really interesting in the human genome
Lex Fridman (1:14:37.620)
is that we have coopted many of these immune genes
Manolis Kellis (1:14:40.340)
to carry out nonimmune functions.
Lex Fridman (1:14:42.420)
For example, in our brain,
Manolis Kellis (1:14:43.980)
we use immune cells to cleave off neuronal connections
Lex Fridman (1:14:48.860)
that don't get used.
Manolis Kellis (1:14:50.140)
This whole use it or lose it, we know the mechanism.
Lex Fridman (1:14:52.820)
It's microglia that cleave off neuronal synaptic connections
Manolis Kellis (1:14:57.820)
that are just not utilized.
Lex Fridman (1:14:59.900)
When you utilize them, you mark them in a particular way
Manolis Kellis (1:15:02.020)
that basically when the microglia come,
Lex Fridman (1:15:04.380)
tell it, don't kill this one, it's used now.
Lex Fridman (1:15:07.820)
And the microglia will go off
Lex Fridman (1:15:08.940)
and kill the ones you don't use.
Manolis Kellis (1:15:10.340)
This is an immune function,
Lex Fridman (1:15:12.780)
which is coopted to do nonimmune things.
Manolis Kellis (1:15:14.980)
If you look at our adipocytes,
Lex Fridman (1:15:16.780)
M1 versus M2 macrophages inside our fat
Manolis Kellis (1:15:19.900)
will basically determine whether you're obese or not.
Lex Fridman (1:15:22.620)
And these are again, immune cells that are resident
Lex Fridman (1:15:24.740)
and living within these tissues.
Lex Fridman (1:15:27.060)
So many disease associations.
Manolis Kellis (1:15:30.220)
That's it, that we coopt these kinds of things
Lex Fridman (1:15:33.660)
for incredibly complicated functions.
Manolis Kellis (1:15:36.660)
Exactly, evolution works in so many different ways,
Lex Fridman (1:15:39.860)
which are all beautiful and mysterious.
Lex Fridman (1:15:41.980)
But not intelligent.
Lex Fridman (1:15:43.340)
Not intelligent, it's in the eye of the beholder.
Lex Fridman (1:15:45.740)
But the point that I'm trying to make is that
Lex Fridman (1:15:51.060)
if you look at the imprint that COVID will have,
Manolis Kellis (1:15:54.260)
hopefully it will not be big.
Lex Fridman (1:15:55.980)
Hopefully the US will get attacked together
Lex Fridman (1:15:57.980)
and stop the virus from spreading further.
Lex Fridman (1:16:00.420)
But if it doesn't, it's having an imprint
Manolis Kellis (1:16:03.500)
on individuals who have particular genetic repertoires.
Lex Fridman (1:16:07.340)
So if you look at now the genetic associations
Manolis Kellis (1:16:10.060)
of blood type and immune function cells, et cetera,
Lex Fridman (1:16:13.620)
there's actually association, genetic variation
Manolis Kellis (1:16:15.740)
that basically says how much more likely am I or you to die
Lex Fridman (1:16:18.540)
if we contact the virus.
Lex Fridman (1:16:20.220)
And it's through these rounds of shaping the human genome
Lex Fridman (1:16:24.540)
that humans have basically made it so far.
Lex Fridman (1:16:27.380)
And selection is ruthless and it's brutal
Lex Fridman (1:16:32.620)
and it only comes with a lot of killing.
Lex Fridman (1:16:34.380)
But this is the way that viruses and environments
Lex Fridman (1:16:38.140)
have shaped the human genome.
Manolis Kellis (1:16:39.540)
Basically, when you go through periods of famine,
Lex Fridman (1:16:41.420)
you select for particular genes.
Lex Fridman (1:16:43.660)
And what's left is not necessarily better,
Lex Fridman (1:16:46.540)
it's just whatever survived.
Lex Fridman (1:16:49.020)
And it might have been the surviving one back then,
Lex Fridman (1:16:51.980)
not because it was better,
Manolis Kellis (1:16:53.140)
maybe the ones that ran slower survived.
Lex Fridman (1:16:54.980)
I mean, again, not necessarily better,
Lex Fridman (1:16:57.420)
but the surviving ones are basically the ones
Lex Fridman (1:17:00.020)
that then are shaped for any kind
Manolis Kellis (1:17:02.420)
of subsequent evolutionary condition
Lex Fridman (1:17:05.420)
and environmental condition.
Lex Fridman (1:17:07.260)
But if you look at, for example, obesity,
Lex Fridman (1:17:09.580)
obesity was selected for basically the genes
Manolis Kellis (1:17:12.420)
that now predisposes to obesity
Lex Fridman (1:17:14.420)
were at 2% frequency in Africa.
Manolis Kellis (1:17:16.660)
They rose to 44% frequency in Europe.
Lex Fridman (1:17:19.020)
Wow, that's fascinating.
Manolis Kellis (1:17:20.300)
Because you basically went through the ice ages
Lex Fridman (1:17:22.940)
and there was a scarcity of food.
Lex Fridman (1:17:24.620)
So there was a selection to being able to store
Lex Fridman (1:17:27.140)
every single calorie you consume.
Manolis Kellis (1:17:29.260)
Eventually, environment changes.
Lex Fridman (1:17:31.860)
So the better allele, which was the fat storing allele,
Manolis Kellis (1:17:35.140)
became the worst allele
Lex Fridman (1:17:36.500)
because it's the fat storing allele.
Manolis Kellis (1:17:38.780)
It still has the same function.
Lex Fridman (1:17:40.940)
So if you look at my genome, speaking of mom calling,
Manolis Kellis (1:17:44.100)
mom gave me a bad copy of that gene, this FTO locus.
Lex Fridman (1:17:48.500)
Basically, makes me.
Manolis Kellis (1:17:49.340)
The one that has to do with.
Lex Fridman (1:17:50.460)
Obesity.
Manolis Kellis (1:17:51.300)
With obesity.
Lex Fridman (1:17:52.140)
Yeah, I basically now have a bad copy from mom
Manolis Kellis (1:17:54.500)
that makes me more likely to be obese.
Lex Fridman (1:17:56.340)
And I also have a bad copy from dad
Manolis Kellis (1:17:59.180)
that makes me more likely to be obese.
Lex Fridman (1:18:00.020)
So homozygous.
Lex Fridman (1:18:01.860)
And that's the allele, it's still the minor allele,
Lex Fridman (1:18:05.860)
but it's at 44% frequency in Southeast Asia,
Manolis Kellis (1:18:09.140)
42% frequency in Europe, even though it started at 2%.
Lex Fridman (1:18:12.740)
It was an awesome allele to have 100 years ago.
Manolis Kellis (1:18:16.060)
Right now, it's pretty terrible allele.
Lex Fridman (1:18:17.900)
So the other concept is that diversity matters.
Manolis Kellis (1:18:21.980)
If we had 100 million nuclear physicists
Lex Fridman (1:18:25.660)
living the earth right now, we'd be in trouble.
Manolis Kellis (1:18:28.420)
You need diversity, you need artists
Lex Fridman (1:18:31.820)
and you need musicians and you need mathematicians
Lex Fridman (1:18:33.980)
and you need politicians, yes, even those.
Lex Fridman (1:18:37.100)
And you need like.
Manolis Kellis (1:18:37.940)
Well, let's not get crazy.
Lex Fridman (1:18:39.580)
But because then if a virus comes along or whatever.
Manolis Kellis (1:18:43.100)
Exactly, exactly.
Lex Fridman (1:18:44.900)
So, no, there's two reasons.
Manolis Kellis (1:18:45.980)
Number one, you want diversity in the immune repertoire
Lex Fridman (1:18:48.820)
and we have built in diversity.
Lex Fridman (1:18:50.820)
So basically, they are the most diverse.
Lex Fridman (1:18:53.380)
Basically, if you look at our immune system,
Manolis Kellis (1:18:54.900)
there's layers and layers of diversity.
Lex Fridman (1:18:57.100)
Like the way that you create your cells generates diversity
Manolis Kellis (1:19:01.540)
because of the selection for the VDJ recombination
Lex Fridman (1:19:04.820)
that basically eventually leads
Manolis Kellis (1:19:06.580)
to a huge number of repertoires.
Lex Fridman (1:19:08.140)
But that's only one small component of diversity.
Manolis Kellis (1:19:10.220)
The blood type is another one.
Lex Fridman (1:19:11.540)
The major histocompatibility complex, the HLA alleles
Manolis Kellis (1:19:15.900)
are another source of diversity.
Lex Fridman (1:19:18.020)
So the immune system of humans is by nature,
Manolis Kellis (1:19:21.660)
incredibly diverse and that basically leads to resilience.
Lex Fridman (1:19:25.580)
So basically what I'm saying that I don't worry
Manolis Kellis (1:19:27.460)
for the human species because we are so diverse immunologically,
Lex Fridman (1:19:32.460)
we are likely to be very resilient
Manolis Kellis (1:19:34.860)
against so many different attacks like this current virus.
Lex Fridman (1:19:39.020)
So you're saying natural pandemics may not be something
Manolis Kellis (1:19:42.100)
that you're really afraid of because of the diversity
Lex Fridman (1:19:45.180)
in our genetic makeup.
Lex Fridman (1:19:48.380)
What about engineered pandemics?
Lex Fridman (1:19:50.380)
Do you have fears of us messing with the makeup of viruses
Manolis Kellis (1:19:55.740)
or well, yeah, let's say with the makeup of viruses
Lex Fridman (1:19:58.860)
to create something that we can't control
Lex Fridman (1:20:00.820)
and would be much more destructive
Lex Fridman (1:20:02.700)
than it would come about naturally?
Lex Fridman (1:20:05.300)
Remember how we were talking about how smart evolution is?
Lex Fridman (1:20:08.020)
Humans are much dumber.
Manolis Kellis (1:20:09.300)
So.
Lex Fridman (1:20:10.140)
You mean like human scientists, engineers?
Manolis Kellis (1:20:11.860)
Yeah, humans, humans just like.
Lex Fridman (1:20:13.180)
Humans overall?
Manolis Kellis (1:20:14.020)
Yeah, humans overall.
Lex Fridman (1:20:14.860)
Okay.
Lex Fridman (1:20:15.700)
But I mean, even the sort of synthetic biologists
Lex Fridman (1:20:19.660)
you know, basically if you were to create,
Manolis Kellis (1:20:25.700)
you know, virus like SARS that will kill a lot of people,
Lex Fridman (1:20:29.700)
you would probably start with SARS.
Lex Fridman (1:20:32.860)
So whoever, you know, would like to design such a thing
Lex Fridman (1:20:37.460)
would basically start with a SARS tree
Manolis Kellis (1:20:39.980)
or at least some relative of SARS.
Lex Fridman (1:20:42.460)
The source genome for the current virus
Manolis Kellis (1:20:45.580)
was something completely different.
Lex Fridman (1:20:47.140)
It was something that has never infected anyone
Lex Fridman (1:20:49.100)
and never infected humans.
Lex Fridman (1:20:50.620)
No one in their right mind would have started there.
Lex Fridman (1:20:52.900)
But when you say sources like the nearest.
Lex Fridman (1:20:55.020)
The nearest relative.
Manolis Kellis (1:20:56.260)
Relative.
Lex Fridman (1:20:57.100)
Is in a whole other branch.
Manolis Kellis (1:20:58.420)
Interesting.
Lex Fridman (1:20:59.260)
No species of which has ever infected humans
Manolis Kellis (1:21:00.980)
in that branch.
Lex Fridman (1:21:02.580)
So, you know, let's put this to rest.
Manolis Kellis (1:21:05.340)
This was not designed by someone to kill off the human race.
Lex Fridman (1:21:08.580)
So you don't believe it was engineered?
Manolis Kellis (1:21:12.020)
The. Or likely.
Lex Fridman (1:21:13.100)
Yeah, the path to engineering a deadly virus
Manolis Kellis (1:21:16.140)
did not come from this strain that was used.
Lex Fridman (1:21:21.220)
Moreover, there's been various claims of,
Manolis Kellis (1:21:26.940)
ha ha, this was mixed and matched in lab
Lex Fridman (1:21:29.300)
because the S1 protein has three different components,
Manolis Kellis (1:21:32.580)
each of which has a different evolutionary tree.
Lex Fridman (1:21:34.700)
So, you know, a lot of popular press basically said,
Manolis Kellis (1:21:37.300)
aha, this came from pangolin
Lex Fridman (1:21:39.260)
and this came from, you know, all kinds of other species.
Manolis Kellis (1:21:42.980)
This is what has been happening
Lex Fridman (1:21:44.900)
throughout the coronavirus tree.
Lex Fridman (1:21:46.900)
So basically the S1 protein has been recombining
Lex Fridman (1:21:49.380)
across species all the time.
Manolis Kellis (1:21:50.420)
Remember when I was talking about the positive strand,
Lex Fridman (1:21:52.020)
the negative strand, sub genomic RNAs,
Manolis Kellis (1:21:54.340)
these can actually recombine.
Lex Fridman (1:21:55.780)
And if you have two different viruses
Manolis Kellis (1:21:57.140)
infecting the same cell,
Lex Fridman (1:21:58.540)
they can actually mix and match
Manolis Kellis (1:21:59.780)
between the positive strand and the negative strand
Lex Fridman (1:22:01.340)
and basically create a new hybrid virus with recombination
Manolis Kellis (1:22:04.700)
that now has the S1 from one
Lex Fridman (1:22:06.700)
and the rest of the genome from another.
Lex Fridman (1:22:08.780)
And this is something that happens a lot in S1,
Lex Fridman (1:22:10.580)
in Orfet, et cetera.
Lex Fridman (1:22:12.060)
And that's something that's true of the whole tree.
Lex Fridman (1:22:13.940)
For the whole family of viruses.
Lex Fridman (1:22:15.780)
So it's not like someone has been messing with this
Lex Fridman (1:22:18.020)
for millions of years and, you know, changing.
Manolis Kellis (1:22:20.860)
This happens naturally.
Lex Fridman (1:22:21.700)
That's, again, beautiful that that somehow happens,
Manolis Kellis (1:22:24.420)
that they recombine.
Lex Fridman (1:22:25.900)
So two different strands can infect the body
Lex Fridman (1:22:27.740)
and then recombine.
Lex Fridman (1:22:30.460)
So all of this actually magic happens inside hosts.
Manolis Kellis (1:22:35.100)
Like all, like.
Lex Fridman (1:22:36.300)
Yeah, that's why classification wise,
Manolis Kellis (1:22:39.220)
virus is not thought to be alive
Lex Fridman (1:22:40.700)
because it doesn't self replicate.
Manolis Kellis (1:22:41.940)
It's not autonomous.
Lex Fridman (1:22:43.020)
It's something that enters a living cell
Lex Fridman (1:22:45.740)
and then co ops it to basically make it its own.
Lex Fridman (1:22:48.740)
But by itself, people ask me,
Lex Fridman (1:22:50.660)
how do we kill this bastard?
Lex Fridman (1:22:51.580)
I'm like, you stop it from replicating.
Manolis Kellis (1:22:54.180)
It's not like a bacterium that will just live
Lex Fridman (1:22:57.660)
in a, you know, puddle or something.
Manolis Kellis (1:23:01.060)
It's a virus.
Lex Fridman (1:23:02.460)
Viruses don't live without their host.
Lex Fridman (1:23:04.420)
And they only live with their host for very little time.
Lex Fridman (1:23:07.380)
So if you stop it from replicating,
Manolis Kellis (1:23:09.100)
it'll stop from spreading.
Lex Fridman (1:23:11.260)
I mean, it's not like HIV, which can stay dormant
Manolis Kellis (1:23:13.220)
for a long time.
Lex Fridman (1:23:14.060)
Basically, coronaviruses just don't do that.
Manolis Kellis (1:23:15.580)
They're not integrating genomes.
Lex Fridman (1:23:16.780)
They're RNA genomes.
Lex Fridman (1:23:18.060)
So if it's not expressed, it degrades.
Lex Fridman (1:23:20.220)
RNA degrades.
Manolis Kellis (1:23:21.180)
It doesn't just stick around.
Lex Fridman (1:23:23.380)
Well, let me ask also about the immune system you mentioned.
Manolis Kellis (1:23:27.340)
A lot of people kind of ask, you know,
Lex Fridman (1:23:31.500)
how can we strengthen the immune system
Manolis Kellis (1:23:34.140)
to respond to this particular virus,
Lex Fridman (1:23:36.300)
but the viruses in general.
Lex Fridman (1:23:37.740)
Do you have from a biological perspective,
Lex Fridman (1:23:40.420)
thoughts on what we can do as humans
Lex Fridman (1:23:43.140)
to strengthen our immune system?
Lex Fridman (1:23:43.980)
If you look at the death rates across different countries,
Manolis Kellis (1:23:46.620)
people with less vaccination have been dying more.
Lex Fridman (1:23:49.700)
If you look at North Italy,
Manolis Kellis (1:23:51.380)
the vaccination rates are abysmal there.
Lex Fridman (1:23:53.940)
And a lot of people have been dying.
Manolis Kellis (1:23:55.860)
If you look at Greece, very good vaccination rates.
Lex Fridman (1:23:58.780)
Almost no one has been dying.
Lex Fridman (1:24:00.300)
So yes, there's a policy component.
Lex Fridman (1:24:03.580)
So Italy reacted very slowly.
Manolis Kellis (1:24:05.980)
Greece reacted very fast.
Lex Fridman (1:24:07.460)
So yeah, many fewer people died in Greece,
Lex Fridman (1:24:09.780)
but there might actually be a component
Lex Fridman (1:24:11.740)
of genetic immune repertoire.
Manolis Kellis (1:24:14.100)
Basically, how did people die off, you know,
Lex Fridman (1:24:16.700)
in the history of the Greek population
Manolis Kellis (1:24:19.020)
versus the Italian population.
Lex Fridman (1:24:20.740)
Wow. There's a...
Manolis Kellis (1:24:22.300)
That's interesting to think about.
Lex Fridman (1:24:24.580)
And then there's a component
Manolis Kellis (1:24:25.980)
of what vaccinations did you have as a kid
Lex Fridman (1:24:28.940)
and what are the off target effects of those vaccinations?
Lex Fridman (1:24:32.460)
So basically a vaccination can have two components.
Lex Fridman (1:24:34.900)
One is training your immune system
Manolis Kellis (1:24:37.620)
against that specific insult.
Lex Fridman (1:24:39.500)
The second one is boosting up your immune system
Manolis Kellis (1:24:42.140)
for all kinds of other things.
Lex Fridman (1:24:44.580)
If you look at allergies,
Manolis Kellis (1:24:47.100)
Northern Europe, super clean environments,
Lex Fridman (1:24:50.220)
tons of allergies.
Manolis Kellis (1:24:51.420)
Southern Europe, my kids grew up eating dirt.
Lex Fridman (1:24:54.980)
No allergies.
Lex Fridman (1:24:57.060)
So growing up, I never had even heard of what allergies are.
Lex Fridman (1:25:00.420)
Like, was it really allergies?
Lex Fridman (1:25:01.940)
And the reason is that I was playing in the garden.
Lex Fridman (1:25:03.580)
I was putting all kinds of stuff in my mouth from,
Manolis Kellis (1:25:05.940)
you know, all kinds of dirt and stuff,
Lex Fridman (1:25:07.380)
tons of viruses there, tons of bacteria there.
Manolis Kellis (1:25:09.620)
You know, my immune system was built up.
Lex Fridman (1:25:11.500)
So the more you protect your immune system from exposure,
Manolis Kellis (1:25:16.700)
the less opportunity it has to learn
Lex Fridman (1:25:18.820)
about non self repertoire in a way that prepares it
Manolis Kellis (1:25:23.100)
for the next insult.
Lex Fridman (1:25:24.380)
So that's the horizontal thing too,
Manolis Kellis (1:25:25.860)
like the, so it's throughout your lifetime
Lex Fridman (1:25:28.100)
and the lifetime of the people that, your ancestors,
Manolis Kellis (1:25:33.220)
that kind of thing.
Lex Fridman (1:25:34.060)
What about the...
Lex Fridman (1:25:35.060)
So again, it returns against free will.
Lex Fridman (1:25:37.900)
On the free will side of things,
Manolis Kellis (1:25:39.540)
is there something we could do
Lex Fridman (1:25:40.980)
to strengthen our immune system in 2020?
Manolis Kellis (1:25:44.780)
Is there like, you know, exercise, diet,
Lex Fridman (1:25:49.100)
all that kind of stuff?
Lex Fridman (1:25:50.700)
So it's kind of funny.
Lex Fridman (1:25:52.900)
There's a cartoon that basically shows two windows
Manolis Kellis (1:25:55.940)
with a teller in each window.
Lex Fridman (1:25:58.300)
One has a humongous line and the other one has no one.
Manolis Kellis (1:26:02.220)
The one that has no one above says health.
Lex Fridman (1:26:04.700)
No, it says exercise and diet.
Lex Fridman (1:26:07.220)
And the other one says pill.
Lex Fridman (1:26:10.300)
And there's a huge line for pill.
Lex Fridman (1:26:12.140)
So we're looking basically for magic bullets
Lex Fridman (1:26:13.940)
for sort of ways that we can, you know,
Manolis Kellis (1:26:16.860)
beat cancer and beat coronavirus and beat this
Lex Fridman (1:26:18.980)
and beat that.
Lex Fridman (1:26:19.820)
And it turns out that the window with like,
Lex Fridman (1:26:21.420)
just diet and exercise is the best way
Manolis Kellis (1:26:23.980)
to boost every aspect of your health.
Lex Fridman (1:26:26.100)
If you look at Alzheimer's, exercise and nutrition.
Lex Fridman (1:26:31.220)
I mean, you're like, really?
Lex Fridman (1:26:32.580)
For my brain, neurodegeneration?
Manolis Kellis (1:26:34.700)
Absolutely.
Lex Fridman (1:26:36.140)
If you look at cancer, exercise and nutrition.
Manolis Kellis (1:26:40.420)
If you look at coronavirus, exercise and nutrition,
Lex Fridman (1:26:43.780)
every single aspect of human health gets improved.
Lex Fridman (1:26:47.300)
And one of the studies we're doing now
Lex Fridman (1:26:48.620)
is basically looking at what are the effects
Lex Fridman (1:26:51.260)
of diet and exercise?
Lex Fridman (1:26:52.940)
How similar are they to each other?
Manolis Kellis (1:26:55.340)
We basically take in diet intervention
Lex Fridman (1:26:58.220)
and exercise intervention in human and in mice.
Lex Fridman (1:27:01.380)
And we're basically doing single cell profiling
Lex Fridman (1:27:03.580)
of a bunch of different tissues
Manolis Kellis (1:27:04.980)
to basically understand how are the cells,
Lex Fridman (1:27:08.020)
both the stromal cells and the immune cells
Manolis Kellis (1:27:10.900)
of each of these tissues responding
Lex Fridman (1:27:13.260)
to the effect of exercise.
Lex Fridman (1:27:15.100)
What are the communication networks
Lex Fridman (1:27:16.900)
between different cells?
Manolis Kellis (1:27:18.540)
Where the muscle that exercises sends signals
Lex Fridman (1:27:23.900)
through the bloodstream, through the lymphatic system,
Manolis Kellis (1:27:25.980)
through all kinds of other systems
Lex Fridman (1:27:27.660)
that give signals to other cells that I have exercised
Lex Fridman (1:27:31.580)
and you should change in this particular way,
Lex Fridman (1:27:33.980)
which basically reconfigure those receptor cells
Manolis Kellis (1:27:37.620)
with the effect of exercise.
Lex Fridman (1:27:39.860)
How well understood is those reconfigurations?
Manolis Kellis (1:27:43.860)
Very little.
Lex Fridman (1:27:44.700)
We're just starting now, basically.
Manolis Kellis (1:27:46.940)
Is the hope there to understand the effect on,
Lex Fridman (1:27:52.420)
so like the effect on the immune system?
Manolis Kellis (1:27:54.300)
On the immune system, the effect on brain,
Lex Fridman (1:27:56.220)
the effect on your liver, on your digestive system,
Lex Fridman (1:27:59.060)
on your adipocytes?
Lex Fridman (1:28:00.900)
Adipose, the most misunderstood organ.
Manolis Kellis (1:28:03.620)
Everybody thinks, oh, fat, terrible.
Lex Fridman (1:28:05.780)
No, fat is awesome.
Manolis Kellis (1:28:07.460)
Your fat cells is what's keeping you alive
Lex Fridman (1:28:09.940)
because if you didn't have your fat cells,
Manolis Kellis (1:28:11.420)
all those lipids and all those calories
Lex Fridman (1:28:13.940)
would be floating around in your blood
Lex Fridman (1:28:15.420)
and you'd be dead by now.
Lex Fridman (1:28:16.940)
Your adipocytes are your best friend.
Manolis Kellis (1:28:18.460)
They're basically storing all these excess calories
Lex Fridman (1:28:21.820)
so that they don't hurt all of the rest of the body.
Lex Fridman (1:28:24.900)
And they're also fat burning in many ways.
Lex Fridman (1:28:28.940)
So, again, when you don't have
Manolis Kellis (1:28:31.540)
the homozygous version that I have,
Lex Fridman (1:28:33.460)
your cells are able to burn calories much more easily
Manolis Kellis (1:28:36.500)
by sort of flipping a master metabolic switch
Lex Fridman (1:28:39.980)
that involves this FTO locus that I mentioned earlier
Lex Fridman (1:28:42.380)
and its target genes, RX3 and RX5,
Lex Fridman (1:28:45.060)
that basically switch your adipocytes
Manolis Kellis (1:28:47.540)
during their three first days of differentiation
Lex Fridman (1:28:50.780)
as they're becoming mature adipocytes
Manolis Kellis (1:28:52.300)
to basically become either fat burning
Lex Fridman (1:28:54.340)
or fat storing fat cells.
Lex Fridman (1:28:57.100)
And the fat burning fat cells are your best friend.
Lex Fridman (1:28:58.980)
They're much closer to muscle
Manolis Kellis (1:29:00.460)
than they are to white adipocytes.
Lex Fridman (1:29:02.820)
Is there a lot of difference between people
Manolis Kellis (1:29:05.340)
that you could give, science could eventually give advice
Lex Fridman (1:29:09.540)
that is very generalizable
Manolis Kellis (1:29:12.260)
or is our differences in our genetic makeup,
Lex Fridman (1:29:16.100)
like you mentioned, is that going to be basically
Manolis Kellis (1:29:18.700)
something we have to be very specialized individuals,
Lex Fridman (1:29:22.800)
any advice we give in terms of diet,
Lex Fridman (1:29:24.900)
like what we were just talking about?
Lex Fridman (1:29:25.740)
Believe it or not, the most personalized advice
Manolis Kellis (1:29:28.380)
that you give for nutrition
Lex Fridman (1:29:29.620)
don't have to do with your genome.
Manolis Kellis (1:29:31.460)
They have to do with your gut microbiome,
Lex Fridman (1:29:34.420)
with the bacteria that live inside you.
Lex Fridman (1:29:35.900)
So most of your digestion is actually happening
Lex Fridman (1:29:37.940)
by species that are not human inside you.
Manolis Kellis (1:29:40.800)
You have more nonhuman cells than you have human cells.
Lex Fridman (1:29:43.060)
You're basically a giant bag of bacteria
Manolis Kellis (1:29:46.740)
with a few human cells along.
Lex Fridman (1:29:48.320)
And those do not necessarily have to do
Manolis Kellis (1:29:53.120)
with your genetic makeup.
Lex Fridman (1:29:54.900)
They interact with your genetic makeup.
Manolis Kellis (1:29:56.760)
They interact with your epigenome.
Lex Fridman (1:29:58.000)
They interact with your nutrition.
Manolis Kellis (1:29:59.600)
They interact with your environment.
Lex Fridman (1:30:01.280)
They're basically an additional source of variation.
Lex Fridman (1:30:07.000)
So when you're thinking about sort of
Lex Fridman (1:30:08.120)
personalized nutritional advice,
Lex Fridman (1:30:10.080)
part of that is actually how do you match your microbiome?
Lex Fridman (1:30:13.640)
And part of that is how do we match your genetics?
Lex Fridman (1:30:17.080)
But again, this is a very diverse set of contributors.
Lex Fridman (1:30:22.160)
And the effect sizes are not enormous.
Lex Fridman (1:30:24.640)
So I think the science for that is not fully developed yet.
Lex Fridman (1:30:27.920)
Speaking of diets,
Manolis Kellis (1:30:28.760)
because I've wrestled in combat sports,
Lex Fridman (1:30:30.640)
but sports my whole life were weight matters.
Lex Fridman (1:30:32.640)
So you have to cut and all that stuff.
Lex Fridman (1:30:35.340)
One thing I've learned a lot about my body,
Lex Fridman (1:30:38.240)
and it seems to be, I think,
Lex Fridman (1:30:39.900)
true about other people's bodies,
Manolis Kellis (1:30:41.680)
is that you can adjust to a lot of things.
Lex Fridman (1:30:45.120)
That's the miraculous thing about this biological system,
Manolis Kellis (1:30:48.280)
is like I fast often.
Lex Fridman (1:30:52.360)
I used to eat like five, six times a day
Lex Fridman (1:30:54.920)
and thought that was absolutely necessary.
Lex Fridman (1:30:57.020)
How could you not eat that often?
Lex Fridman (1:30:58.960)
And then when I started fasting,
Lex Fridman (1:31:01.320)
your body adjusted to that.
Lex Fridman (1:31:02.720)
And you learn how to not eat.
Lex Fridman (1:31:04.360)
And it was, if you just give it a chance
Manolis Kellis (1:31:07.600)
for a few weeks, actually,
Lex Fridman (1:31:09.140)
over a period of a few weeks,
Manolis Kellis (1:31:10.320)
your body can adjust to anything.
Lex Fridman (1:31:11.800)
And that's a miraculous, that's such a beautiful thing.
Lex Fridman (1:31:14.120)
So I'm a computer scientist,
Lex Fridman (1:31:15.480)
and I've basically gone through periods of 24 hours
Manolis Kellis (1:31:18.040)
without eating or stopping.
Lex Fridman (1:31:19.680)
And then I'm like, oh, must eat.
Lex Fridman (1:31:22.080)
And I eat a ton.
Lex Fridman (1:31:23.080)
I used to order two pizzas just with my brother.
Lex Fridman (1:31:27.440)
So I've gone through these extremes as well,
Lex Fridman (1:31:29.720)
and I've gone the whole intermittent fasting thing.
Lex Fridman (1:31:32.160)
So I can sympathize with you both on the seven meals a day
Lex Fridman (1:31:35.200)
to the zero meals a day.
Lex Fridman (1:31:37.580)
So I think when I say everything with moderation,
Lex Fridman (1:31:40.820)
I actually think your body responds interestingly
Manolis Kellis (1:31:44.400)
to these different changes in diet.
Lex Fridman (1:31:47.400)
I think part of the reason why we lose weight
Manolis Kellis (1:31:49.920)
with pretty much every kind of change in behavior
Lex Fridman (1:31:52.220)
is because our epigenome and the set of proteins
Lex Fridman (1:31:55.960)
and enzymes that are expressed and our microbiome
Lex Fridman (1:31:58.640)
are not well suited to that nutritional source.
Lex Fridman (1:32:02.080)
And therefore, they will not be able
Lex Fridman (1:32:03.880)
to sort of catch everything that you give them.
Lex Fridman (1:32:06.680)
And then a lot of that will go undigested.
Lex Fridman (1:32:09.160)
And that basically means that your body can then
Manolis Kellis (1:32:11.920)
lose weight in the short term,
Lex Fridman (1:32:13.200)
but very quickly will adjust to that new normal.
Lex Fridman (1:32:16.160)
And then we'll be able to sort of perhaps gain
Lex Fridman (1:32:18.200)
a lot of weight from the diet.
Lex Fridman (1:32:20.400)
So anyway, I mean, there's also studies in factories
Lex Fridman (1:32:24.400)
where basically people dim the lights
Lex Fridman (1:32:27.200)
and then suddenly everybody started working better.
Lex Fridman (1:32:28.720)
It was like, wow, that's amazing.
Manolis Kellis (1:32:30.160)
Three weeks later, they made the lights a little brighter.
Lex Fridman (1:32:32.840)
Everybody started working better.
Lex Fridman (1:32:34.360)
So any kind of intervention has a placebo effect of,
Lex Fridman (1:32:39.440)
wow, now I'm healthier and I'm gonna be running
Manolis Kellis (1:32:41.320)
more often, et cetera.
Lex Fridman (1:32:42.160)
So it's very hard to uncouple the placebo effect
Manolis Kellis (1:32:44.600)
of, wow, I'm doing something to intervene on my diet
Lex Fridman (1:32:47.080)
from the, wow, this is actually the right thing for me.
Manolis Kellis (1:32:50.280)
So, you know.
Lex Fridman (1:32:51.120)
Yeah, from the perspective from a nutrition science,
Manolis Kellis (1:32:53.200)
psychology, both things I'm interested in,
Lex Fridman (1:32:55.760)
especially psychology, it seems that it's extremely difficult
Manolis Kellis (1:32:59.560)
to do good science because there's so many variables
Lex Fridman (1:33:03.440)
involved, it's so difficult to control the variables,
Lex Fridman (1:33:06.560)
so difficult to do sufficiently large scale experiments,
Lex Fridman (1:33:10.280)
both sort of in terms of the number of subjects
Lex Fridman (1:33:12.840)
and temporal, like how long you do the study for,
Lex Fridman (1:33:17.040)
that it just seems like it's not even a real science
Manolis Kellis (1:33:20.720)
for now, like nutrition science.
Lex Fridman (1:33:22.640)
I wanna jump into the whole placebo effect
Manolis Kellis (1:33:24.840)
for a little bit here.
Lex Fridman (1:33:25.840)
And basically talk about the implications of that.
Manolis Kellis (1:33:30.200)
If I give you a sugar pill and I tell you it's a sugar pill,
Lex Fridman (1:33:33.400)
you won't get any better.
Lex Fridman (1:33:35.520)
But if I tell you a sugar pill and I tell you,
Lex Fridman (1:33:38.240)
wow, this is an amazing drug,
Manolis Kellis (1:33:40.000)
it actually will stop your cancer,
Lex Fridman (1:33:42.240)
your cancer will actually stop with much higher probability.
Lex Fridman (1:33:46.280)
What does that mean?
Lex Fridman (1:33:47.120)
That's so amazing.
Manolis Kellis (1:33:47.960)
That means that if I can trick your brain
Lex Fridman (1:33:49.920)
into thinking that I'm healing you,
Manolis Kellis (1:33:51.840)
your brain will basically figure out a way to heal itself,
Lex Fridman (1:33:54.560)
to heal the body.
Lex Fridman (1:33:55.920)
And that tells us that there's so much
Lex Fridman (1:33:58.600)
that we don't understand in the interplay
Manolis Kellis (1:34:01.440)
between our cognition and our biology,
Lex Fridman (1:34:05.320)
that if we were able to better harvest
Manolis Kellis (1:34:08.400)
the power of our brain to sort of impact the body
Lex Fridman (1:34:12.320)
through the placebo effect,
Manolis Kellis (1:34:14.200)
we would be so much better in so many different things.
Lex Fridman (1:34:17.280)
Just by tricking yourself into thinking
Manolis Kellis (1:34:19.200)
that you're doing better, you're actually doing better.
Lex Fridman (1:34:21.560)
So there's something to be said
Manolis Kellis (1:34:22.640)
about sort of positive thinking, about optimism,
Lex Fridman (1:34:25.040)
about sort of just getting your brain
Lex Fridman (1:34:30.040)
and your mind into the right mindset
Lex Fridman (1:34:33.000)
that helps your body and helps your entire biology.
Manolis Kellis (1:34:36.920)
Yeah, from a science perspective, that's just fascinating.
Lex Fridman (1:34:39.840)
Obviously most things about the brain
Manolis Kellis (1:34:41.600)
is a total mystery for now,
Lex Fridman (1:34:43.640)
but that's a fascinating interplay
Manolis Kellis (1:34:46.160)
that the brain can help cure cancer.
Lex Fridman (1:34:54.240)
I don't even know what to do with that.
Manolis Kellis (1:34:55.840)
I mean, the way to think about that is the following.
Lex Fridman (1:34:59.120)
The converse of the equation is something
Manolis Kellis (1:35:01.240)
that we are much more comfortable with.
Lex Fridman (1:35:03.120)
Like, oh, if you're stressed,
Manolis Kellis (1:35:05.720)
then your heart rate might rise
Lex Fridman (1:35:08.200)
and all kinds of sort of toxins might be released
Lex Fridman (1:35:10.960)
and that can have a detrimental effect in your body,
Lex Fridman (1:35:13.600)
et cetera, et cetera, et cetera.
Lex Fridman (1:35:14.800)
So maybe it's easier to understand your body
Lex Fridman (1:35:18.080)
healing from your mind
Manolis Kellis (1:35:20.040)
by your mind is not killing your body,
Lex Fridman (1:35:23.040)
or at least it's killing it less.
Lex Fridman (1:35:24.560)
So I think that aspect of the stress equation
Lex Fridman (1:35:28.000)
is a little easier for most of us to conceptualize,
Lex Fridman (1:35:31.800)
but then the healing part is perhaps the same pathways,
Lex Fridman (1:35:35.160)
perhaps different pathways,
Lex Fridman (1:35:36.160)
but again, something that is totally untapped scientifically.
Lex Fridman (1:35:39.520)
I think we try to bring this question up a couple of times,
Lex Fridman (1:35:42.920)
but let's return to it again,
Lex Fridman (1:35:44.640)
is what do you think is the difference
Manolis Kellis (1:35:46.440)
between the way a computer represents information,
Lex Fridman (1:35:49.480)
the human genome represents and stores information?
Lex Fridman (1:35:53.120)
And maybe broadly, what is the difference
Lex Fridman (1:35:55.560)
between how you think about computers
Lex Fridman (1:35:57.840)
and how you think about biological systems?
Lex Fridman (1:36:00.400)
So I made a very provocative claim earlier
Manolis Kellis (1:36:02.520)
that we are a digital computer.
Lex Fridman (1:36:04.360)
Like I said, at the core lies a digital code
Lex Fridman (1:36:06.360)
and that's true in many ways,
Lex Fridman (1:36:07.520)
but surrounding that digital core,
Manolis Kellis (1:36:09.480)
there's a huge amount of analog.
Lex Fridman (1:36:11.440)
If you look at our brain, it's not really digital.
Manolis Kellis (1:36:13.680)
If you look at our sort of RNA
Lex Fridman (1:36:15.760)
and all of that stuff inside our cell,
Manolis Kellis (1:36:17.160)
it's not really digital.
Lex Fridman (1:36:18.000)
It's really analog in many ways,
Lex Fridman (1:36:21.000)
but let's start with the code
Lex Fridman (1:36:22.800)
and then we'll expand to the rest.
Lex Fridman (1:36:24.800)
So the code itself is digital.
Lex Fridman (1:36:27.800)
So there's genes.
Manolis Kellis (1:36:28.680)
You can think of the genes as, I don't know,
Lex Fridman (1:36:30.840)
the procedures, the functions inside your language.
Lex Fridman (1:36:33.840)
And then somehow you have to turn these functions on.
Lex Fridman (1:36:36.400)
How do you call a gene?
Lex Fridman (1:36:37.320)
How do you call that function?
Lex Fridman (1:36:39.360)
The way that you would do it in old programming languages
Manolis Kellis (1:36:41.880)
is go to address whatever in your memory
Lex Fridman (1:36:44.560)
and then you'd start running from there.
Lex Fridman (1:36:46.240)
And modern programming languages
Lex Fridman (1:36:48.920)
have encapsulated this into functions
Lex Fridman (1:36:50.800)
and objects and all of that.
Lex Fridman (1:36:52.000)
And it's nice and cute, but in the end, deep down,
Manolis Kellis (1:36:54.560)
there's still an assembly code
Lex Fridman (1:36:55.560)
that says go to that instruction
Lex Fridman (1:36:57.600)
and it runs that instruction.
Lex Fridman (1:36:59.680)
If you look at the human genome
Lex Fridman (1:37:01.600)
and the genome of pretty much most species out there,
Lex Fridman (1:37:06.800)
there's no go to function.
Manolis Kellis (1:37:08.160)
You just don't start transcribing in position 13,000,
Lex Fridman (1:37:15.280)
13,527 in chromosome 12.
Manolis Kellis (1:37:18.760)
You instead have content based indexing.
Lex Fridman (1:37:21.920)
So at every location in the genome,
Manolis Kellis (1:37:25.120)
in front of the genes that need to be turned on,
Lex Fridman (1:37:28.760)
I don't know, when you drink coffee,
Manolis Kellis (1:37:30.560)
there's a little coffee marker in front of all of them.
Lex Fridman (1:37:34.440)
And whenever your cells that metabolize coffee
Manolis Kellis (1:37:38.480)
need to metabolize coffee,
Lex Fridman (1:37:39.760)
they basically see coffee and they're like,
Manolis Kellis (1:37:41.280)
ooh, let's go turn on all the coffee marked genes.
Lex Fridman (1:37:44.720)
So there's basically these small motifs,
Manolis Kellis (1:37:48.040)
these small sequences that we call regulatory motifs.
Lex Fridman (1:37:50.840)
They're like patterns of DNA.
Manolis Kellis (1:37:52.040)
They're only eight characters long or so,
Lex Fridman (1:37:54.680)
like GAT, GCA, et cetera.
Lex Fridman (1:37:57.920)
And these motifs work in combinations
Lex Fridman (1:38:01.560)
and every one of them has some recruitment affinity
Manolis Kellis (1:38:06.320)
for a different protein that will then come and bind it
Lex Fridman (1:38:09.520)
and together collections of these motifs
Manolis Kellis (1:38:11.840)
create regions that we call regulatory regions
Lex Fridman (1:38:15.440)
that can be either promoters near the beginning of the gene
Lex Fridman (1:38:19.280)
and that basically tells you
Lex Fridman (1:38:20.160)
where the function actually starts, where you call it,
Lex Fridman (1:38:22.480)
and then enhancers that are looping around of the DNA
Lex Fridman (1:38:26.200)
that basically bring the machinery
Manolis Kellis (1:38:28.200)
that binds those enhancers
Lex Fridman (1:38:29.800)
and then bring it onto the promoter,
Manolis Kellis (1:38:32.520)
which then recruits the right sort of the ribosome
Lex Fridman (1:38:36.080)
and the polymerase and all of that thing,
Manolis Kellis (1:38:37.800)
which will first transcribe and then export
Lex Fridman (1:38:40.560)
and then eventually translate in the cytoplasm,
Manolis Kellis (1:38:42.680)
you know, whatever RNA molecule.
Lex Fridman (1:38:45.520)
So the beauty of the way
Manolis Kellis (1:38:50.520)
that the digital computer that's the genome works
Lex Fridman (1:38:54.280)
is that it's extremely fault tolerant.
Manolis Kellis (1:38:57.960)
If I took your hard drive
Lex Fridman (1:38:59.480)
and I messed with 20% of the letters in it,
Manolis Kellis (1:39:03.160)
of the zeros and ones and I flipped them,
Lex Fridman (1:39:05.880)
you'd be in trouble.
Manolis Kellis (1:39:07.400)
If I take the genome and I flipped 20% of the letters,
Lex Fridman (1:39:11.400)
you probably won't even notice.
Lex Fridman (1:39:13.880)
And that resilience.
Lex Fridman (1:39:15.320)
That's fascinating, yeah.
Manolis Kellis (1:39:16.720)
Is a key design principle.
Lex Fridman (1:39:18.840)
And again, I'm anthropomorphizing here,
Lex Fridman (1:39:20.760)
but it's a key driving principle
Lex Fridman (1:39:22.400)
of how biological systems work.
Manolis Kellis (1:39:24.320)
They're first resilient and then anything else.
Lex Fridman (1:39:27.880)
And when you look at this incredible beauty of life
Manolis Kellis (1:39:32.160)
from the most, I don't know, beautiful,
Lex Fridman (1:39:35.480)
I don't know, human genome maybe of humanity
Lex Fridman (1:39:38.280)
and all of the ideals that should come with it
Lex Fridman (1:39:40.840)
to the most terrifying genome,
Manolis Kellis (1:39:42.360)
like, I don't know, COVID 19, SARS COVID 2
Lex Fridman (1:39:45.520)
and the current pandemic,
Manolis Kellis (1:39:47.640)
you basically see this elegance
Lex Fridman (1:39:50.280)
as the epitome of clean design,
Lex Fridman (1:39:54.520)
but it's dirty.
Lex Fridman (1:39:55.920)
It's a mess.
Manolis Kellis (1:39:57.280)
It's, you know, the way to get there is hugely messy.
Lex Fridman (1:40:02.200)
And that's something that we as computer scientists
Manolis Kellis (1:40:04.400)
don't embrace.
Lex Fridman (1:40:06.120)
We like to have clean code.
Manolis Kellis (1:40:08.040)
You know, like in engineering,
Lex Fridman (1:40:10.360)
they teach you about compartmentalization,
Manolis Kellis (1:40:12.240)
about sort of separating functions,
Lex Fridman (1:40:13.720)
about modularity, about hierarchical design.
Manolis Kellis (1:40:17.080)
None of that applies in biology.
Lex Fridman (1:40:19.040)
Testing.
Manolis Kellis (1:40:19.880)
Testing, sure.
Lex Fridman (1:40:22.320)
Yeah, biology does plenty of that.
Lex Fridman (1:40:24.200)
But I mean, through evolutionary exploration.
Lex Fridman (1:40:26.800)
But if you look at biological systems,
Manolis Kellis (1:40:31.040)
first they are robust
Lex Fridman (1:40:33.440)
and then they specialize to become anything else.
Lex Fridman (1:40:36.680)
And if you look at viruses,
Lex Fridman (1:40:38.240)
the reason why they're so elegant
Manolis Kellis (1:40:41.040)
when you look at the design of this, you know, genome,
Lex Fridman (1:40:44.600)
it seems so elegant.
Lex Fridman (1:40:46.120)
And the reason for that is that it's been stripped down
Lex Fridman (1:40:49.680)
from something much larger
Manolis Kellis (1:40:51.600)
because of the pressure to keep it compact.
Lex Fridman (1:40:53.960)
So many compact genomes out there
Manolis Kellis (1:40:56.040)
have ancestors that were much larger.
Lex Fridman (1:40:58.680)
You don't start small and become big.
Manolis Kellis (1:41:00.760)
You go through a loop of add a bunch of stuff,
Lex Fridman (1:41:03.640)
increase complexity, and then, you know, slim it down.
Lex Fridman (1:41:07.240)
And one of my early papers was in fact on genome duplication.
Lex Fridman (1:41:12.120)
One of the things we found is that baker's yeast,
Manolis Kellis (1:41:14.080)
which is the, you know, yeast that you use to make bread,
Lex Fridman (1:41:17.600)
but also the yeast that you use to make wine,
Manolis Kellis (1:41:19.480)
which is basically the dominant species
Lex Fridman (1:41:20.960)
when you go in the fields of Tuscany
Lex Fridman (1:41:22.360)
and you say, you know, what's out there,
Lex Fridman (1:41:24.040)
it's basically saccharomyces cerevisiae,
Manolis Kellis (1:41:26.320)
or the way my Italian friends say,
Lex Fridman (1:41:27.880)
saccharomyces cerevisiae.
Manolis Kellis (1:41:30.120)
So, so.
Lex Fridman (1:41:33.000)
Oh, which means what?
Manolis Kellis (1:41:34.480)
Oh, saccharomyces, okay, I'm sorry, I'm Greek.
Lex Fridman (1:41:36.680)
So yeah, zacharo, mikis, zacharo is sugar,
Manolis Kellis (1:41:39.680)
mikis is fungus.
Lex Fridman (1:41:41.120)
Yes, cerevisiae, cerveza, beer.
Lex Fridman (1:41:44.520)
So it means the sugar fungus of beer.
Lex Fridman (1:41:47.200)
Yeah.
Manolis Kellis (1:41:48.040)
You know, less, less sounding to the ear.
Lex Fridman (1:41:51.080)
Still poetic, yeah.
Lex Fridman (1:41:52.840)
So anyway, saccharomyces cerevisiae,
Lex Fridman (1:41:54.960)
basically the major baker's yeast out there
Manolis Kellis (1:41:57.040)
is the descendant of a whole genome duplication.
Lex Fridman (1:42:00.440)
Why would a whole gene duplication even happen?
Manolis Kellis (1:42:02.880)
When it happened is coinciding
Lex Fridman (1:42:06.080)
with about a hundred million years ago
Lex Fridman (1:42:08.240)
and the emergence of fruit bearing plants.
Lex Fridman (1:42:14.320)
Why fruit bearing plants?
Manolis Kellis (1:42:15.560)
Because animals would eat the fruit
Lex Fridman (1:42:19.040)
and would walk around and poop huge amounts of nutrients
Manolis Kellis (1:42:23.640)
along with a seed for the plants to spread.
Lex Fridman (1:42:26.480)
Before that, plants were not spreading through animals,
Manolis Kellis (1:42:29.000)
they were spreading through wind
Lex Fridman (1:42:30.480)
and all kinds of other ways.
Lex Fridman (1:42:32.360)
But basically the moment you have fruit bearing plants,
Lex Fridman (1:42:34.720)
these plants are basically creating this abundance
Manolis Kellis (1:42:38.760)
of sugar in the environment.
Lex Fridman (1:42:40.240)
So there's an evolutionary niche that gets created.
Lex Fridman (1:42:42.920)
And in that evolutionary niche,
Lex Fridman (1:42:44.080)
you basically have enough sugar
Manolis Kellis (1:42:46.640)
that a whole genome duplication,
Lex Fridman (1:42:48.680)
which initially is a very messy event,
Manolis Kellis (1:42:51.040)
allows you to then, you know,
Lex Fridman (1:42:53.760)
relieve some of that complexity.
Lex Fridman (1:42:56.000)
So I had to pause, what does genome duplication mean?
Lex Fridman (1:42:59.520)
That basically means that instead of having eight chromosomes,
Manolis Kellis (1:43:03.200)
you can now have 16 chromosomes.
Lex Fridman (1:43:06.240)
So, but the duplication at first,
Manolis Kellis (1:43:09.600)
when you go to 16, you're not using that.
Lex Fridman (1:43:13.760)
Oh yeah, you are.
Manolis Kellis (1:43:15.120)
Yeah, so basically from one day to the next,
Lex Fridman (1:43:17.280)
you went from having eight chromosomes
Manolis Kellis (1:43:18.600)
to having 16 chromosomes.
Lex Fridman (1:43:20.280)
Probably a non disjunction event during a duplication,
Manolis Kellis (1:43:22.920)
during a division.
Lex Fridman (1:43:24.160)
So you basically divide the cell
Manolis Kellis (1:43:25.800)
instead of half the genome going this way
Lex Fridman (1:43:27.760)
and half the genome going the other way
Manolis Kellis (1:43:29.000)
after duplication of the genome,
Lex Fridman (1:43:30.640)
you basically have all of it going to one cell
Lex Fridman (1:43:33.000)
and then there's sufficient messiness there
Lex Fridman (1:43:35.840)
that you end up with slight differences
Manolis Kellis (1:43:38.320)
that make most of these chromosomes
Lex Fridman (1:43:39.760)
be actually preserved.
Manolis Kellis (1:43:42.280)
It's a long story short to me.
Lex Fridman (1:43:43.120)
But that's a big upgrade, right?
Lex Fridman (1:43:45.000)
So that's...
Lex Fridman (1:43:45.840)
Not necessarily,
Manolis Kellis (1:43:46.760)
because what happens immediately thereafter
Lex Fridman (1:43:48.480)
is that you start massively losing
Manolis Kellis (1:43:50.320)
tons of those duplicated genes.
Lex Fridman (1:43:52.280)
So 90% of those genes were actually lost
Manolis Kellis (1:43:55.360)
very rapidly after whole gene duplication.
Lex Fridman (1:43:58.160)
And the reason for that is that biology is not intelligent,
Manolis Kellis (1:44:01.680)
it's just ruthless selection, random mutation.
Lex Fridman (1:44:06.480)
So the ruthless selection basically means
Manolis Kellis (1:44:08.520)
that as soon as one of the random mutations hit one gene,
Lex Fridman (1:44:11.400)
ruthless selection just kills off that gene.
Manolis Kellis (1:44:13.360)
It's just,
Lex Fridman (1:44:16.680)
if you have a pressure to maintain a small compact genome,
Manolis Kellis (1:44:19.520)
you will very rapidly lose the second copy
Lex Fridman (1:44:21.680)
of most of your genes and a small number 10%
Manolis Kellis (1:44:24.240)
were kept in two copies.
Lex Fridman (1:44:25.720)
And those had to do a lot with environment adaptation,
Manolis Kellis (1:44:28.800)
with the speed of replication,
Lex Fridman (1:44:31.080)
with the speed of translation and with sugar processing.
Lex Fridman (1:44:34.240)
So I'm making a long story short
Lex Fridman (1:44:36.000)
to basically say that evolution is messy.
Manolis Kellis (1:44:38.720)
The only way...
Lex Fridman (1:44:39.960)
Like, so the example that I was giving
Manolis Kellis (1:44:42.160)
of messing with 20% of your bits in your computer,
Lex Fridman (1:44:45.840)
totally bogus.
Manolis Kellis (1:44:47.200)
Duplicating all your functions
Lex Fridman (1:44:48.720)
and just throwing them out there in the same function,
Manolis Kellis (1:44:51.880)
just totally bogus.
Lex Fridman (1:44:52.840)
Like this would never work in an engineer system.
Lex Fridman (1:44:55.200)
But biological systems,
Lex Fridman (1:44:56.960)
because of this content based indexing
Lex Fridman (1:44:59.080)
and because of this modularity that comes
Lex Fridman (1:45:01.880)
from the fact that the gene is controlled
Manolis Kellis (1:45:04.200)
by a series of tags.
Lex Fridman (1:45:05.200)
And now if you need this gene in another setting,
Manolis Kellis (1:45:08.160)
you just add some more tags
Lex Fridman (1:45:09.760)
that will basically turn it on also in those settings.
Lex Fridman (1:45:12.560)
So this gene is now pressured to do two different functions
Lex Fridman (1:45:17.280)
and it builds up complexity.
Manolis Kellis (1:45:19.760)
I see a whole gene duplication
Lex Fridman (1:45:21.240)
and gene duplication in general
Manolis Kellis (1:45:22.560)
as a way to relieve that complexity.
Lex Fridman (1:45:24.560)
So you have this gradual buildup of complexity
Manolis Kellis (1:45:26.640)
as functions get sort of added onto the existing genes.
Lex Fridman (1:45:30.920)
And then boom, you duplicate your workforce.
Lex Fridman (1:45:34.160)
And you now have two copies of this gene.
Lex Fridman (1:45:36.720)
One will probably specialize to do one
Lex Fridman (1:45:38.760)
and the other one will specialize to do the other
Lex Fridman (1:45:40.440)
or one will maintain the ancestral function.
Manolis Kellis (1:45:42.160)
The other one will sort of be free to evolve
Lex Fridman (1:45:44.800)
and specialize while losing the ancestral function
Lex Fridman (1:45:47.720)
and so on and so forth.
Lex Fridman (1:45:48.640)
So that's how genomes evolve.
Manolis Kellis (1:45:49.960)
They're just messy things,
Lex Fridman (1:45:52.040)
but they're extremely fault tolerant
Lex Fridman (1:45:54.600)
and they're extremely able to deal with mutations
Lex Fridman (1:45:58.400)
because that's the very way that you generate new functions.
Lex Fridman (1:46:03.800)
So new functionalization comes
Lex Fridman (1:46:05.440)
from the very thing that breaks it.
Lex Fridman (1:46:07.720)
So even in the current pandemic,
Lex Fridman (1:46:09.120)
many people are asking me which mutations matter the most.
Lex Fridman (1:46:12.560)
And what I tell them is,
Lex Fridman (1:46:13.800)
well, we can study the evolutionary dynamics
Manolis Kellis (1:46:16.240)
of the current genome to then understand
Lex Fridman (1:46:19.520)
which mutations have previously happened or not.
Lex Fridman (1:46:23.120)
And which mutations happen in genes
Lex Fridman (1:46:26.040)
that evolve rapidly or not.
Lex Fridman (1:46:28.040)
And one of the things we found, for example,
Lex Fridman (1:46:29.720)
is that the genes that evolved rapidly in the past
Manolis Kellis (1:46:33.840)
are still evolving rapidly now in the current pandemic.
Lex Fridman (1:46:36.720)
The genes that evolved slowly in the past
Manolis Kellis (1:46:38.680)
are still evolving slowly.
Lex Fridman (1:46:40.040)
Which means that they're useful?
Manolis Kellis (1:46:41.720)
Which means that they're under
Lex Fridman (1:46:43.360)
the same evolutionary pressures.
Lex Fridman (1:46:45.280)
But then the question is what happens in specific mutations?
Lex Fridman (1:46:49.440)
So if you look at the D614 gene mutations,
Manolis Kellis (1:46:52.560)
that's been all over the news.
Lex Fridman (1:46:53.840)
So in position 614, in the amino acids 614 of the S protein,
Manolis Kellis (1:46:59.720)
there's a D2 gene mutation
Lex Fridman (1:47:02.120)
that sort of has creeped over the population.
Manolis Kellis (1:47:07.040)
That mutation, we found out through my work,
Lex Fridman (1:47:10.040)
disrupts a perfectly conserved nucleotide position
Manolis Kellis (1:47:13.480)
that has never been changed in the history
Lex Fridman (1:47:15.800)
of millions of years of equivalent
Manolis Kellis (1:47:17.920)
per million evolution of these viruses.
Lex Fridman (1:47:23.080)
That basically means that it's a completely new adaptation
Manolis Kellis (1:47:25.960)
to human.
Lex Fridman (1:47:27.480)
And that mutation has now gone from 1% frequency
Manolis Kellis (1:47:30.840)
to 90% frequency in almost all outbreaks.
Lex Fridman (1:47:33.800)
So this mutation, I like how you say the 416,
Lex Fridman (1:47:38.880)
what was it, okay.
Lex Fridman (1:47:39.720)
Yeah, 614, sorry.
Manolis Kellis (1:47:40.640)
614.
Lex Fridman (1:47:41.480)
D614G.
Manolis Kellis (1:47:43.200)
D614, so literally, so what you're saying
Lex Fridman (1:47:46.560)
is this is like a chess move.
Lex Fridman (1:47:48.440)
So it just mutated one letter to another.
Lex Fridman (1:47:50.560)
Exactly.
Lex Fridman (1:47:51.400)
And that hasn't happened before.
Lex Fridman (1:47:53.000)
Yeah, never.
Lex Fridman (1:47:54.320)
And this somehow, this mutation is really useful.
Lex Fridman (1:47:58.120)
It's really useful in the current environment of the genome,
Manolis Kellis (1:48:02.000)
which is moving from human to human.
Lex Fridman (1:48:04.920)
When it was moving from bat to bat,
Manolis Kellis (1:48:06.840)
it couldn't care less for that mutation,
Lex Fridman (1:48:08.680)
but it's environment specific.
Lex Fridman (1:48:09.960)
So now that it's moving from human to human,
Lex Fridman (1:48:12.560)
it's moving way better, like by orders of magnitude.
Lex Fridman (1:48:15.920)
What do you, okay, so you're like tracking
Lex Fridman (1:48:18.440)
this evolutionary dynamics, which is fascinating,
Lex Fridman (1:48:22.520)
but what do you do with that?
Lex Fridman (1:48:24.200)
So what does that mean?
Lex Fridman (1:48:25.320)
What does this mean, what do you make,
Lex Fridman (1:48:27.560)
what do you make of this mutation
Manolis Kellis (1:48:29.120)
in trying to anticipate, I guess,
Lex Fridman (1:48:31.560)
is one of the things you're trying to do
Manolis Kellis (1:48:34.120)
is anticipate where, how this unrolls into the future,
Lex Fridman (1:48:37.840)
this evolutionary dynamics.
Manolis Kellis (1:48:39.760)
Such a great question.
Lex Fridman (1:48:40.640)
So there's two things.
Manolis Kellis (1:48:42.880)
Remember when I was saying earlier,
Lex Fridman (1:48:44.680)
mutation is the path to new things,
Lex Fridman (1:48:47.040)
but also the path to break old things.
Lex Fridman (1:48:49.720)
So what we know is that this position
Manolis Kellis (1:48:52.960)
was extremely preserved through gazillions of mutations.
Lex Fridman (1:48:56.600)
That mutation was never tolerated
Manolis Kellis (1:48:58.440)
when it was moving from bats to bats.
Lex Fridman (1:49:00.200)
So that basically means that that position
Manolis Kellis (1:49:02.480)
is extremely important in the function of that protein.
Lex Fridman (1:49:05.600)
That's the first thing it tells.
Manolis Kellis (1:49:06.920)
The second one is that that position
Lex Fridman (1:49:09.280)
was very well suited to bat transmission,
Lex Fridman (1:49:12.360)
but now is not well suited to human transmission,
Lex Fridman (1:49:14.720)
so it got rid of it.
Lex Fridman (1:49:15.840)
And it now has a new version of that amino acid
Lex Fridman (1:49:18.840)
that basically makes it much easier
Manolis Kellis (1:49:20.920)
to transmit from human to human.
Lex Fridman (1:49:22.720)
So in terms of the evolutionary history
Manolis Kellis (1:49:27.440)
teaching us about the future,
Lex Fridman (1:49:29.800)
it basically tells us here's the regions
Manolis Kellis (1:49:31.960)
that are currently mutating.
Lex Fridman (1:49:34.760)
Here's the regions that are most likely
Manolis Kellis (1:49:36.400)
to mutate going forward.
Lex Fridman (1:49:37.880)
As you're building a vaccine,
Manolis Kellis (1:49:39.440)
here's what you should be focusing on
Lex Fridman (1:49:41.680)
in terms of the most stable regions
Manolis Kellis (1:49:43.520)
that are the least likely to mutate.
Lex Fridman (1:49:45.440)
Or here's the newly evolved functions
Manolis Kellis (1:49:48.200)
that are the most likely to be important
Lex Fridman (1:49:50.240)
because they've overcome this local maximum
Manolis Kellis (1:49:54.560)
that it had reached in the bat transmission.
Lex Fridman (1:49:59.360)
So anyway, it's a tangent to basically say
Manolis Kellis (1:50:01.760)
that evolution works in messy ways.
Lex Fridman (1:50:04.080)
And the thing that you would break
Manolis Kellis (1:50:07.400)
is the thing that actually allows you
Lex Fridman (1:50:10.320)
to first go through a lull
Lex Fridman (1:50:12.160)
and then reaching new local maximum.
Lex Fridman (1:50:15.200)
And I often like to say that if engineers
Manolis Kellis (1:50:18.920)
had basically designed evolution,
Lex Fridman (1:50:21.200)
we would still be perfectly replicating bacteria
Manolis Kellis (1:50:26.640)
because it's my making the bacterium worse
Lex Fridman (1:50:29.400)
that you allow evolution to reach a new optimum.
Manolis Kellis (1:50:32.200)
That's, just to pause on that,
Lex Fridman (1:50:34.520)
that's so profound.
Manolis Kellis (1:50:35.840)
That's so profound for the entirety
Lex Fridman (1:50:39.400)
of this scientific and engineering disciplines.
Manolis Kellis (1:50:44.680)
Exactly.
Lex Fridman (1:50:45.520)
We as engineers need to embrace breaking things.
Manolis Kellis (1:50:48.520)
We as engineers need to embrace robustness
Lex Fridman (1:50:50.960)
as the first principle beyond perfection
Manolis Kellis (1:50:54.240)
because nothing's gonna ever be perfect.
Lex Fridman (1:50:56.040)
And when you're sending a satellite to Mars,
Manolis Kellis (1:50:58.440)
when something goes wrong, it'll break down.
Lex Fridman (1:51:01.160)
As opposed to building systems that tolerate failure
Lex Fridman (1:51:04.600)
and are resilient to that.
Lex Fridman (1:51:08.840)
And in fact, get better through that.
Lex Fridman (1:51:11.080)
So the SpaceX approach versus NASA for the...
Lex Fridman (1:51:14.400)
For example.
Manolis Kellis (1:51:16.200)
Is there something we can learn about the incredible,
Lex Fridman (1:51:21.320)
take lessons from the incredible biological systems
Manolis Kellis (1:51:23.960)
in their resilience, in the mushiness, the messiness
Lex Fridman (1:51:27.600)
to our computing systems, to our computers?
Manolis Kellis (1:51:31.880)
It would basically be starting from scratch in many ways.
Lex Fridman (1:51:35.280)
It would basically be building new paradigms
Manolis Kellis (1:51:38.960)
that don't try to get the right answer all the time,
Lex Fridman (1:51:42.760)
but try to get the right answer most of the time
Manolis Kellis (1:51:45.600)
or a lot of the time.
Lex Fridman (1:51:47.000)
Do you see deep learning systems in the whole world
Manolis Kellis (1:51:49.280)
of machine learning as kind of taking a step
Lex Fridman (1:51:51.120)
in that direction?
Manolis Kellis (1:51:52.000)
Absolutely, absolutely.
Lex Fridman (1:51:53.600)
Basically by allowing this much more natural evolution
Manolis Kellis (1:51:57.560)
of these parameters, you basically...
Lex Fridman (1:52:01.080)
And if you look at sort of deep learning systems again,
Manolis Kellis (1:52:04.000)
they're not inspired by the genome aspect of biology,
Lex Fridman (1:52:07.440)
they're inspired by the brain aspect of biology.
Lex Fridman (1:52:10.160)
And again, I want you to pause for a second
Lex Fridman (1:52:12.560)
and realize the complexity of the entire human brain
Manolis Kellis (1:52:18.720)
with trillions of connections within our neurons,
Lex Fridman (1:52:22.760)
with millions of cells talking to each other,
Manolis Kellis (1:52:26.640)
is still encoded within that same genome.
Lex Fridman (1:52:29.040)
That same genome encodes every single freaking cell type
Manolis Kellis (1:52:36.080)
of the entire body.
Lex Fridman (1:52:37.920)
Every single cell is encoded by the same code.
Lex Fridman (1:52:41.040)
And yet specialization allows you to have
Lex Fridman (1:52:45.240)
the single viral like genome that self replicates,
Manolis Kellis (1:52:50.040)
the single module, modular automaton,
Lex Fridman (1:52:54.200)
work with other copies of itself, it's mind boggling.
Manolis Kellis (1:52:57.360)
Create complex organs through which blood flows.
Lex Fridman (1:53:01.600)
And what is that blood?
Manolis Kellis (1:53:02.680)
The same freaking genome.
Lex Fridman (1:53:05.560)
Create organs that communicate with each other.
Lex Fridman (1:53:09.840)
And what are these organs?
Lex Fridman (1:53:11.080)
The exact same genome.
Manolis Kellis (1:53:13.120)
Create a brain that is innervated by massive amounts
Lex Fridman (1:53:17.560)
of blood pumping energy to it,
Manolis Kellis (1:53:21.240)
20% of our energetic needs to the brain from the same genome.
Lex Fridman (1:53:28.240)
And all of the neuronal connections,
Manolis Kellis (1:53:30.120)
all of the auxiliary cells, all of the immune cells,
Lex Fridman (1:53:33.920)
the astrocytes, the ligodendrocytes, the neurons,
Manolis Kellis (1:53:35.920)
the excitatory, the inhibitory neurons,
Lex Fridman (1:53:37.360)
all of the different classes of parasites,
Manolis Kellis (1:53:39.480)
the blood brain barrier, all of that, same genome.
Lex Fridman (1:53:42.880)
One way to see that in a sad, this one is beautiful.
Manolis Kellis (1:53:47.680)
The sad thing is thinking about the trillions
Lex Fridman (1:53:50.800)
of organisms that died to create that.
Lex Fridman (1:53:55.240)
You mean on the evolutionary path to humans?
Lex Fridman (1:53:57.080)
On the evolutionary path to humans.
Manolis Kellis (1:53:59.600)
It's crazy, there's two descendant of apes
Lex Fridman (1:54:02.680)
just talking on a podcast.
Manolis Kellis (1:54:04.760)
Okay, it's just so mind boggling.
Lex Fridman (1:54:08.520)
Just to boggle our minds a little bit more.
Manolis Kellis (1:54:11.120)
Us talking to each other,
Lex Fridman (1:54:13.920)
we are basically generating a series of vocal utterances
Manolis Kellis (1:54:18.440)
through our pulsating of vocal cords received through this.
Lex Fridman (1:54:23.440)
The people who listen to this
Manolis Kellis (1:54:26.160)
are taking a completely different path
Lex Fridman (1:54:29.240)
to that information transfer, yet through language.
Lex Fridman (1:54:32.880)
But imagine if we could connect these brains
Lex Fridman (1:54:36.160)
directly to each other.
Manolis Kellis (1:54:38.920)
The amount of information that I'm condensing
Lex Fridman (1:54:41.600)
into a small number of words is a huge funnel,
Manolis Kellis (1:54:46.360)
which then you receive and you expand
Lex Fridman (1:54:49.480)
into a huge number of thoughts from that small funnel.
Manolis Kellis (1:54:55.760)
In many ways, engineers would love
Lex Fridman (1:54:58.080)
to have the whole information transfer,
Manolis Kellis (1:54:59.880)
just take the whole set of neurons and throw them away.
Lex Fridman (1:55:02.640)
I mean, throw them to the other person.
Manolis Kellis (1:55:05.440)
This might actually not be better
Lex Fridman (1:55:07.280)
because in your misinterpretation
Manolis Kellis (1:55:10.640)
of every word that I'm saying,
Lex Fridman (1:55:13.000)
you are creating new interpretation
Manolis Kellis (1:55:14.680)
that might actually be way better
Lex Fridman (1:55:16.080)
than what I meant in the first place.
Manolis Kellis (1:55:17.920)
The ambiguity of language perhaps
Lex Fridman (1:55:21.760)
might be the secret to creativity.
Manolis Kellis (1:55:25.000)
Every single time you work on a project by yourself,
Lex Fridman (1:55:28.400)
you only bounce ideas with one person
Lex Fridman (1:55:31.120)
and your neurons are basically fully cognizant
Lex Fridman (1:55:33.760)
of what these ideas are.
Lex Fridman (1:55:35.880)
But the moment you interact with another person,
Lex Fridman (1:55:37.720)
the misinterpretations that happen
Manolis Kellis (1:55:41.080)
might be the most creative part of the process.
Lex Fridman (1:55:43.760)
With my students, every time we have a research meeting,
Manolis Kellis (1:55:45.600)
I very often pause and say,
Lex Fridman (1:55:47.560)
let me repeat what you just said in a different way.
Lex Fridman (1:55:50.400)
And I sort of go on and brainstorm
Lex Fridman (1:55:52.400)
with what they were saying,
Lex Fridman (1:55:53.680)
but by the third time,
Lex Fridman (1:55:55.960)
it's not what they were saying at all.
Lex Fridman (1:55:58.000)
And when they pick up what I'm saying,
Lex Fridman (1:55:59.480)
they're like, oh, well, dah, dah, dah.
Manolis Kellis (1:56:01.160)
Now they've sort of learned something very different
Lex Fridman (1:56:04.160)
from what I was saying.
Lex Fridman (1:56:05.000)
And that is the same kind of messiness
Lex Fridman (1:56:08.480)
that I'm describing in the genome itself.
Manolis Kellis (1:56:10.960)
It's sort of embracing the messiness.
Lex Fridman (1:56:13.600)
And that's a feature, not a book.
Manolis Kellis (1:56:15.400)
Exactly.
Lex Fridman (1:56:16.240)
And in the same way, when you're thinking
Manolis Kellis (1:56:17.560)
about sort of these deep learning systems
Lex Fridman (1:56:19.960)
that will allow us to sort of be more creative perhaps
Manolis Kellis (1:56:23.600)
or learn better approximations of these complex functions,
Lex Fridman (1:56:27.560)
again, tuned to the universe that we inhabit,
Manolis Kellis (1:56:30.720)
you have to embrace the breaking.
Lex Fridman (1:56:33.680)
You have to embrace the,
Lex Fridman (1:56:35.400)
how do we get out of these local optima?
Lex Fridman (1:56:38.000)
And a lot of the design paradigms
Manolis Kellis (1:56:40.960)
that have made deep learning so successful
Lex Fridman (1:56:43.400)
are ways to get away from that,
Manolis Kellis (1:56:45.400)
ways to get better training
Lex Fridman (1:56:47.360)
by sort of sending long range messages,
Manolis Kellis (1:56:50.520)
these LSTM models and the sort of feed forward loops
Lex Fridman (1:56:55.920)
that sort of jump through layers
Manolis Kellis (1:56:59.280)
of a convolutional neural network.
Lex Fridman (1:57:00.920)
All of these things are basically ways to push you out
Manolis Kellis (1:57:04.960)
of these local maxima.
Lex Fridman (1:57:07.360)
And that's sort of what evolution does.
Manolis Kellis (1:57:08.840)
That's what language does.
Lex Fridman (1:57:09.840)
That's what conversation and brainstorming does.
Manolis Kellis (1:57:12.360)
That's what our brain does.
Lex Fridman (1:57:14.120)
So this design paradigm is something that's pervasive
Lex Fridman (1:57:18.320)
and yet not taught in schools,
Lex Fridman (1:57:20.560)
not taught in engineering schools
Manolis Kellis (1:57:22.280)
where everything's minutely modularized
Lex Fridman (1:57:24.520)
to make sure that we never deviate
Manolis Kellis (1:57:26.040)
from whatever signal we're trying to emit
Lex Fridman (1:57:28.640)
as opposed to let all hell breaks loose
Manolis Kellis (1:57:31.440)
because that's the path to paradise.
Lex Fridman (1:57:34.000)
The path to paradise.
Manolis Kellis (1:57:35.440)
Yeah, I mean, it's difficult to know how to teach that
Lex Fridman (1:57:38.040)
and what to do with it.
Manolis Kellis (1:57:39.320)
I mean, it's difficult to know how to build up
Lex Fridman (1:57:43.680)
the scientific method around messiness.
Manolis Kellis (1:57:46.640)
I mean, it's not all messiness.
Lex Fridman (1:57:49.960)
We need some cleanness.
Lex Fridman (1:57:51.960)
And going back to the example with Mars,
Lex Fridman (1:57:54.400)
that's probably the place where I want
Manolis Kellis (1:57:55.520)
to sort of moderate error as much as possible
Lex Fridman (1:57:58.800)
and sort of control the environment as much as possible.
Lex Fridman (1:58:01.080)
But if you're trying to repopulate Mars,
Lex Fridman (1:58:03.160)
well, maybe messiness is a good thing then.
Manolis Kellis (1:58:05.320)
On that, you quickly mentioned this
Lex Fridman (1:58:09.280)
in terms of us using our vocal cords
Manolis Kellis (1:58:12.920)
to speak on a podcast.
Lex Fridman (1:58:15.120)
So Elon Musk and Neuralink are working
Manolis Kellis (1:58:17.800)
on trying to plug, as per our discussion
Lex Fridman (1:58:22.680)
with computers and biological systems,
Manolis Kellis (1:58:24.920)
to connect the two.
Lex Fridman (1:58:25.840)
He's trying to connect our brain to a computer
Manolis Kellis (1:58:30.640)
to create a brain computer interface
Lex Fridman (1:58:32.840)
where they can communicate back and forth.
Manolis Kellis (1:58:36.160)
On this line of thinking, do you think this is possible
Lex Fridman (1:58:40.960)
to bridge the gap between our engineered computing systems
Lex Fridman (1:58:45.200)
and the messy biological systems?
Lex Fridman (1:58:49.280)
My answer would be absolutely.
Manolis Kellis (1:58:51.920)
You know, there's no doubt that we can understand
Lex Fridman (1:58:54.440)
more and more about what goes on in the brain
Lex Fridman (1:58:57.120)
and we can sort of train the brain.
Lex Fridman (1:59:00.520)
I don't know if you remember the Palm Pilot.
Manolis Kellis (1:59:03.600)
Yeah, Palm Pilot, yeah.
Lex Fridman (1:59:04.720)
Remember this whole sort of alphabet that they had created?
Lex Fridman (1:59:08.440)
Am I thinking of the same thing?
Lex Fridman (1:59:10.920)
It's basically, you had a little pen
Lex Fridman (1:59:13.280)
and for every character, you had a little scribble
Lex Fridman (1:59:17.000)
that was unique that the machine could understand.
Lex Fridman (1:59:19.800)
And that instead of trying the machine
Lex Fridman (1:59:22.520)
and trying to teach the machine
Manolis Kellis (1:59:23.640)
to recognize human characters,
Lex Fridman (1:59:25.200)
you had basically, they figured out
Manolis Kellis (1:59:27.200)
that it's better and easier to train humans
Lex Fridman (1:59:29.960)
to create human like characters
Manolis Kellis (1:59:31.840)
that the machine is better at recognizing.
Lex Fridman (1:59:34.760)
So in the same way, I think what will happen
Manolis Kellis (1:59:38.320)
is that humans will be trained
Lex Fridman (1:59:40.600)
to be able to create the mind pattern
Manolis Kellis (1:59:43.200)
that the machine will respond to
Lex Fridman (1:59:45.160)
before the machine truly comprehends our thoughts.
Lex Fridman (1:59:47.760)
So the first human brain interfaces
Lex Fridman (1:59:50.160)
will be tricking humans to speak the machine language
Manolis Kellis (1:59:53.640)
where with the right set of electrodes,
Lex Fridman (1:59:55.640)
I can sort of trick my brain into doing this.
Lex Fridman (1:59:57.640)
And this is the same way that many people teach,
Lex Fridman (20:00.860)
Others may have been misleading the public
Manolis Kellis (20:02.820)
for the greater good, such as don't wear masks
Lex Fridman (20:05.540)
because we don't want the mask to run out.
Manolis Kellis (20:07.220)
I mean, that was very silly in my view
Lex Fridman (20:09.060)
and a very big mistake.
Lex Fridman (20:11.360)
But the spread of knowledge
Lex Fridman (20:15.060)
from the scientific community was phenomenal.
Lex Fridman (20:17.300)
And some people will point out to bogus articles
Lex Fridman (20:20.680)
that snuck in and made the front page.
Manolis Kellis (20:22.860)
Yeah, they did.
Lex Fridman (20:23.700)
But within 24 hours, they were debunked
Lex Fridman (20:26.100)
and went out of the front page.
Lex Fridman (20:27.500)
And I think that's the beauty of science today.
Manolis Kellis (20:30.160)
The fact that it's not, oh, knowledge is fixed.
Lex Fridman (20:33.220)
It's the ability to embrace that nothing is permanent
Manolis Kellis (20:36.980)
when it comes to knowledge,
Lex Fridman (20:37.860)
that everything is the current best hypothesis
Lex Fridman (20:40.020)
and the current best model that best fits the current data
Lex Fridman (20:42.900)
and the willingness to be wrong.
Manolis Kellis (20:45.780)
The expectation that we're gonna be wrong
Lex Fridman (20:48.260)
and the celebration of success based on
Lex Fridman (20:50.660)
how long was I not proven wrong for,
Lex Fridman (20:52.700)
rather than, wow, I was exactly right.
Manolis Kellis (20:55.620)
Because no one is gonna be exactly right
Lex Fridman (20:57.020)
with partial knowledge.
Lex Fridman (20:58.940)
But the arc towards perfection,
Lex Fridman (21:03.140)
I think is so much more important
Manolis Kellis (21:05.280)
than how far you are in your first step.
Lex Fridman (21:08.700)
And I think that's what sort of
Manolis Kellis (21:10.380)
the current pandemic has taught us.
Lex Fridman (21:13.420)
The fact that, yeah, no, of course,
Manolis Kellis (21:14.780)
we're gonna make mistakes,
Lex Fridman (21:16.140)
but at least we're gonna learn from those mistakes
Lex Fridman (21:18.300)
and become better and learn better
Lex Fridman (21:20.340)
and spread information better.
Lex Fridman (21:21.340)
So if I were to answer the question of,
Lex Fridman (21:23.320)
where would you go to learn about coronavirus?
Manolis Kellis (21:27.700)
First textbook, it all starts with a textbook.
Lex Fridman (21:29.960)
Just open up a chapter on virology
Lex Fridman (21:32.580)
and how coronaviruses work.
Lex Fridman (21:34.420)
Then some basic epidemiology
Lex Fridman (21:36.940)
and sort of how pandemics have worked in the past.
Lex Fridman (21:39.860)
What are the basic principles surrounding
Lex Fridman (21:41.860)
these first wave, second wave?
Lex Fridman (21:43.480)
Why do they even exist?
Manolis Kellis (21:45.420)
Then understanding about growth,
Lex Fridman (21:47.240)
understanding about the R0 and RT
Manolis Kellis (21:50.260)
at various time points.
Lex Fridman (21:52.420)
And then understanding the means of spread,
Lex Fridman (21:55.260)
how it spreads from person to person.
Lex Fridman (21:57.340)
Then how does it get into your cells?
Manolis Kellis (22:00.080)
From when it gets into the cells,
Lex Fridman (22:01.580)
what are the paths that it takes?
Lex Fridman (22:03.260)
What are the cell types that express
Lex Fridman (22:05.180)
the particular ACE2 receptor?
Lex Fridman (22:07.300)
How is your immune system interacting with the virus?
Lex Fridman (22:09.940)
And once your immune system launches a defense,
Lex Fridman (22:12.300)
how is that helping or actually hurting your health?
Lex Fridman (22:15.520)
What about the cytokine storm?
Lex Fridman (22:16.920)
What are most people dying from?
Lex Fridman (22:18.560)
Why are the comorbidities
Lex Fridman (22:20.020)
and these risk factors even applying?
Lex Fridman (22:23.900)
What makes obese people respond more
Manolis Kellis (22:25.940)
or elderly people respond more to the virus
Lex Fridman (22:28.580)
while kids are completely,
Lex Fridman (22:32.500)
very often not even aware that they're spreading it?
Lex Fridman (22:36.380)
So I think there's some basic questions
Manolis Kellis (22:41.180)
that you would start from.
Lex Fridman (22:42.740)
And then I'm sorry to say,
Lex Fridman (22:44.340)
but Wikipedia is pretty awesome.
Lex Fridman (22:45.700)
Yeah, it is. Google is pretty awesome.
Manolis Kellis (22:47.780)
It used to be a time,
Lex Fridman (22:48.620)
it used to be a time maybe five years ago.
Manolis Kellis (22:50.500)
I forget when,
Lex Fridman (22:52.000)
but people kind of made fun of Wikipedia
Manolis Kellis (22:54.260)
for being an unreliable source.
Lex Fridman (22:57.260)
I never quite understood it.
Manolis Kellis (22:58.500)
I thought from the early days, it was pretty reliable
Lex Fridman (23:01.260)
or better than a lot of the alternatives.
Lex Fridman (23:03.660)
But at this point,
Lex Fridman (23:04.780)
it's kind of like a solid accessible survey paper
Manolis Kellis (23:08.340)
on every subject ever.
Lex Fridman (23:10.620)
There's an ascertainment bias and a writing bias.
Lex Fridman (23:14.620)
So I think this is related to sort of people saying,
Lex Fridman (23:17.820)
oh, so many nature papers are wrong.
Lex Fridman (23:20.720)
And they're like, why would you publish in nature?
Lex Fridman (23:22.540)
So many nature papers are wrong.
Lex Fridman (23:23.680)
And my answer is no, no, no.
Lex Fridman (23:26.120)
So many nature papers are scrutinized.
Lex Fridman (23:29.440)
And just because more of them are being proven wrong
Lex Fridman (23:31.980)
than in other articles is actually evidence
Manolis Kellis (23:35.380)
that they're actually better papers overall
Lex Fridman (23:37.060)
because they're being scrutinized at a rate
Manolis Kellis (23:39.140)
much higher than any other journal.
Lex Fridman (23:41.040)
So if you basically judge Wikipedia
Manolis Kellis (23:45.560)
by not the initial content,
Lex Fridman (23:49.820)
but by the number of revisions,
Manolis Kellis (23:52.420)
then of course it's gonna be the best source
Lex Fridman (23:53.900)
of knowledge eventually.
Manolis Kellis (23:55.300)
It's still very superficial.
Lex Fridman (23:57.120)
You then have to go into the review papers,
Manolis Kellis (23:58.780)
et cetera, et cetera, et cetera.
Lex Fridman (24:00.180)
But I mean, for most scientific topics,
Manolis Kellis (24:03.380)
it's extremely superficial,
Lex Fridman (24:05.060)
but it is quite authoritative
Manolis Kellis (24:07.660)
because it is the place that everybody likes to criticize
Lex Fridman (24:10.780)
as being wrong.
Manolis Kellis (24:11.660)
You say that it's superficial.
Lex Fridman (24:13.620)
And a lot of topics that I've studied a lot of,
Manolis Kellis (24:18.340)
I find it, I don't know if superficial is the right word.
Lex Fridman (24:24.380)
Because superficial kind of implies that it's not correct.
Manolis Kellis (24:27.700)
No, no, no.
Lex Fridman (24:29.100)
I don't mean any implication of it not being correct.
Manolis Kellis (24:31.660)
It's just superficial.
Lex Fridman (24:32.860)
It's basically only scratching the surface.
Manolis Kellis (24:35.540)
For depth, you don't go to Wikipedia.
Lex Fridman (24:37.140)
You go to the review articles.
Lex Fridman (24:38.340)
But it can be profound in the way that articles rarely,
Lex Fridman (24:41.860)
one of the frustrating things to me
Manolis Kellis (24:43.300)
about certain computer science,
Lex Fridman (24:46.580)
like in the machine learning world,
Manolis Kellis (24:48.380)
articles, they don't as often take the bigger picture view.
Lex Fridman (24:54.900)
There's a kind of data set and you show that it works
Lex Fridman (24:57.440)
and you kind of show that here's an architecture thing
Lex Fridman (24:59.660)
that creates an improvement and so on and so forth.
Lex Fridman (25:02.300)
But you don't say, well, what does this mean
Lex Fridman (25:05.300)
for the nature of intelligence for future data sets
Lex Fridman (25:08.580)
we haven't even thought about?
Lex Fridman (25:10.080)
Or if you were trying to implement this,
Manolis Kellis (25:11.940)
like if we took this data set of 100,000 examples
Lex Fridman (25:15.940)
and scale it to 100 billion examples with this method,
Manolis Kellis (25:19.260)
like look at the bigger picture,
Lex Fridman (25:21.220)
which is what a Wikipedia article would actually try to do,
Manolis Kellis (25:25.540)
which is like, what does this mean in the context
Lex Fridman (25:28.540)
of the broad field of computer vision or something like that?
Manolis Kellis (25:32.380)
Yeah, no, I agree with you completely, but it depends
Lex Fridman (25:35.340)
on the topic.
Manolis Kellis (25:36.180)
I mean, for some topics, there's been a huge amount of work.
Lex Fridman (25:38.420)
For other topics, it's just a stub.
Manolis Kellis (25:40.300)
So, you know.
Lex Fridman (25:41.580)
I got it.
Manolis Kellis (25:42.420)
Yeah.
Lex Fridman (25:43.240)
Well, yeah, actually the, which we'll talk on,
Manolis Kellis (25:46.380)
genomics was not great.
Lex Fridman (25:48.060)
Yeah, it's very shallow, yeah, yeah.
Manolis Kellis (25:50.460)
It's not wrong, it's just shallow.
Lex Fridman (25:51.860)
It's shallow.
Manolis Kellis (25:52.700)
Yeah, every time I criticize something,
Lex Fridman (25:54.700)
I should feel partly responsible.
Manolis Kellis (25:56.320)
Basically, if more people from my community went there
Lex Fridman (25:58.740)
and edited, it would not be shallow.
Manolis Kellis (26:01.120)
It's just that there's different modes of communication
Lex Fridman (26:04.020)
in different fields.
Lex Fridman (26:05.280)
And in some fields, the experts have embraced Wikipedia.
Lex Fridman (26:08.980)
In other fields, it's relegated.
Lex Fridman (26:11.180)
And perhaps the reason is that if it was any better
Lex Fridman (26:15.860)
to start with, people would invest more time.
Lex Fridman (26:18.000)
But if it's not great to start with,
Lex Fridman (26:19.980)
then you need a few initial pioneers who will basically
Manolis Kellis (26:22.860)
go in and say, ah, enough, we're just gonna fix that.
Lex Fridman (26:26.460)
And then I think it'll catch on much more.
Lex Fridman (26:29.120)
So if it's okay, before we go on to genomics,
Lex Fridman (26:32.220)
can we linger a little bit longer on the beauty
Lex Fridman (26:35.420)
of the human genome?
Lex Fridman (26:37.140)
You've given me a few notes.
Lex Fridman (26:38.580)
What else do you find beautiful about the human genome?
Lex Fridman (26:41.660)
So the last aspect of what makes the human genome unique,
Manolis Kellis (26:44.860)
in addition to the, you know, similarity and the differences
Lex Fridman (26:49.980)
and the individuality is that, so very early on,
Manolis Kellis (26:56.340)
people would basically say, oh, you don't do that
Lex Fridman (26:58.020)
experiment in human, you have to learn about that in fly,
Manolis Kellis (27:01.260)
or you have to learn about that in yeast first,
Lex Fridman (27:03.140)
or in mouse first, or in a primate first.
Lex Fridman (27:05.900)
And the human genome was in fact relegated to sort of,
Lex Fridman (27:09.040)
oh, the last place that you're gonna go
Manolis Kellis (27:11.020)
to learn something new.
Lex Fridman (27:12.640)
That has dramatically changed.
Lex Fridman (27:14.220)
And the reason that changed is human genetics.
Lex Fridman (27:18.620)
We are the species in the planet
Manolis Kellis (27:22.500)
that's the most studied right now.
Lex Fridman (27:24.660)
It's embarrassing to say that,
Lex Fridman (27:26.260)
but this was not the case a few years ago.
Lex Fridman (27:28.380)
It used to be, you know, first viruses, then bacteria,
Manolis Kellis (27:33.840)
then yeast, then the fruit fly and the worm,
Lex Fridman (27:37.880)
then the mouse, and eventually human was very far last.
Lex Fridman (27:42.420)
So it's embarrassing that it took us this long
Lex Fridman (27:44.740)
to focus on it, or the...
Manolis Kellis (27:46.580)
It's embarrassing that the model organisms
Lex Fridman (27:49.220)
have been taken over because of the power of human genetics.
Manolis Kellis (27:52.660)
That right now, it's actually simpler to figure out
Lex Fridman (27:55.420)
the phenotype of something by mining
Manolis Kellis (27:58.820)
this massive amount of human data
Lex Fridman (28:01.380)
than by going back to any of the other species.
Lex Fridman (28:04.020)
And the reason for that is that if you look
Lex Fridman (28:05.540)
at the natural variation that happens
Manolis Kellis (28:07.360)
in a population of seven billion,
Lex Fridman (28:09.700)
you basically have a mutation in almost every nucleotide.
Lex Fridman (28:13.380)
So every nucleotide you wanna perturb,
Lex Fridman (28:15.680)
you can go find a living, breathing human being
Lex Fridman (28:18.780)
and go test the function of that nucleotide
Lex Fridman (28:20.360)
by sort of searching the database and finding that person.
Lex Fridman (28:22.620)
Wait, why is that embarrassing?
Lex Fridman (28:23.660)
It's a beautiful data set.
Manolis Kellis (28:24.700)
It's a beautiful data set.
Lex Fridman (28:26.380)
It's embarrassing for the model organism.
Manolis Kellis (28:29.300)
For the flies.
Lex Fridman (28:30.140)
Yeah, exactly.
Manolis Kellis (28:30.980)
I mean, do you feel on a small tangent,
Lex Fridman (28:34.940)
is there something of value in the genome of a fly
Lex Fridman (28:40.060)
and other of these model organisms that you miss
Lex Fridman (28:43.740)
that we wish we would be looking at deeper?
Lex Fridman (28:47.420)
So directed perturbation, of course.
Lex Fridman (28:49.900)
So I think the place where humans are still lagging
Manolis Kellis (28:54.140)
is the fact that in an animal model,
Lex Fridman (28:55.700)
you can go and say,
Manolis Kellis (28:56.540)
well, let me knock out this gene completely
Lex Fridman (28:58.620)
and let me knock out these three genes completely.
Lex Fridman (29:00.580)
And the moment you get into combinatorics,
Lex Fridman (29:02.780)
it's something you can't do in the human
Manolis Kellis (29:04.180)
because there just simply aren't enough humans
Lex Fridman (29:05.980)
on the planet.
Lex Fridman (29:07.060)
And again, let me be honest,
Lex Fridman (29:08.820)
we haven't sequenced all seven billion people.
Manolis Kellis (29:11.180)
It's not like we have every mutation,
Lex Fridman (29:12.820)
but we know that there's a carrier out there.
Lex Fridman (29:15.060)
So if you look at the trend and the speed
Lex Fridman (29:17.500)
with which human genetics has progressed,
Manolis Kellis (29:19.460)
we can now find thousands of genes involved
Lex Fridman (29:23.300)
in human cognition, in human psychology,
Manolis Kellis (29:27.060)
in the emotions and the feelings
Lex Fridman (29:29.100)
that we used to think are uniquely learned.
Manolis Kellis (29:31.780)
It turns out there's a genetic basis to a lot of that.
Lex Fridman (29:34.380)
So the human genome has continued to elucidate
Manolis Kellis (29:42.540)
through these studies of genetic variation,
Lex Fridman (29:44.860)
so many different processes that we previously thought
Manolis Kellis (29:47.500)
were something like free will.
Lex Fridman (29:52.300)
Free will is this beautiful concept
Manolis Kellis (29:54.260)
that humans have had for a long time.
Lex Fridman (29:58.060)
In the end, it's just a bunch of chemical reactions
Manolis Kellis (29:59.860)
happening in your brain.
Lex Fridman (2:00:00.240)
like learn to control artificial limbs.
Manolis Kellis (2:00:02.960)
You basically try a bunch of stuff
Lex Fridman (2:00:04.520)
and eventually you figure out how your limbs work.
Manolis Kellis (2:00:06.880)
That might not be very different
Lex Fridman (2:00:08.200)
from how humans learn to use their natural limbs
Manolis Kellis (2:00:11.480)
when they first grow up.
Lex Fridman (2:00:13.080)
Basically, you have these, you know,
Manolis Kellis (2:00:14.920)
neoteny period of, you know,
Lex Fridman (2:00:17.960)
this puddle of soup inside your brain,
Manolis Kellis (2:00:21.320)
trying to figure out how to even make neural connections
Lex Fridman (2:00:23.840)
before you're born and then learning sounds
Manolis Kellis (2:00:27.280)
in utero of, you know, all kinds of echoes
Lex Fridman (2:00:31.480)
and, you know, eventually getting out in the real world.
Lex Fridman (2:00:35.840)
And I don't know if you've seen newborns,
Lex Fridman (2:00:37.280)
but they just stare around a lot.
Manolis Kellis (2:00:39.840)
You know, one way to think about this
Lex Fridman (2:00:41.680)
as a machine learning person is,
Manolis Kellis (2:00:43.000)
oh, they're just training their edge detectors.
Lex Fridman (2:00:46.080)
And eventually they figure out
Lex Fridman (2:00:47.320)
how to train their edge detectors.
Lex Fridman (2:00:48.680)
They work through the second layer of the visual cortex
Lex Fridman (2:00:50.800)
and the third layer and so on and so forth.
Lex Fridman (2:00:52.640)
And you basically have this learning
Lex Fridman (2:00:58.320)
how to control your limbs
Lex Fridman (2:00:59.360)
that probably comes at the same time.
Manolis Kellis (2:01:01.000)
You're sort of, you know, throwing random things there
Lex Fridman (2:01:03.280)
and you realize that, oh, wow,
Manolis Kellis (2:01:04.720)
when I do this thing, my limb moves.
Lex Fridman (2:01:08.200)
Let's do the following experiment.
Manolis Kellis (2:01:09.240)
Take a breath.
Lex Fridman (2:01:11.880)
What muscles did you flex?
Lex Fridman (2:01:13.480)
Now take another breath and think what muscles do I flex?
Lex Fridman (2:01:16.600)
The first thing that you're thinking
Manolis Kellis (2:01:17.920)
when you're taking a breath
Lex Fridman (2:01:19.840)
is the impact that it has on your lungs.
Manolis Kellis (2:01:22.320)
You're like, oh, I'm now gonna increase my lungs
Lex Fridman (2:01:24.080)
or I'm not gonna bring air in.
Lex Fridman (2:01:25.440)
But what you're actually doing
Lex Fridman (2:01:26.360)
is just changing your diaphragm.
Manolis Kellis (2:01:29.160)
That's not conscious, of course.
Lex Fridman (2:01:31.880)
You never think of the diaphragm as a thing.
Lex Fridman (2:01:34.920)
And why is that?
Lex Fridman (2:01:36.000)
That's probably the same reason
Lex Fridman (2:01:37.360)
why I think of moving my finger
Lex Fridman (2:01:38.800)
when I actually move my finger.
Manolis Kellis (2:01:40.520)
I think of the effect instead of actually thinking
Lex Fridman (2:01:42.520)
of whatever muscle is twitching
Manolis Kellis (2:01:44.080)
that actually causes my finger to move.
Lex Fridman (2:01:46.440)
So we basically in our first years of life
Manolis Kellis (2:01:49.240)
build up this massive lookup table
Lex Fridman (2:01:52.360)
between whatever neuronal firing we do
Lex Fridman (2:01:55.400)
and whatever action happens in our body that we control.
Lex Fridman (2:02:00.880)
If you have a kid grow up with a third limb,
Manolis Kellis (2:02:04.280)
I'm sure they'll figure out how to control them
Lex Fridman (2:02:06.600)
probably at the same rate as their natural limbs.
Lex Fridman (2:02:09.440)
And a lot of the work would be done by the...
Lex Fridman (2:02:13.320)
If a third limb is a computer,
Manolis Kellis (2:02:15.520)
you kind of have a, not a faith, but a thought
Lex Fridman (2:02:20.440)
that the brain might be able to figure out...
Manolis Kellis (2:02:24.360)
The plasticity would come from the brain.
Lex Fridman (2:02:26.840)
The brain would be cleverer than the machine at first.
Manolis Kellis (2:02:28.960)
When I talk about a third limb,
Lex Fridman (2:02:29.960)
that's exactly what I'm saying, an artificial limb
Manolis Kellis (2:02:32.240)
that basically just controls your mouse while you're typing.
Lex Fridman (2:02:35.640)
Perfectly natural thing.
Manolis Kellis (2:02:36.600)
I mean, again, in a few hundred years.
Lex Fridman (2:02:40.320)
Maybe sooner than that.
Lex Fridman (2:02:41.600)
But basically, as long as the machine is consistent
Lex Fridman (2:02:46.040)
in the way that it will respond to your brain impulses,
Manolis Kellis (2:02:49.760)
you'll figure out how to control that
Lex Fridman (2:02:51.680)
and you could play tennis with your third limb.
Lex Fridman (2:02:53.920)
And let me go back to consistency.
Lex Fridman (2:02:57.480)
People who have dramatic accidents
Manolis Kellis (2:03:01.280)
that basically take out a whole chunk of their brain
Lex Fridman (2:03:03.920)
can be taught to coopt other parts of the brain
Manolis Kellis (2:03:07.000)
to then control that part.
Lex Fridman (2:03:08.560)
You can basically build up that tissue again
Lex Fridman (2:03:10.840)
and eventually train your body how to walk again
Lex Fridman (2:03:13.480)
and how to read again and how to play again
Lex Fridman (2:03:15.400)
and how to think again, how to speak a language again,
Lex Fridman (2:03:17.160)
et cetera.
Lex Fridman (2:03:18.080)
So there's a massive amount of malleability
Lex Fridman (2:03:21.280)
that happens naturally in our way of controlling our body,
Manolis Kellis (2:03:26.600)
our brain, our thoughts, our vocal cords, our limbs,
Lex Fridman (2:03:29.720)
et cetera.
Lex Fridman (2:03:30.760)
And human machine interfaces are inevitable
Lex Fridman (2:03:35.640)
if we sort of figure out how to read these electric impulses,
Lex Fridman (2:03:39.240)
but the resolution at which we can understand human thought
Lex Fridman (2:03:43.400)
right now is nil, is ridiculous.
Lex Fridman (2:03:46.560)
So how are human thoughts encoded?
Lex Fridman (2:03:49.120)
It's basically combinations of neurons that cofire
Lex Fridman (2:03:53.560)
and these create these things called engrams
Lex Fridman (2:03:55.720)
that eventually form memories and so on and so forth.
Manolis Kellis (2:03:58.920)
We know nothing of all that stuff.
Lex Fridman (2:04:01.920)
So before we can actually read into your brain
Manolis Kellis (2:04:05.600)
that you wanna build a program
Lex Fridman (2:04:06.680)
that does this and this and this and that,
Manolis Kellis (2:04:08.920)
we need a lot of neuroscience.
Lex Fridman (2:04:10.960)
Well, so to push back on that,
Lex Fridman (2:04:13.480)
do you think it's possible that without understanding
Lex Fridman (2:04:16.680)
the functionally about the brain or from the neuroscience
Manolis Kellis (2:04:20.000)
or the cognitive science or psychology,
Lex Fridman (2:04:22.080)
whichever level of the brain we'll look at,
Lex Fridman (2:04:24.220)
do you think if we just connect them,
Lex Fridman (2:04:26.700)
just like per your previous point,
Manolis Kellis (2:04:29.200)
if we just have a high enough resolution
Lex Fridman (2:04:30.840)
between connection between a Wikipedia and your brain,
Manolis Kellis (2:04:34.400)
the brain will just figure it out with us understanding
Lex Fridman (2:04:38.160)
because that's one of the innovations of Neuralink
Manolis Kellis (2:04:40.320)
is they're increasing the number of connections
Lex Fridman (2:04:43.540)
to the brain to like several thousand,
Manolis Kellis (2:04:45.320)
which before was in the dozens or whatever.
Lex Fridman (2:04:48.280)
You're still off by a few orders of magnitude
Manolis Kellis (2:04:51.000)
on the order of seven.
Lex Fridman (2:04:52.160)
Right, but the thing is, the hope is if you increase
Manolis Kellis (2:04:57.480)
that number more and more and more,
Lex Fridman (2:04:58.800)
maybe you don't need to understand anything
Manolis Kellis (2:05:00.600)
about the actual how human thought
Lex Fridman (2:05:03.780)
is represented in the brain.
Manolis Kellis (2:05:04.960)
You can just let it figure it out by itself.
Lex Fridman (2:05:08.520)
Keanu Reeves waking up and saying, I know cook food.
Manolis Kellis (2:05:10.680)
Yeah, exactly.
Lex Fridman (2:05:13.160)
So yeah, sure.
Manolis Kellis (2:05:14.600)
You don't have faith in the plasticity of the brain
Lex Fridman (2:05:16.720)
to that degree.
Manolis Kellis (2:05:18.240)
It's not about brain plasticity.
Lex Fridman (2:05:19.840)
It's about the input aspect.
Manolis Kellis (2:05:21.880)
Basically, I think on the output aspect,
Lex Fridman (2:05:23.720)
being able to control a machine is something
Manolis Kellis (2:05:25.440)
that you can probably train your neural impulses
Lex Fridman (2:05:28.440)
that you're sending out to sort of match
Manolis Kellis (2:05:30.940)
whatever response you see in the environment.
Lex Fridman (2:05:33.280)
If this thing moved every single time I thought
Manolis Kellis (2:05:35.600)
a particular thought, then I could figure out,
Lex Fridman (2:05:37.340)
I could hack my way into moving this thing
Manolis Kellis (2:05:39.520)
with just a series of thoughts.
Lex Fridman (2:05:40.960)
I could think guitar, piano, tennis ball,
Lex Fridman (2:05:45.880)
and then this thing would be moving.
Lex Fridman (2:05:47.120)
And then I would just have the series of thoughts
Manolis Kellis (2:05:50.640)
that would sort of result in the impulses
Lex Fridman (2:05:52.640)
that will move this thing the way that I want it.
Lex Fridman (2:05:54.040)
And then eventually it'll become natural
Lex Fridman (2:05:55.560)
because I won't even think about it.
Manolis Kellis (2:05:57.640)
I mean, in the same way that we control our limbs
Lex Fridman (2:05:59.120)
in a very natural way, but babies don't do that.
Manolis Kellis (2:06:01.360)
Babies have to figure it out.
Lex Fridman (2:06:03.160)
And some of that is hard coded,
Lex Fridman (2:06:04.840)
but some of that is actually learned
Lex Fridman (2:06:06.800)
based on whatever soup of neurons you ended up with,
Manolis Kellis (2:06:10.320)
whatever connections you pruned them to,
Lex Fridman (2:06:13.440)
and eventually you were born with.
Manolis Kellis (2:06:15.360)
A lot of that is coded in the genome,
Lex Fridman (2:06:17.740)
but a huge chunk of that is stochastic.
Lex Fridman (2:06:19.680)
And sort of the way that you sort of create
Lex Fridman (2:06:21.320)
all these neurons, they migrate, they form connections,
Manolis Kellis (2:06:23.440)
they sort of spread out,
Lex Fridman (2:06:25.140)
they have particular branching patterns,
Lex Fridman (2:06:26.520)
but then the connectivity itself,
Lex Fridman (2:06:28.200)
unique in every single new person.
Manolis Kellis (2:06:30.120)
All this to say that on the output side,
Lex Fridman (2:06:34.000)
absolutely, I'm very, very, you know,
Manolis Kellis (2:06:36.920)
hopeful that we can have machines
Lex Fridman (2:06:38.640)
that read thousands of these neuronal connections
Manolis Kellis (2:06:41.920)
on the output side, but on the input side, oh boy.
Lex Fridman (2:06:47.960)
I don't expect any time in the near future
Manolis Kellis (2:06:51.240)
we'll be able to sort of send a series of impulses
Lex Fridman (2:06:53.400)
that will tell me, oh, earth to sun distance,
Manolis Kellis (2:06:56.280)
7.5 million, et cetera, et cetera.
Lex Fridman (2:06:58.960)
Like nowhere.
Manolis Kellis (2:07:00.720)
I mean, I think language will still be the input way
Lex Fridman (2:07:04.480)
rather than sort of any kind of more complex.
Manolis Kellis (2:07:07.360)
It's a really interesting notion
Lex Fridman (2:07:08.760)
that the ambiguity of language is a feature.
Lex Fridman (2:07:12.600)
And we evolved for millions of years
Lex Fridman (2:07:16.520)
to take advantage of that ambiguity.
Manolis Kellis (2:07:19.520)
Exactly.
Lex Fridman (2:07:20.520)
And yet no one teaches us the subtle differences
Manolis Kellis (2:07:23.380)
between words that are near cognates,
Lex Fridman (2:07:26.080)
and yet evoke so much more than, you know,
Manolis Kellis (2:07:29.320)
one from the other.
Lex Fridman (2:07:30.760)
And yet, you know, when you're choosing words
Manolis Kellis (2:07:34.520)
from a list of 20 synonyms,
Lex Fridman (2:07:36.840)
you know exactly the connotation
Manolis Kellis (2:07:38.400)
of every single one of them.
Lex Fridman (2:07:40.040)
And that's something that, you know, is there.
Lex Fridman (2:07:42.600)
So yes, there's ambiguity,
Lex Fridman (2:07:45.120)
but there's all kinds of connotations.
Lex Fridman (2:07:46.800)
And in the way that we select our words,
Lex Fridman (2:07:48.880)
we have so much baggage that we're sending along,
Manolis Kellis (2:07:51.320)
the way that we're emoting,
Lex Fridman (2:07:52.960)
the way that we're moving our hands
Manolis Kellis (2:07:54.720)
every single time we speak,
Lex Fridman (2:07:56.080)
the, you know, the pauses, the eye contact, et cetera.
Lex Fridman (2:07:58.880)
So much higher baud rate than just a vocal,
Lex Fridman (2:08:01.800)
you know, string of characters.
Manolis Kellis (2:08:04.040)
Well, let me just take a small tangent on that.
Lex Fridman (2:08:07.120)
Oh, tangent?
Manolis Kellis (2:08:07.960)
We haven't done that yet.
Lex Fridman (2:08:08.800)
It's a good idea.
Manolis Kellis (2:08:09.640)
Let's do a tangent.
Lex Fridman (2:08:10.480)
We'll return to the origin of life after.
Manolis Kellis (2:08:16.200)
So, I mean, you're Greek,
Lex Fridman (2:08:17.840)
but I'm going on this personal journey.
Manolis Kellis (2:08:20.880)
I'm going to Paris for the explicit purpose
Lex Fridman (2:08:25.120)
of talking to one of the most famous,
Manolis Kellis (2:08:29.360)
a couple who's a famous translators of Russian literature,
Lex Fridman (2:08:33.200)
Dostoevsky, Tolstoy, and they go,
Manolis Kellis (2:08:36.280)
that's their art is the translation.
Lex Fridman (2:08:38.440)
And everything I've learned about the translation art,
Manolis Kellis (2:08:44.320)
it makes me feel,
Lex Fridman (2:08:46.120)
it's so profound in a way that's so much more profound
Manolis Kellis (2:08:53.240)
than the natural language processing papers
Lex Fridman (2:08:55.400)
I read in the machine learning community,
Manolis Kellis (2:08:57.440)
that there's such depth to language
Lex Fridman (2:09:00.440)
that I don't know what to do with.
Manolis Kellis (2:09:03.160)
I don't know if you've experienced that in your own life
Lex Fridman (2:09:05.720)
with knowing multiple languages.
Manolis Kellis (2:09:08.760)
I don't know what to,
Lex Fridman (2:09:09.840)
I don't know how to make sense of it,
Lex Fridman (2:09:11.680)
but there's so much loss in translation
Lex Fridman (2:09:13.600)
between Russian and English,
Lex Fridman (2:09:15.320)
and getting a sense of that.
Lex Fridman (2:09:17.440)
Like, for example,
Manolis Kellis (2:09:19.640)
there's like just taking a single sentence
Lex Fridman (2:09:22.000)
from Dostoevsky, and like, there's a lot of them.
Manolis Kellis (2:09:25.440)
You could talk for hours
Lex Fridman (2:09:27.160)
about how to translate that sentence properly.
Manolis Kellis (2:09:30.120)
That captures the meaning, the period,
Lex Fridman (2:09:34.360)
the culture, the humor, the wit,
Manolis Kellis (2:09:36.560)
the suffering that was in the context of the time,
Lex Fridman (2:09:39.760)
all of that could be a single sentence.
Manolis Kellis (2:09:42.280)
You could talk forever about what it takes
Lex Fridman (2:09:46.120)
to translate that correctly.
Manolis Kellis (2:09:47.160)
I don't know what to do with that.
Lex Fridman (2:09:48.720)
So being Greek, it's very hard for me
Manolis Kellis (2:09:51.640)
to think of a sentence or even a word
Lex Fridman (2:09:54.480)
without going into the full etymology of that word,
Manolis Kellis (2:09:59.000)
breaking up every single atom of that sentence
Lex Fridman (2:10:04.840)
and every single atom of these words
Lex Fridman (2:10:07.080)
and rebuilding it back up.
Lex Fridman (2:10:09.800)
I have three kids.
Lex Fridman (2:10:11.200)
And the way that I teach them Greek
Lex Fridman (2:10:13.680)
is the same way that, you know,
Manolis Kellis (2:10:16.240)
the documentary I was mentioning earlier
Lex Fridman (2:10:17.640)
about sort of understanding the deep roots
Manolis Kellis (2:10:19.720)
of all of these, you know, words.
Lex Fridman (2:10:23.880)
And it's very interesting
Manolis Kellis (2:10:29.120)
that every single time I hear a new word
Lex Fridman (2:10:31.320)
that I've never heard before,
Manolis Kellis (2:10:33.000)
I go and figure out the etymology of that word
Lex Fridman (2:10:34.720)
because I will never appreciate that word
Manolis Kellis (2:10:36.760)
without understanding how it was initially formed.
Lex Fridman (2:10:40.160)
Interesting, but how does that help?
Manolis Kellis (2:10:42.080)
Because that's not the full picture.
Lex Fridman (2:10:44.080)
No, no, of course, of course.
Lex Fridman (2:10:44.920)
But what I'm trying to say is that knowing the components
Lex Fridman (2:10:48.360)
teaches you about the context of the formation of that word
Lex Fridman (2:10:52.280)
and sort of the original usage of that word.
Lex Fridman (2:10:54.880)
And then of course the word takes new meaning
Manolis Kellis (2:10:57.360)
as you create it, you know, from its parts.
Lex Fridman (2:11:00.840)
And that meaning then gets augmented.
Lex Fridman (2:11:04.120)
And two synonyms that sort of have different roots
Lex Fridman (2:11:08.160)
will actually have implications
Manolis Kellis (2:11:09.200)
that carry a lot of that baggage
Lex Fridman (2:11:11.440)
of the historical provenance of these words.
Lex Fridman (2:11:14.240)
So before working on genome evolution,
Lex Fridman (2:11:16.640)
my passion was evolution of language
Lex Fridman (2:11:19.920)
and sort of tracing cognates across different languages
Lex Fridman (2:11:24.640)
through their etymologies.
Manolis Kellis (2:11:27.280)
That's fascinating that there's parallels between,
Lex Fridman (2:11:30.320)
I mean, the idea that there's evolutionary dynamics
Manolis Kellis (2:11:34.280)
to our language.
Lex Fridman (2:11:35.520)
Yeah, every single word that you utter, parallels, parallels.
Lex Fridman (2:11:41.560)
What does parallels mean?
Lex Fridman (2:11:42.680)
Para means side by side.
Manolis Kellis (2:11:44.800)
Alleles from alleles, which means identical twins.
Lex Fridman (2:11:48.760)
Parallels.
Manolis Kellis (2:11:49.600)
I mean, name any word and there's so much baggage,
Lex Fridman (2:11:53.200)
so much beauty in how that word came to be
Lex Fridman (2:11:56.880)
and how this word took a new meaning
Lex Fridman (2:11:58.920)
than the sum of its parts.
Manolis Kellis (2:12:02.240)
Yeah, and there's just, there's so many different words
Lex Fridman (2:12:05.440)
that are just words.
Manolis Kellis (2:12:06.280)
They don't have any physical grounding.
Lex Fridman (2:12:08.800)
And now you take these words
Lex Fridman (2:12:10.240)
and you weave them into a sentence.
Lex Fridman (2:12:13.600)
The emotional invocations of that weaving are fathomless.
Lex Fridman (2:12:19.280)
And all of those emotions all live in the brains of humans.
Lex Fridman (2:12:25.480)
In the eye of the beholder.
Manolis Kellis (2:12:28.680)
No, seriously, you have to embrace this concept
Lex Fridman (2:12:30.840)
of the eye of the beholder.
Manolis Kellis (2:12:32.440)
It's the conceptualization that nothing takes meaning
Lex Fridman (2:12:37.960)
with one person creating it.
Manolis Kellis (2:12:39.400)
Everything takes meaning in the receiving end
Lex Fridman (2:12:42.480)
and the emergent properties of these communication networks
Manolis Kellis (2:12:47.760)
where every single, you know,
Lex Fridman (2:12:49.320)
if you look at the network of our cells
Lex Fridman (2:12:50.960)
and how they're communicating with each other,
Lex Fridman (2:12:52.480)
every cell has its own code.
Manolis Kellis (2:12:54.200)
This code is modulated by the epigenome.
Lex Fridman (2:12:56.200)
This creates a bunch of different cell types.
Manolis Kellis (2:12:57.960)
Each cell type now has its own identity.
Lex Fridman (2:12:59.960)
Yet they all have the common root of the stem cells
Manolis Kellis (2:13:02.200)
that sort of led to them.
Lex Fridman (2:13:04.760)
Each of these identities is now communicating
Manolis Kellis (2:13:06.600)
with each other.
Lex Fridman (2:13:08.120)
They take meaning in their interaction.
Manolis Kellis (2:13:11.800)
There's an emergent property that comes
Lex Fridman (2:13:13.760)
from a bunch of cells being together
Manolis Kellis (2:13:15.680)
that is not in any one of the parts.
Lex Fridman (2:13:17.920)
If you look at neurons communicating,
Manolis Kellis (2:13:19.320)
again, these engrams don't exist in any one neuron.
Lex Fridman (2:13:23.360)
They exist in the connection and the combination of neurons.
Lex Fridman (2:13:26.480)
And the meaning of the words that I'm telling you
Lex Fridman (2:13:29.040)
is empty until it reaches you
Lex Fridman (2:13:31.880)
and it affects you in a very different way
Lex Fridman (2:13:34.120)
than it affects whoever's listening
Manolis Kellis (2:13:35.440)
to this conversation now.
Lex Fridman (2:13:37.520)
Because of the emotional baggage that I've grown up with,
Manolis Kellis (2:13:40.400)
that you've grown up with, and that they've grown up with.
Lex Fridman (2:13:43.280)
And that's, I think, the magic of translation.
Manolis Kellis (2:13:46.800)
If you start thinking of translation
Lex Fridman (2:13:48.720)
as just simply capturing that emotional set of reactions
Manolis Kellis (2:13:53.720)
that you evoke, you need a different set of words
Lex Fridman (2:13:57.880)
to evoke that same set of reactions to a French person
Manolis Kellis (2:14:01.240)
than to a Russian person,
Lex Fridman (2:14:02.760)
because of the baggage of the culture that we grew up in.
Manolis Kellis (2:14:05.480)
Yeah, I mean, there's...
Lex Fridman (2:14:07.320)
So basically, you shouldn't find the best word.
Manolis Kellis (2:14:10.440)
Sometimes it's a completely different sentence structure
Lex Fridman (2:14:13.120)
that you will need,
Manolis Kellis (2:14:15.160)
matched to the cultural context
Lex Fridman (2:14:18.960)
of the target audience that you have.
Manolis Kellis (2:14:20.560)
Yeah, there's a lot of different words
Lex Fridman (2:14:22.360)
in the target audience that you have.
Manolis Kellis (2:14:23.800)
Yeah, it's, I mean, you're just...
Lex Fridman (2:14:26.480)
I usually don't think about this,
Lex Fridman (2:14:27.720)
but right now, there's this feeling,
Lex Fridman (2:14:30.000)
as a reminder, that it's just you and I talking,
Lex Fridman (2:14:32.680)
but there's several hundred thousand people
Lex Fridman (2:14:35.280)
will listen to this.
Manolis Kellis (2:14:36.440)
There's some guy in Russia right now running,
Lex Fridman (2:14:40.840)
like in Moscow, listening to us.
Manolis Kellis (2:14:44.120)
There's somebody in India, I guarantee you.
Lex Fridman (2:14:46.680)
There's somebody in China and South America.
Manolis Kellis (2:14:48.680)
There's somebody in Texas,
Lex Fridman (2:14:51.240)
they all have different...
Manolis Kellis (2:14:53.120)
Emotional baggage.
Lex Fridman (2:14:54.120)
They probably got angry earlier on
Manolis Kellis (2:14:56.160)
about the whole discussion about coronavirus
Lex Fridman (2:14:58.360)
and about some aspect of it.
Manolis Kellis (2:15:02.080)
Yeah, and there's that network effect that's...
Lex Fridman (2:15:06.960)
It's a beautiful thing.
Lex Fridman (2:15:08.000)
And this lateral transfer of information,
Lex Fridman (2:15:10.880)
that's what makes the collective, quote unquote,
Manolis Kellis (2:15:12.920)
genome of humanity so unique from any other species.
Lex Fridman (2:15:17.920)
Yeah.
Lex Fridman (2:15:19.920)
So you somehow miraculously wrapped it back
Lex Fridman (2:15:22.640)
to the very beginning of when we were talking
Manolis Kellis (2:15:25.160)
about the beauty of the human genome.
Lex Fridman (2:15:29.120)
So I think this is the right time,
Manolis Kellis (2:15:31.240)
unless we wanna go for a six to eight hour conversation.
Lex Fridman (2:15:34.880)
We're gonna have to talk again,
Lex Fridman (2:15:36.000)
but I think for now, to wrap it up,
Lex Fridman (2:15:39.120)
this is the right time to talk about
Manolis Kellis (2:15:41.080)
the biggest, most ridiculous question of all,
Lex Fridman (2:15:44.960)
meaning of life.
Manolis Kellis (2:15:45.900)
Off mic, you mentioned to me
Lex Fridman (2:15:47.340)
that you had your 42nd birthday.
Manolis Kellis (2:15:52.480)
42nd being a very special, absurdly special number.
Lex Fridman (2:15:58.480)
And you had a kind of get together with friends
Manolis Kellis (2:16:03.040)
to discuss the meaning of life.
Lex Fridman (2:16:04.400)
So let me ask you,
Manolis Kellis (2:16:05.920)
in your, as a biologist, as a computer scientist,
Lex Fridman (2:16:09.800)
and as a human, what is the meaning of life?
Manolis Kellis (2:16:14.640)
I've been asking this question for a long time,
Lex Fridman (2:16:18.880)
ever since my 42nd birthday,
Lex Fridman (2:16:21.160)
but well before that,
Lex Fridman (2:16:22.080)
in even planning the meaning of life symposium.
Lex Fridman (2:16:25.280)
And symposium, sim means together,
Lex Fridman (2:16:29.800)
posy actually means to drink together.
Lex Fridman (2:16:31.520)
So symposium is actually a drinking party.
Lex Fridman (2:16:33.560)
So the meaning.
Lex Fridman (2:16:36.040)
Can you actually elaborate about this meaning of life
Lex Fridman (2:16:37.880)
symposium that you put together?
Manolis Kellis (2:16:39.480)
It's like the most genius idea I've ever heard.
Lex Fridman (2:16:42.280)
So 42 is obviously the answer to life,
Manolis Kellis (2:16:44.640)
the universe and everything,
Lex Fridman (2:16:45.600)
from the Hitchhiker's Guide to the Galaxy.
Lex Fridman (2:16:47.640)
And as I was turning 42,
Lex Fridman (2:16:49.560)
I've had the theme for every one of my birthdays.
Manolis Kellis (2:16:51.800)
When I was turning 32, it's one, zero, zero, zero, zero, zero
Lex Fridman (2:16:55.640)
in binary.
Lex Fridman (2:16:56.640)
So I celebrated my 100,000th binary birthday,
Lex Fridman (2:17:00.080)
and I had a theme of going back 100,000 years,
Manolis Kellis (2:17:03.640)
let's dress something in the last 100,000 years.
Lex Fridman (2:17:07.160)
Anyway, it was, I've always had these.
Manolis Kellis (2:17:09.600)
It's such an interesting human being.
Lex Fridman (2:17:12.320)
Okay, that's awesome.
Manolis Kellis (2:17:13.160)
I've always had these sort of numerology
Lex Fridman (2:17:17.400)
related announcements for my birthday parties.
Lex Fridman (2:17:21.800)
So what came out of that meaning of life symposium
Lex Fridman (2:17:27.360)
is that I basically asked 42 of my colleagues,
Manolis Kellis (2:17:29.720)
42 of my friends, 42 of my collaborators,
Lex Fridman (2:17:33.080)
to basically give seven minutes species
Manolis Kellis (2:17:35.520)
on the meaning of life, each from their perspective.
Lex Fridman (2:17:38.520)
And I really encourage you to go there
Manolis Kellis (2:17:40.600)
because it's mind boggling
Lex Fridman (2:17:42.560)
that every single person said a different answer.
Manolis Kellis (2:17:46.280)
Every single person started with,
Lex Fridman (2:17:48.480)
I don't know what the meaning of life is, but,
Lex Fridman (2:17:50.920)
and then give this beautifully eloquently answer,
Lex Fridman (2:17:54.240)
eloquent answer.
Lex Fridman (2:17:55.440)
And they were all different,
Lex Fridman (2:17:57.320)
but they all were consistent with each other
Lex Fridman (2:18:01.360)
and mutually synergistic and together forming
Lex Fridman (2:18:04.360)
a beautiful view of what it means to be human in many ways.
Manolis Kellis (2:18:08.560)
Some people talked about the loss of their loved one,
Lex Fridman (2:18:12.280)
their life partner for many, many years
Lex Fridman (2:18:14.520)
and how their life changed through that.
Lex Fridman (2:18:16.520)
Some people talked about the origin of life.
Manolis Kellis (2:18:19.280)
Some people talked about the difference
Lex Fridman (2:18:21.080)
between purpose and meaning.
Manolis Kellis (2:18:24.160)
I'll maybe quote one of the answers,
Lex Fridman (2:18:28.600)
which is this linguistics professor,
Manolis Kellis (2:18:30.880)
friend of mine at Harvard, who basically said,
Lex Fridman (2:18:35.880)
that she was gonna, she's Greek as well.
Lex Fridman (2:18:37.800)
And she said, I will give a very Pythian answer.
Lex Fridman (2:18:40.120)
So Pythia was the Oracle of Delphi,
Manolis Kellis (2:18:42.960)
who would basically give these very cryptic answers,
Lex Fridman (2:18:45.280)
very short, but interpretable in many different ways.
Manolis Kellis (2:18:48.320)
There was this whole set of priests
Lex Fridman (2:18:50.480)
who were tasked with interpreting what Pythia had said.
Lex Fridman (2:18:53.440)
And very often you would not get a clean interpretation,
Lex Fridman (2:18:56.440)
but she said, I will be like Pythia
Lex Fridman (2:18:59.000)
and give you a very short and multiply interpretable answer.
Lex Fridman (2:19:02.520)
But unlike her, I will actually also give you
Manolis Kellis (2:19:04.800)
three interpretations.
Lex Fridman (2:19:07.000)
And she said, the answer to the meaning of life
Manolis Kellis (2:19:09.800)
is become one.
Lex Fridman (2:19:12.840)
And the first interpretation is like a child,
Manolis Kellis (2:19:16.320)
become one year old with the excitement
Lex Fridman (2:19:18.720)
of discovering everything about the world.
Manolis Kellis (2:19:21.400)
Second interpretation, in whatever you take on,
Lex Fridman (2:19:25.040)
become one, the first, the best, excel,
Manolis Kellis (2:19:28.840)
drive yourself to perfection for every one of your tasks
Lex Fridman (2:19:32.600)
and become one when people are separate,
Manolis Kellis (2:19:38.480)
become one, come together, learn to understand each other.
Lex Fridman (2:19:43.600)
Damn, that's an answer.
Lex Fridman (2:19:45.400)
And one way to summarize
Lex Fridman (2:19:46.880)
this whole meaning of life symposium
Manolis Kellis (2:19:48.760)
is that the very symposium was illustrating
Lex Fridman (2:19:52.920)
the quest for meaning,
Manolis Kellis (2:19:54.680)
which might itself be the meaning of life.
Lex Fridman (2:19:58.120)
This constant quest for something sublime,
Manolis Kellis (2:20:01.400)
something human, something intangible,
Lex Fridman (2:20:04.880)
some aspect of what defines us as a species
Lex Fridman (2:20:09.680)
and as an individual.
Lex Fridman (2:20:11.320)
Both the quest of me as a person through my own life,
Lex Fridman (2:20:16.360)
but the meaning of life could also be
Lex Fridman (2:20:19.200)
the meaning of all of life.
Lex Fridman (2:20:20.840)
What is the whole point of life?
Lex Fridman (2:20:22.040)
Why life?
Lex Fridman (2:20:22.880)
Why life itself?
Lex Fridman (2:20:24.480)
Because we've been talking about the history
Lex Fridman (2:20:26.720)
and evolution of life,
Lex Fridman (2:20:28.360)
but we haven't talked about why life in the first place?
Lex Fridman (2:20:31.080)
Is life inevitable?
Lex Fridman (2:20:32.520)
Is life part of physics?
Manolis Kellis (2:20:35.880)
Does life transcend physics
Lex Fridman (2:20:37.720)
by fighting against entropy,
Manolis Kellis (2:20:40.320)
by compartmentalizing and increasing concentrations
Lex Fridman (2:20:42.960)
rather than diluting away?
Manolis Kellis (2:20:45.320)
Is life a distinct entity in the universe
Lex Fridman (2:20:51.520)
beyond the traditional very simple physical rules
Manolis Kellis (2:20:55.080)
that govern gravity and electromagnetism
Lex Fridman (2:20:58.560)
and all of these forces?
Lex Fridman (2:21:00.600)
Is life another force?
Lex Fridman (2:21:02.120)
Is there a life force?
Manolis Kellis (2:21:03.120)
Is there a unique kind of set of principles that emerge,
Lex Fridman (2:21:05.920)
of course, built on top of the hardware of physics,
Lex Fridman (2:21:09.080)
but is it sort of a new layer of software
Lex Fridman (2:21:11.840)
or a new layer of a computer system?
Lex Fridman (2:21:14.400)
And so that's at the level of big questions.
Lex Fridman (2:21:18.480)
There's another aspect of gratitude
Manolis Kellis (2:21:21.200)
of basically what I like to say is,
Lex Fridman (2:21:27.000)
during this pandemic,
Manolis Kellis (2:21:27.920)
I've basically worked from 6 a.m. until 7 p.m.
Lex Fridman (2:21:30.800)
every single day, nonstop, including Saturday and Sunday.
Manolis Kellis (2:21:34.280)
I've basically broken all boundaries
Lex Fridman (2:21:36.440)
of where life, personal life begins
Lex Fridman (2:21:39.120)
and work life ends.
Lex Fridman (2:21:42.000)
And that has been exhilarating for me,
Manolis Kellis (2:21:46.280)
just the intellectual pleasure that I get
Lex Fridman (2:21:50.480)
from a day of exhaustion,
Manolis Kellis (2:21:53.840)
where at the end of the day, my brain is hurting.
Lex Fridman (2:21:55.520)
I'm telling my wife, wow, I was useful today.
Lex Fridman (2:22:00.360)
And there's a certain pleasure
Lex Fridman (2:22:04.720)
that comes from feeling useful.
Lex Fridman (2:22:08.360)
And there's a certain pleasure
Lex Fridman (2:22:09.880)
that comes from feeling grateful.
Lex Fridman (2:22:12.440)
So I've written this little sort of prayer for my kids
Lex Fridman (2:22:16.440)
to say at bedtime every night,
Manolis Kellis (2:22:19.520)
where they basically say,
Lex Fridman (2:22:21.000)
thank you, God, for all you have given me
Lex Fridman (2:22:24.720)
and give me the strength to give onto others
Lex Fridman (2:22:28.560)
with the same love that you have given onto me.
Manolis Kellis (2:22:33.160)
We as a species are so special,
Lex Fridman (2:22:36.560)
the only ones who worry about the meaning of life.
Lex Fridman (2:22:40.800)
And maybe that's what makes us human.
Lex Fridman (2:22:44.680)
And what I like to say to my wife and to my students
Manolis Kellis (2:22:47.960)
during this pandemic work extravaganza
Lex Fridman (2:22:53.280)
is every now and then they ask me, but how do you do this?
Lex Fridman (2:22:56.400)
And I'm like, I'm a workaholic.
Lex Fridman (2:22:58.880)
I love this.
Manolis Kellis (2:23:00.880)
This is me in the most unfiltered way.
Lex Fridman (2:23:04.800)
The ability to do something useful,
Manolis Kellis (2:23:07.280)
to feel that my brain is being used,
Lex Fridman (2:23:09.640)
to interact with the smartest people on the planet
Manolis Kellis (2:23:12.600)
day in, day out, and to help them discover aspects
Lex Fridman (2:23:15.880)
of the human genome, of the human brain,
Manolis Kellis (2:23:18.520)
of human disease and the human condition
Lex Fridman (2:23:21.960)
that no one has seen before
Manolis Kellis (2:23:24.560)
with data that we're capturing that has never been observed.
Lex Fridman (2:23:29.880)
And there's another aspect, which is on the personal life.
Lex Fridman (2:23:34.480)
Many people say, oh, I'm not gonna have kids, why bother?
Lex Fridman (2:23:37.560)
I can tell you as a father,
Manolis Kellis (2:23:41.280)
they're missing half the picture, if not the whole picture.
Lex Fridman (2:23:44.560)
Teaching my kids about my view of the world
Lex Fridman (2:23:49.040)
and watching through their eyes
Lex Fridman (2:23:51.200)
the naivete with which they start
Lex Fridman (2:23:53.480)
and the sophistication with which they end up,
Lex Fridman (2:23:56.920)
the understanding that they have
Manolis Kellis (2:24:00.080)
of not just the natural world around them, but of me too.
Lex Fridman (2:24:05.120)
The unfiltered criticism that you get from your own children
Manolis Kellis (2:24:10.120)
that knows no bounds of honesty.
Lex Fridman (2:24:15.200)
And I've grown components of my heart
Manolis Kellis (2:24:18.840)
that I didn't know I had
Lex Fridman (2:24:20.800)
until you sense that fragility,
Manolis Kellis (2:24:25.000)
that vulnerability of the children,
Lex Fridman (2:24:30.360)
that immense love and passion,
Manolis Kellis (2:24:34.040)
the unfiltered egoism,
Lex Fridman (2:24:36.560)
that we as adults learn how to hide so much better.
Manolis Kellis (2:24:40.080)
It's just this back of emotions
Lex Fridman (2:24:43.720)
that tell me about the raw materials that make a human being
Lex Fridman (2:24:48.720)
and how these raw materials can be arranged
Lex Fridman (2:24:50.880)
with more sophistication that we learn through life
Manolis Kellis (2:24:53.880)
to become truly human adults.
Lex Fridman (2:24:57.760)
But there's something so beautiful
Manolis Kellis (2:24:59.960)
about seeing that progression between them
Lex Fridman (2:25:02.480)
and seeing that progress and that progress
Lex Fridman (2:25:05.040)
and that progression between them,
Lex Fridman (2:25:07.280)
the complexity of the language growing
Manolis Kellis (2:25:10.600)
as more neural connections are formed
Lex Fridman (2:25:13.720)
to realize that the hardware is getting rearranged
Manolis Kellis (2:25:18.560)
as their software is getting implemented on that hardware,
Lex Fridman (2:25:22.640)
that their frontal cortex continues to grow
Manolis Kellis (2:25:24.880)
for another 10 years.
Lex Fridman (2:25:27.840)
There's neuronal connections that are continuing to form,
Manolis Kellis (2:25:29.960)
new neurons that actually get replicated and formed.
Lex Fridman (2:25:33.120)
And it's just incredible that we have these,
Manolis Kellis (2:25:38.120)
not just you grow the hardware for 30 years
Lex Fridman (2:25:40.680)
and then you feed it all of the knowledge.
Manolis Kellis (2:25:42.640)
No, no, the knowledge is fed throughout
Lex Fridman (2:25:45.200)
and is shaping these neural connections as they're forming.
Lex Fridman (2:25:48.480)
So seeing that transformation from either your own blood
Lex Fridman (2:25:52.840)
or from an adopted child
Manolis Kellis (2:25:54.560)
is the most beautiful thing you can do as a human being.
Lex Fridman (2:25:57.520)
And it completes you, it completes that path, that journey.
Manolis Kellis (2:26:00.760)
The create life, oh sure, that's at conception, that's easy.
Lex Fridman (2:26:04.880)
But create human life to add the human part,
Manolis Kellis (2:26:08.400)
that takes decades of compassion, of sharing,
Lex Fridman (2:26:13.160)
of love and of anger and of impatience and patience.
Lex Fridman (2:26:18.640)
And as a parent,
Lex Fridman (2:26:21.880)
I think I've become a very different kind of teacher
Manolis Kellis (2:26:25.960)
because again, I'm a professor.
Lex Fridman (2:26:27.080)
My first role is to bring adult human beings
Manolis Kellis (2:26:31.040)
into a more mature level of adulthood
Lex Fridman (2:26:34.440)
where they learn not just to do science,
Lex Fridman (2:26:37.040)
but they learn the process of discovery
Lex Fridman (2:26:39.840)
and the process of collaboration, the process of sharing,
Manolis Kellis (2:26:42.280)
the process of conveying the knowledge
Lex Fridman (2:26:44.840)
of encapsulating something incredibly complex
Lex Fridman (2:26:48.000)
and sort of giving it up in sort of bite sized chunks
Lex Fridman (2:26:51.200)
that the rest of humanity can appreciate.
Manolis Kellis (2:26:54.400)
I tell my students all the time, if you, you know,
Lex Fridman (2:26:57.440)
like when an apple fall,
Manolis Kellis (2:26:58.720)
when a tree falls in the forest
Lex Fridman (2:27:00.840)
and no one's there to listen, has it really fallen?
Manolis Kellis (2:27:03.040)
The same way you do this awesome research,
Lex Fridman (2:27:05.280)
if you write an impenetrable paper that no one will understand,
Manolis Kellis (2:27:08.640)
it's as if you never did the awesome research.
Lex Fridman (2:27:11.040)
So conveying of knowledge, conveying this lateral transfer
Manolis Kellis (2:27:15.200)
that I was talking about at the very beginning
Lex Fridman (2:27:17.520)
of sort of humanity and sort of the sharing of information,
Manolis Kellis (2:27:22.480)
all of that has gotten so much more rich
Lex Fridman (2:27:27.200)
by seeing human beings grow in my own home
Manolis Kellis (2:27:32.240)
because that makes me a better parent
Lex Fridman (2:27:35.080)
and that makes me a better teacher and a better mentor
Manolis Kellis (2:27:38.920)
to the nurturing of my adult children,
Lex Fridman (2:27:42.240)
which are my research group.
Manolis Kellis (2:27:43.960)
First of all, beautifully put, connects beautifully
Lex Fridman (2:27:48.280)
to the vertical and the horizontal inheritance of ideas
Manolis Kellis (2:27:52.240)
that we talked about at the very beginning.
Lex Fridman (2:27:54.440)
I don't think there's a better way to end it
Manolis Kellis (2:27:57.320)
on this poetic and powerful note.
Lex Fridman (2:28:01.320)
Manolis, thank you so much for talking to me.
Manolis Kellis (2:28:02.920)
It was a huge honor.
Lex Fridman (2:28:03.760)
We'll have to talk again about the origin of life,
Manolis Kellis (2:28:07.240)
about epigenetics, epigenomics,
Lex Fridman (2:28:10.520)
and some of the incredible research you're doing.
Manolis Kellis (2:28:13.600)
Truly an honor. Thanks so much for talking to me.
Lex Fridman (2:28:15.320)
Thank you. Such a pleasure. It's such a pleasure.
Manolis Kellis (2:28:17.240)
I mean, your questions are outstanding.
Lex Fridman (2:28:19.080)
I've had such a blast here and I can't wait to be back.
Manolis Kellis (2:28:21.880)
Awesome.
Lex Fridman (2:28:23.240)
Thanks for listening to this conversation
Manolis Kellis (2:28:24.800)
with Manolis Kellis, and thank you to our sponsors,
Lex Fridman (2:28:28.000)
Blinkist, 8sleep, and Masterclass.
Manolis Kellis (2:28:31.360)
Please consider supporting this podcast
Lex Fridman (2:28:33.200)
by going to blinkist.com slash lex,
Manolis Kellis (2:28:35.640)
8sleep.com slash lex, and masterclass.com slash lex.
Lex Fridman (2:28:41.040)
Click the links, buy the stuff, get the discount.
Manolis Kellis (2:28:44.160)
It's the best way to support this podcast.
Lex Fridman (2:28:47.040)
If you enjoy this thing, subscribe on YouTube,
Manolis Kellis (2:28:48.800)
review it with five stars on Apple Podcast,
Lex Fridman (2:28:50.920)
support it on Patreon,
Manolis Kellis (2:28:52.280)
or connect with me on Twitter at lexfreedman.
Lex Fridman (2:28:55.480)
And now let me leave you with some words
Manolis Kellis (2:28:57.360)
from Charles Darwin that I think Manolis
Lex Fridman (2:29:00.080)
represents quite beautifully.
Manolis Kellis (2:29:02.600)
If I had my life to live over again,
Lex Fridman (2:29:04.840)
I would have made a rule to read some poetry
Lex Fridman (2:29:07.560)
and listen to some music at least once every week.
Lex Fridman (2:29:11.640)
Thank you for listening, and hope to see you next time.
Lex Fridman (30:00.740)
And the particular abundance of receptors
Lex Fridman (30:03.140)
that you have this day based on what you ate yesterday
Manolis Kellis (30:06.100)
or that you have been wired with based on your parents
Lex Fridman (30:10.380)
and your upbringing, et cetera,
Manolis Kellis (30:12.580)
determines a lot of that quote unquote free will component
Lex Fridman (30:15.700)
to sort of narrow and narrow sort of slices.
Lex Fridman (30:20.700)
So how much on that point, how much freedom
Lex Fridman (30:24.140)
do you think we have to escape the constraints
Lex Fridman (30:29.020)
of our genome?
Lex Fridman (30:30.420)
You're making it sound like more and more
Manolis Kellis (30:31.980)
we're discovering that our genome is actually has the,
Lex Fridman (30:35.060)
a lot of the story already encoded into it.
Lex Fridman (30:37.740)
How much freedom do we have?
Lex Fridman (30:39.580)
I, so let me describe what that freedom would look like.
Manolis Kellis (30:45.140)
That freedom would be my saying,
Lex Fridman (30:47.620)
ooh, I'm gonna resist the urge to eat that apple
Manolis Kellis (30:51.540)
because I choose not to.
Lex Fridman (30:54.500)
But there are chemical receptors that made me
Manolis Kellis (30:57.620)
not resist the urge to prove my individuality
Lex Fridman (31:01.340)
and my free will by resisting the apple.
Lex Fridman (31:04.100)
So then the next question is,
Lex Fridman (31:05.580)
well, maybe now I'll resist the urge to resist the apple
Lex Fridman (31:08.220)
and I'll go for the chocolate instead
Lex Fridman (31:09.540)
to prove my individuality.
Lex Fridman (31:10.780)
But then what about those other receptors that, you know?
Lex Fridman (31:14.460)
That might be all encoded in there.
Lex Fridman (31:17.780)
So it's kicking the bucket down the road
Lex Fridman (31:19.460)
and basically saying, well, your choice
Manolis Kellis (31:22.020)
will may have actually been driven by other things
Lex Fridman (31:24.900)
that you actually are not choosing.
Lex Fridman (31:27.860)
So that's why it's very hard to answer that question.
Lex Fridman (31:30.020)
It's hard to know what to do with that.
Manolis Kellis (31:31.420)
I mean, if the genome has,
Lex Fridman (31:35.820)
if there's not much freedom, it's a...
Manolis Kellis (31:38.500)
It's the butterfly effect.
Lex Fridman (31:40.500)
It's basically that in the short term,
Manolis Kellis (31:42.900)
you can predict something extremely well
Lex Fridman (31:45.700)
by knowing the current state of the system.
Lex Fridman (31:48.100)
But a few steps down, it's very hard to predict
Lex Fridman (31:50.700)
based on the current knowledge.
Lex Fridman (31:52.380)
Is that because the system is truly free?
Lex Fridman (31:55.220)
When I look at weather patterns,
Manolis Kellis (31:56.340)
I can predict the next 10 days.
Lex Fridman (31:57.860)
Is it because the weather has a lot of freedom
Lex Fridman (32:00.260)
and after 10 days it chooses to do something else?
Lex Fridman (32:03.420)
Or is it because in fact the system is fully deterministic
Lex Fridman (32:07.300)
and there's just a slightly different magnetic field
Lex Fridman (32:10.100)
of the earth, slightly more energy arriving from the sun,
Manolis Kellis (32:12.420)
a slightly different spin of the gravitational pull
Lex Fridman (32:15.140)
of Jupiter that is now causing all kinds of tides
Lex Fridman (32:18.860)
and slight deviation of the moon, et cetera.
Lex Fridman (32:20.860)
Maybe all of that can be fully modeled.
Manolis Kellis (32:22.940)
Maybe the fact that China is emitting
Lex Fridman (32:25.740)
a little more carbon today is actually gonna affect
Manolis Kellis (32:28.180)
the weather in Egypt in three weeks.
Lex Fridman (32:31.460)
And all of that could be fully modeled.
Manolis Kellis (32:33.860)
In the same way, if you take a complete view
Lex Fridman (32:36.780)
of a human being now, I model everything about you.
Lex Fridman (32:42.700)
The question is, can I predict your next step?
Lex Fridman (32:44.860)
Probably, but at how far?
Lex Fridman (32:47.740)
And if it's a little further, is that because of stochasticity
Lex Fridman (32:51.260)
and sort of chaos properties of unpredictability
Lex Fridman (32:54.580)
of beyond a certain level?
Lex Fridman (32:56.100)
Or was that actually true free will?
Manolis Kellis (32:58.260)
Yeah, so the number of variables might be so,
Lex Fridman (33:01.260)
you might need to build an entire universe to be able to model.
Manolis Kellis (33:05.260)
To simulate a human, and then maybe that human
Lex Fridman (33:07.740)
will be fully simulatable.
Lex Fridman (33:09.420)
But maybe aspects of free will will exist.
Lex Fridman (33:12.220)
And where's that free will coming from?
Manolis Kellis (33:13.380)
It's still coming from the same neurons
Lex Fridman (33:14.980)
or maybe from a spirit inhabiting these neurons.
Lex Fridman (33:17.580)
But again, it's very difficult empirically
Lex Fridman (33:19.740)
to sort of evaluate where does free will begin
Lex Fridman (33:22.540)
and sort of chemical reactions and electric signals.
Lex Fridman (33:26.700)
So on that topic, let me ask the most absurd question
Manolis Kellis (33:31.140)
that most MIT faculty rolled their eyes on.
Lex Fridman (33:33.900)
But what do you think about the simulation hypothesis
Lex Fridman (33:38.260)
and the idea that we live in a simulation?
Lex Fridman (33:40.220)
I think it's complete BS.
Manolis Kellis (33:41.580)
Okay.
Lex Fridman (33:44.540)
There's no empirical evidence.
Manolis Kellis (33:45.740)
No, it's not. Absolutely not.
Lex Fridman (33:47.060)
Not in terms of empirical evidence or not,
Lex Fridman (33:49.020)
but in terms of a thought experiment,
Lex Fridman (33:52.380)
does it help you think about the universe?
Manolis Kellis (33:54.860)
I mean, so if you look at the genome,
Lex Fridman (33:57.500)
it's encoding a lot of the information
Manolis Kellis (33:59.180)
that is required to create some of the beautiful
Lex Fridman (34:01.500)
human complexity that we see around us.
Manolis Kellis (34:04.220)
It's an interesting thought experiment.
Lex Fridman (34:05.940)
How much parameters do we need to have
Lex Fridman (34:11.340)
in order to model this full human experience?
Lex Fridman (34:15.300)
Like if we were to build a video game,
Lex Fridman (34:17.540)
how hard it would be to build a video game
Lex Fridman (34:19.980)
that's like convincing enough and fun enough
Lex Fridman (34:22.660)
and it has consistent laws of physics, all that stuff.
Lex Fridman (34:28.340)
It's not interesting to use a thought experiment.
Manolis Kellis (34:31.380)
I mean, it's cute, but it's Occam's razor.
Lex Fridman (34:35.060)
I mean, what's more realistic,
Manolis Kellis (34:36.820)
the fact that you're actually a machine
Lex Fridman (34:38.340)
or that you're a person?
Manolis Kellis (34:39.740)
What's the fact that all of my experiences exist
Lex Fridman (34:43.340)
inside the chemical molecules that I have
Lex Fridman (34:45.540)
or that somebody is actually simulating all that?
Lex Fridman (34:49.540)
Well, you did refer to humans
Manolis Kellis (34:50.860)
as a digital computer earlier.
Lex Fridman (34:52.540)
Of course, of course.
Lex Fridman (34:53.420)
But that does not.
Lex Fridman (34:54.260)
It's a kind of a machine, right?
Manolis Kellis (34:55.300)
I know, I know.
Lex Fridman (34:56.380)
But I think the probability of all that is nil
Lex Fridman (35:01.740)
and let the machines wake me up
Lex Fridman (35:03.500)
and just terminate me now if it's not.
Manolis Kellis (35:07.540)
I challenge you machines.
Lex Fridman (35:08.860)
They're gonna wait a little bit
Manolis Kellis (35:10.940)
to see what you're gonna do next.
Lex Fridman (35:12.380)
It's fun.
Manolis Kellis (35:13.220)
It's fun to watch, especially the clever humans.
Lex Fridman (35:17.380)
What's the difference to you
Manolis Kellis (35:18.540)
between the way a computer stores information
Lex Fridman (35:21.300)
and the human genome stores information?
Lex Fridman (35:24.020)
So you also have roots and your work.
Lex Fridman (35:27.020)
Would you say when you introduce yourself at a bar.
Manolis Kellis (35:31.980)
It depends who I'm talking to.
Lex Fridman (35:34.020)
Would you say it's computational biology?
Lex Fridman (35:36.180)
Do you reveal your expertise in computers?
Lex Fridman (35:43.300)
It depends who I'm talking to, truly.
Manolis Kellis (35:45.340)
I mean, basically, if I meet someone who's in computers,
Lex Fridman (35:47.700)
I'll say, oh, I'm a professor in computer science.
Manolis Kellis (35:51.100)
If I meet someone who's in engineering,
Lex Fridman (35:52.500)
I say computer science and electrical engineering.
Manolis Kellis (35:54.780)
If I meet someone in biology,
Lex Fridman (35:55.980)
I'll say, hey, I work in genomics.
Manolis Kellis (35:57.220)
If I meet someone in medicine,
Lex Fridman (35:58.300)
I'm like, hey, I work on genetics.
Lex Fridman (36:00.740)
So you're a fun person to meet at a bar.
Lex Fridman (36:02.220)
I got you, but so.
Manolis Kellis (36:03.940)
No, no, but what I'm trying to say is that I don't,
Lex Fridman (36:07.460)
I mean, there's no single attribute
Manolis Kellis (36:09.100)
that I will define myself as.
Lex Fridman (36:11.140)
There's a few things I know.
Manolis Kellis (36:12.140)
There's a few things I study.
Lex Fridman (36:13.140)
There's a few things I have degrees on
Lex Fridman (36:15.100)
and there's a few things that I grant degrees in.
Lex Fridman (36:17.980)
And I publish papers across the whole gamut,
Manolis Kellis (36:22.900)
the whole spectrum of computation to biology, et cetera.
Lex Fridman (36:26.380)
I mean, the complete answer is that I use computer science
Manolis Kellis (36:31.580)
to understand biology.
Lex Fridman (36:34.180)
So I develop methods in AI and machine learning,
Manolis Kellis (36:39.460)
statistics and algorithms, et cetera.
Lex Fridman (36:41.700)
But the ultimate goal of my career
Manolis Kellis (36:44.060)
is to really understand biology.
Lex Fridman (36:45.700)
If these things don't advance our understanding
Manolis Kellis (36:47.780)
of biology, I'm not as fascinated by them.
Lex Fridman (36:51.980)
Although there are some beautiful computational problems
Manolis Kellis (36:54.940)
by themselves, I've sort of made it my mission
Lex Fridman (36:57.940)
to apply the power of computer science
Manolis Kellis (37:01.660)
to truly understand the human genome, health, disease,
Lex Fridman (37:07.500)
and the whole gamut of how our brain works,
Lex Fridman (37:10.100)
how our body works and all of that,
Lex Fridman (37:11.740)
which is so fascinating.
Lex Fridman (37:13.980)
And so the dream, there's not an equivalent
Lex Fridman (37:16.940)
sort of complimentary dream of understanding
Manolis Kellis (37:20.940)
human biology in order to create an artificial life
Lex Fridman (37:23.340)
or an artificial brain or artificial intelligence
Manolis Kellis (37:26.060)
that supersedes the intelligence
Lex Fridman (37:27.660)
and the capabilities of us humans.
Manolis Kellis (37:30.740)
It's an interesting question.
Lex Fridman (37:31.860)
It's a fascinating question.
Lex Fridman (37:33.260)
So understanding the human brain is undoubtedly coupled
Lex Fridman (37:39.740)
to how do we make better AI?
Manolis Kellis (37:42.180)
Because so much of AI has in fact been inspired
Lex Fridman (37:46.420)
by the brain.
Manolis Kellis (37:47.260)
It may have taken 50 years
Lex Fridman (37:49.060)
since the early days of neural networks
Manolis Kellis (37:51.140)
till we have all of these amazing progress
Lex Fridman (37:55.420)
that we've seen with deep belief networks
Lex Fridman (38:00.820)
and all of these advances in Go, in Chess,
Lex Fridman (38:06.460)
in image synthesis, in deep fakes, in you name it.
Lex Fridman (38:10.420)
But the underlying architecture is very much inspired
Lex Fridman (38:17.020)
by the human brain,
Manolis Kellis (38:18.060)
which actually posits a very, very interesting question.
Lex Fridman (38:22.580)
Why are neural networks performing so well?
Lex Fridman (38:27.220)
And they perform amazingly well.
Lex Fridman (38:28.980)
Is it because they can simulate any possible function?
Lex Fridman (38:32.580)
And the answer is no, no.
Lex Fridman (38:34.420)
They simulate a very small number of functions.
Manolis Kellis (38:37.180)
Is it because they can simulate every function
Lex Fridman (38:39.420)
in the universe?
Lex Fridman (38:40.500)
And that's where it gets interesting.
Lex Fridman (38:41.540)
The answer is actually, yeah, a little closer to that.
Lex Fridman (38:44.740)
And here's where it gets really fun.
Lex Fridman (38:47.700)
If you look at human brain and human cognition,
Manolis Kellis (38:51.700)
it didn't evolve in a vacuum.
Lex Fridman (38:53.980)
It evolved in a world with physical constraints,
Manolis Kellis (38:58.780)
like the world that inhabits us.
Lex Fridman (39:00.700)
It is the world that we inhabit.
Lex Fridman (39:03.220)
And if you look at our senses, what do they perceive?
Lex Fridman (39:08.220)
They perceive different parts of the electromagnetic spectrum.
Manolis Kellis (39:13.220)
The hearing is just different movements in air,
Lex Fridman (39:17.260)
the touch, et cetera.
Manolis Kellis (39:18.820)
I mean, all of these things,
Lex Fridman (39:20.100)
we've built intuitions for the physical world
Manolis Kellis (39:22.700)
that we inhabit.
Lex Fridman (39:23.980)
And our brains and the brains of all animals evolved
Manolis Kellis (39:27.660)
for that world.
Lex Fridman (39:29.140)
And the AI systems that we have built
Manolis Kellis (39:32.660)
happen to work well with images
Lex Fridman (39:34.700)
of the type that we encounter
Manolis Kellis (39:36.100)
in the physical world that we inhabit.
Lex Fridman (39:38.380)
Whereas if you just take noise and you add random signal
Manolis Kellis (39:42.420)
that doesn't match anything in our world,
Lex Fridman (39:44.420)
neural networks will not do as well.
Lex Fridman (39:46.940)
And that actually basically has this whole loop around this,
Lex Fridman (39:52.900)
which is this was designed by studying our own brain,
Manolis Kellis (39:57.620)
which was evolved for our own world.
Lex Fridman (39:59.580)
And they happen to do well in our own world.
Lex Fridman (40:01.940)
And they happen to make the same types of mistakes
Lex Fridman (40:04.140)
that humans make many times.
Lex Fridman (40:07.020)
And of course you can engineer images
Lex Fridman (40:08.740)
by adding just the right amount of sort of pixel deviations
Manolis Kellis (40:12.580)
to make a zebra look like a bamboo and stuff like that,
Lex Fridman (40:15.780)
or like a table.
Lex Fridman (40:18.380)
But ultimately the undoctored images at least
Lex Fridman (40:23.580)
are very often mistaken, I don't know,
Manolis Kellis (40:25.940)
between muffins and dogs, for example,
Lex Fridman (40:28.860)
in the same way that humans make those mistakes.
Lex Fridman (40:31.220)
So there's no doubt in my view
Lex Fridman (40:35.820)
that the more we understand about the tricks
Manolis Kellis (40:38.580)
that our human brain has evolved
Lex Fridman (40:40.580)
to understand the physical world around us,
Manolis Kellis (40:42.900)
the more we will be able to bring
Lex Fridman (40:44.740)
new computational primitives in our AI systems
Manolis Kellis (40:48.780)
to again better understand not just the world around us,
Lex Fridman (40:52.220)
but maybe even the world inside us,
Lex Fridman (40:54.460)
and maybe even the computational problems that arise
Lex Fridman (40:57.180)
from new types of data that we haven't been exposed to,
Lex Fridman (41:00.380)
but are yet inhabiting the same universe that we live in
Lex Fridman (41:03.460)
with a very tiny little subset of functions
Manolis Kellis (41:06.100)
from all possible mathematical functions.
Lex Fridman (41:08.140)
Yeah, and that small subset of functions,
Manolis Kellis (41:10.220)
all that matters to us humans really, that's what makes.
Lex Fridman (41:12.940)
It's all that has mattered so far.
Lex Fridman (41:14.860)
And even within our scientific realm,
Lex Fridman (41:17.100)
it's all that seems to continue to matter.
Lex Fridman (41:19.740)
But I mean, I always like to think about our senses
Lex Fridman (41:24.860)
and how much of the physical world around us we perceive.
Lex Fridman (41:29.380)
And if you look at the LIGO experiment
Lex Fridman (41:35.020)
over the last year and a half has been all over the news.
Lex Fridman (41:38.220)
What did LIGO do?
Lex Fridman (41:39.660)
It created a new sense for human beings,
Manolis Kellis (41:42.940)
a sense that has never been sensed
Lex Fridman (41:45.820)
in the history of our planet.
Manolis Kellis (41:48.980)
Gravitational waves have been traversing the earth
Lex Fridman (41:53.020)
since its creation a few billion years ago.
Manolis Kellis (41:55.220)
Life has evolved senses to sense things
Lex Fridman (41:59.580)
that were never before sensed.
Manolis Kellis (42:02.220)
Light was not perceived by early life.
Lex Fridman (42:05.780)
No one cared.
Lex Fridman (42:07.580)
And eventually photoreceptors evolved
Lex Fridman (42:11.260)
and the ability to sense colors
Manolis Kellis (42:14.220)
by sort of catching different parts
Lex Fridman (42:16.060)
of that electromagnetic spectrum.
Lex Fridman (42:19.140)
And hearing evolved and touch evolved, et cetera.
Lex Fridman (42:23.380)
But no organism evolved a way to sense neutrinos
Manolis Kellis (42:27.700)
floating through earth or gravitational waves
Lex Fridman (42:29.940)
flowing through earth, et cetera.
Lex Fridman (42:31.260)
And I find it so beautiful in the history
Lex Fridman (42:33.860)
of not just humanity, but life on the planet
Manolis Kellis (42:37.020)
that we are now able to capture additional signals
Lex Fridman (42:40.460)
from the physical world than we ever knew before.
Lex Fridman (42:43.620)
And axions, for example, have been all over the news
Lex Fridman (42:46.340)
in the last few weeks.
Lex Fridman (42:47.460)
And the concept that we can capture and perceive
Lex Fridman (42:53.580)
more of that physical world is as exciting
Manolis Kellis (42:57.660)
as the fact that we were blind to it
Lex Fridman (43:01.980)
is traumatizing before.
Manolis Kellis (43:04.620)
Because that also tells us, you know, we're in 2020.
Lex Fridman (43:09.380)
Picture yourself in 3020 or in 20, you know.
Lex Fridman (43:12.820)
What new senses might we discover?
Lex Fridman (43:15.580)
Is it, you know, could it be that we're missing
Lex Fridman (43:19.580)
nine tenths of physics?
Lex Fridman (43:21.900)
That like, there's a lot of physics out there
Manolis Kellis (43:23.980)
that we're just blind to, completely oblivious to it.
Lex Fridman (43:27.900)
And yet they're permeating us all the time.
Manolis Kellis (43:29.340)
Yeah, so it might be right in front of us.
Lex Fridman (43:31.140)
So when you're thinking about premonitions,
Manolis Kellis (43:35.060)
yeah, a lot of that is ascertainment bias.
Lex Fridman (43:37.580)
Like, yeah, you know, every now and then you're like,
Manolis Kellis (43:39.420)
oh, I remember my friend.
Lex Fridman (43:41.020)
And then my friend doesn't appear
Lex Fridman (43:42.660)
and I'll forget that I remembered my friend.
Lex Fridman (43:44.540)
But every now and then my friend will actually appear.
Manolis Kellis (43:45.980)
I'm like, oh my God, I thought about you a minute ago.
Lex Fridman (43:48.340)
You just called me, that's amazing.
Manolis Kellis (43:50.140)
So, you know, some of that is this,
Lex Fridman (43:51.980)
but some of that might be that there are,
Manolis Kellis (43:55.060)
within our brain, sensors for waves
Lex Fridman (43:59.900)
that we emit that we're not even aware of.
Lex Fridman (44:03.220)
And this whole concept of when I hug my children,
Lex Fridman (44:07.020)
there's such an emotional transfer there
Manolis Kellis (44:10.460)
that we don't comprehend.
Lex Fridman (44:12.220)
I mean, sure, yeah, of course we're all like hard wire
Manolis Kellis (44:15.100)
for all kinds of touchy feely things
Lex Fridman (44:16.700)
between parents and kids, it's beautiful,
Manolis Kellis (44:18.220)
between partners, it's beautiful, et cetera.
Lex Fridman (44:20.660)
But then there are intangible aspects
Manolis Kellis (44:24.900)
of human communication
Lex Fridman (44:27.340)
that I don't think it's unfathomable
Manolis Kellis (44:30.020)
that our brain has actually evolved waves and sensors
Lex Fridman (44:32.060)
for it that we just don't capture.
Manolis Kellis (44:33.940)
We don't understand the function
Lex Fridman (44:35.220)
of the vast majority of our neurons.
Lex Fridman (44:37.420)
And maybe our brain is already sensing it,
Lex Fridman (44:40.100)
but even worse, maybe our brain is not sensing it at all.
Lex Fridman (44:43.940)
And we're oblivious to this until we build a machine
Lex Fridman (44:46.620)
that suddenly is able to sort of capture
Lex Fridman (44:48.300)
so much more of what's happening in the natural world.
Lex Fridman (44:50.340)
So what you're saying is physics
Manolis Kellis (44:52.620)
is going to discover a sensor for love.
Lex Fridman (44:57.220)
And maybe dogs are off scale for that.
Lex Fridman (45:01.460)
And we've been oblivious to it the whole time
Lex Fridman (45:04.140)
because we didn't have the right sensor.
Lex Fridman (45:05.740)
And now you're gonna have a little wrist that says,
Lex Fridman (45:07.420)
oh my God, I feel all this love in the house.
Manolis Kellis (45:09.660)
I sense a disturbance in the forest.
Lex Fridman (45:11.860)
It's all around us.
Lex Fridman (45:13.660)
And dogs and cats will have zero.
Lex Fridman (45:15.780)
None. None.
Manolis Kellis (45:17.020)
It's just.
Lex Fridman (45:17.860)
Oh, no signal.
Lex Fridman (45:20.100)
But let's take a step back to our unfortunate place.
Lex Fridman (45:24.540)
To one of the 400 topics that we had actually planned for.
Lex Fridman (45:29.580)
But to our sad time in 2020
Lex Fridman (45:31.820)
when we only have just a few sensors
Lex Fridman (45:33.860)
and very primitive early computers.
Lex Fridman (45:37.620)
So you have a foot in computer science
Lex Fridman (45:41.820)
and a foot in biology.
Lex Fridman (45:43.500)
In your sense, how do computers represent information
Lex Fridman (45:48.300)
differently than like the genome or biological systems?
Lex Fridman (45:52.300)
So first of all, let me correct
Manolis Kellis (45:55.540)
that no, we're in an amazing time in 2020.
Lex Fridman (46:00.340)
Computer science is totally awesome.
Lex Fridman (46:02.460)
And physics is totally awesome.
Lex Fridman (46:03.980)
And we have understood so much of the natural world
Manolis Kellis (46:06.900)
than ever before.
Lex Fridman (46:08.500)
So I am extremely grateful and feeling extremely lucky
Manolis Kellis (46:13.140)
to be living in the time that we are.
Lex Fridman (46:16.180)
Cause you know, first of all,
Manolis Kellis (46:17.540)
who knows when the asteroid will hit.
Lex Fridman (46:20.140)
And second, you know, of all times in humanity,
Manolis Kellis (46:26.140)
this is probably the best time to be a human being.
Lex Fridman (46:29.380)
And this might actually be the best place
Manolis Kellis (46:31.100)
to be a human being.
Lex Fridman (46:31.940)
So anyway, you know, for anyone who loves science,
Manolis Kellis (46:34.420)
this is it.
Lex Fridman (46:35.260)
This is awesome.
Manolis Kellis (46:36.100)
This is a great time.
Lex Fridman (46:36.940)
At the same time, just a quick comment.
Manolis Kellis (46:39.300)
All I meant is that if we look several hundred years
Lex Fridman (46:43.060)
from now and we end up somehow not destroying ourselves,
Manolis Kellis (46:48.500)
people will probably look back at this time
Lex Fridman (46:50.300)
in computer science and at your work of Manos at MIT.
Manolis Kellis (46:55.620)
As infantile.
Lex Fridman (46:56.580)
As infantile and silly and how ignorant it all was.
Manolis Kellis (46:59.620)
I like to joke very often with my students
Lex Fridman (47:02.500)
that, you know, we've written so many papers.
Manolis Kellis (47:04.220)
We've published so much.
Lex Fridman (47:05.260)
We've been citing so much.
Lex Fridman (47:06.460)
And every single time I tell my students, you know,
Lex Fridman (47:08.380)
the best is ahead of us.
Lex Fridman (47:09.700)
What we're working on now
Lex Fridman (47:11.380)
is the most exciting thing I've ever worked on.
Lex Fridman (47:13.860)
So in a way, I do have this sense of, yeah,
Lex Fridman (47:16.420)
even the papers I wrote 10 years ago,
Manolis Kellis (47:18.540)
they were awesome at the time,
Lex Fridman (47:20.300)
but I'm so much more excited about where we're heading now.
Lex Fridman (47:22.380)
And I don't mean to minimize any of the stuff
Lex Fridman (47:24.500)
we've done in the past,
Lex Fridman (47:25.460)
but you know, there's just this sense of excitement
Lex Fridman (47:29.020)
about what you're working on now
Manolis Kellis (47:30.980)
that as soon as a paper is submitted,
Lex Fridman (47:33.380)
it's like, ugh, it's old.
Manolis Kellis (47:35.540)
You know, I can't talk about that anymore.
Lex Fridman (47:37.140)
I'm not gonna talk about it.
Manolis Kellis (47:37.980)
At the same time, you're not,
Lex Fridman (47:38.820)
you probably are not going to be able to predict
Lex Fridman (47:41.340)
what are the most impactful papers and ideas
Lex Fridman (47:45.500)
when people look back 200 years from now at your work,
Lex Fridman (47:47.860)
what would be the most exciting papers.
Lex Fridman (47:50.740)
And it may very well be not the thing that you expected.
Manolis Kellis (47:54.220)
Or the things you got awards for or, you know.
Lex Fridman (47:58.100)
This might be true in some fields.
Manolis Kellis (47:59.980)
I don't know.
Lex Fridman (48:00.820)
I feel slightly differently about it in our field.
Manolis Kellis (48:02.340)
I feel that I kind of know what are the important ones.
Lex Fridman (48:05.660)
And there's a very big difference
Manolis Kellis (48:07.300)
between what the press picks up on
Lex Fridman (48:09.180)
and what's actually fundamentally important for the field.
Lex Fridman (48:11.620)
And I think for the fundamentally important ones,
Lex Fridman (48:13.380)
we kind of have a pretty good idea what they are.
Lex Fridman (48:15.580)
And it's hard to sometimes get the press excited
Lex Fridman (48:18.140)
about the fundamental advances,
Lex Fridman (48:20.100)
but you know, we take what we get
Lex Fridman (48:23.700)
and celebrate what we get.
Lex Fridman (48:24.740)
And sometimes, you know, one of our papers,
Lex Fridman (48:27.180)
which was in a minor journal,
Manolis Kellis (48:28.500)
made the front page of Reddit
Lex Fridman (48:30.180)
and suddenly had like hundreds of thousands of views.
Manolis Kellis (48:33.500)
Even though it was in a minor journal
Lex Fridman (48:34.980)
because, you know, somebody pitched it the right way
Manolis Kellis (48:37.020)
that it suddenly caught everybody's attention.
Lex Fridman (48:39.380)
Whereas other papers that are sort of truly fundamental,
Manolis Kellis (48:42.020)
you know, we have a hard time
Lex Fridman (48:43.660)
getting the editors even excited about them
Manolis Kellis (48:46.060)
when so many hundreds of people
Lex Fridman (48:47.860)
are already using the results and building upon them.
Lex Fridman (48:50.860)
So I do appreciate that there's a discrepancy
Lex Fridman (48:54.420)
between the perception and the perceived success
Lex Fridman (48:57.420)
and the awards that you get for various papers.
Lex Fridman (48:59.500)
But I think that fundamentally, I know that, you know,
Manolis Kellis (49:02.500)
some paper, I'm so, so when you write.
Lex Fridman (49:04.380)
So is there a paper that you're most proud of?
Manolis Kellis (49:06.820)
See, now you just, you trapped yourself.
Lex Fridman (49:09.340)
No, no, no, no, I mean.
Manolis Kellis (49:10.340)
Is there a line of work that you have a sense
Lex Fridman (49:14.620)
is really powerful that you've done to date?
Manolis Kellis (49:17.580)
You've done so much work in so many directions,
Lex Fridman (49:20.180)
which is interesting.
Lex Fridman (49:21.860)
Is there something where you think is quite special?
Lex Fridman (49:25.340)
I mean, it's like asking me to say
Manolis Kellis (49:28.740)
which of my three children I love best.
Lex Fridman (49:30.380)
I mean.
Manolis Kellis (49:34.060)
Exactly.
Lex Fridman (49:34.900)
So, I mean, and it's such a gimme question
Manolis Kellis (49:38.500)
that is so, so difficult not to brag
Lex Fridman (49:42.580)
about the awesome work that my team
Lex Fridman (49:44.740)
and my students have done.
Lex Fridman (49:47.060)
And I'll just mention a few off the top of my head.
Manolis Kellis (49:49.940)
I mean, basically there's a few landmark papers
Lex Fridman (49:53.060)
that I think have shaped my scientific path.
Manolis Kellis (49:56.780)
And, you know, I like to somehow describe it
Lex Fridman (50:00.380)
as a linear continuation of one thing led to another
Lex Fridman (50:03.620)
and led to another led to another.
Lex Fridman (50:05.340)
And, you know, it kind of all started with,
Manolis Kellis (50:11.020)
skip, skip, skip, skip, skip.
Lex Fridman (50:12.340)
Let me try to start somewhere in the middle.
Lex Fridman (50:14.020)
So my first PhD paper was the first comparative analysis
Lex Fridman (50:20.460)
of multiple species.
Lex Fridman (50:21.820)
So multiple complete genomes.
Lex Fridman (50:23.580)
So for the first time we basically developed the concept
Manolis Kellis (50:27.340)
of genome wide evolutionary signatures.
Lex Fridman (50:29.980)
The fact that you could look across the entire genome
Lex Fridman (50:32.940)
and understand how things evolve.
Lex Fridman (50:35.660)
And from these signatures of evolution
Manolis Kellis (50:38.260)
you could go back and study any one region
Lex Fridman (50:41.540)
and say, that's a protein coding gene.
Manolis Kellis (50:44.020)
That's an RNA gene.
Lex Fridman (50:45.540)
That's a regulatory motif.
Manolis Kellis (50:47.260)
That's a, you know, binding site and so on and so forth.
Lex Fridman (50:50.100)
So.
Manolis Kellis (50:50.940)
I'm sorry, so comparing different.
Lex Fridman (50:52.660)
Different species.
Manolis Kellis (50:53.780)
Species of the same.
Lex Fridman (50:55.060)
So take human, mouse, rat and dog.
Manolis Kellis (50:57.140)
Yeah.
Lex Fridman (50:58.060)
You know, they're all animals, they're all mammals.
Manolis Kellis (50:59.980)
They're all performing similar functions with their heart,
Lex Fridman (51:02.820)
with their brain, with their lungs, et cetera, et cetera.
Lex Fridman (51:05.860)
So there's many functional elements
Lex Fridman (51:08.140)
that make us uniquely mammalian.
Lex Fridman (51:10.900)
And those mammalian elements are actually conserved.
Lex Fridman (51:14.620)
99% of our genome does not code for protein.
Manolis Kellis (51:18.940)
1% codes for protein.
Lex Fridman (51:20.780)
The other 99%, we frankly didn't know what it does
Manolis Kellis (51:25.100)
until we started doing this comparative genomic studies.
Lex Fridman (51:28.140)
So basically these series of papers in my career
Manolis Kellis (51:32.060)
have basically first developed that concept
Lex Fridman (51:34.540)
of evolutionary signatures and then apply them to yeast,
Manolis Kellis (51:37.460)
apply them to flies, apply them to four mammals,
Lex Fridman (51:40.140)
apply them to 17 fungi,
Manolis Kellis (51:41.620)
apply them to 12 Drosophila species,
Lex Fridman (51:43.700)
apply them to then 29 mammals and now 200 mammals.
Lex Fridman (51:46.900)
So sorry, so can we.
Lex Fridman (51:48.820)
So the evolutionary signatures seems like
Manolis Kellis (51:51.380)
it's such a fascinating idea.
Lex Fridman (51:53.580)
And we're probably gonna linger on your early PhD work
Manolis Kellis (51:57.380)
for two hours.
Lex Fridman (51:58.220)
But what is, how can you reveal something interesting
Manolis Kellis (52:04.260)
about the genome by looking at the multiple,
Lex Fridman (52:08.500)
multiple species and looking at the evolutionary signatures?
Manolis Kellis (52:11.900)
Yeah, so you basically align
Lex Fridman (52:16.900)
the matching regions.
Lex Fridman (52:20.820)
So everything evolved from a common ancestor way, way back.
Lex Fridman (52:23.980)
And mammals evolved from a common ancestor
Manolis Kellis (52:26.020)
about 60 million years back.
Lex Fridman (52:27.860)
So after the meteor that killed off the dinosaurs landed
Manolis Kellis (52:35.460)
near Machu Picchu, we know the crater.
Lex Fridman (52:38.780)
It didn't allegedly land.
Manolis Kellis (52:41.700)
That was the aliens, okay.
Lex Fridman (52:42.860)
No, just slightly north of Machu Picchu
Manolis Kellis (52:44.660)
in the Gulf of Mexico, there's a giant hole
Lex Fridman (52:47.100)
that that meteor impact.
Lex Fridman (52:49.060)
Sorry, is that definitive to people?
Lex Fridman (52:51.380)
Have people conclusively figured out
Lex Fridman (52:56.380)
what killed the dinosaurs?
Lex Fridman (52:58.180)
I think so.
Lex Fridman (52:59.220)
So it was a meteor?
Lex Fridman (53:00.540)
Well, volcanic activity, all kinds of other stuff
Manolis Kellis (53:04.860)
is coinciding, but the meteor is pretty unique
Lex Fridman (53:09.580)
and we now have. That's also terrifying.
Manolis Kellis (53:11.180)
I wouldn't, we still have a lot of 2020 left,
Lex Fridman (53:14.940)
so if anything.
Manolis Kellis (53:15.780)
No, no, but think about it this way.
Lex Fridman (53:17.220)
So the dinosaurs ruled the earth for 175 million years.
Lex Fridman (53:24.420)
We humans have been around for what?
Lex Fridman (53:28.380)
Less than 1 million years.
Manolis Kellis (53:29.940)
If you're super generous about what you call humans
Lex Fridman (53:32.900)
and you include chimps basically.
Lex Fridman (53:35.340)
So we are just getting warmed up
Lex Fridman (53:38.580)
and we've ruled the planet much more ruthlessly
Manolis Kellis (53:42.500)
than Tyrannosaurus Rex.
Lex Fridman (53:46.220)
T Rex had much less of an environmental impact
Manolis Kellis (53:48.340)
than we did.
Lex Fridman (53:49.580)
And if you give us another 174 million years,
Manolis Kellis (53:54.020)
humans will look very different if we make it that far.
Lex Fridman (53:58.380)
So I think dinosaurs basically are much more
Manolis Kellis (54:02.180)
of life history on earth than we are in all respects.
Lex Fridman (54:06.100)
But look at the bright side, when they were killed off,
Manolis Kellis (54:08.740)
another life form emerged, mammals.
Lex Fridman (54:10.860)
And that's that whole evolutionary branching
Manolis Kellis (54:14.620)
that's happened.
Lex Fridman (54:15.460)
So you kind of have,
Manolis Kellis (54:17.060)
when you have these evolutionary signatures,
Lex Fridman (54:19.180)
is there basically a map of how the genome changed?
Manolis Kellis (54:22.660)
Yeah, exactly, exactly.
Lex Fridman (54:23.540)
So now you can go back to this early mammal
Manolis Kellis (54:26.180)
that was hiding in caves and you can basically ask
Lex Fridman (54:29.260)
what happened after the dinosaurs were wiped out.
Manolis Kellis (54:31.300)
A ton of evolutionary niches opened up
Lex Fridman (54:34.060)
and the mammals started populating all of these niches.
Lex Fridman (54:37.460)
And in that diversification,
Lex Fridman (54:40.660)
there was room for expansion of new types of functions.
Lex Fridman (54:44.740)
So some of them populated the air with bats flying,
Lex Fridman (54:50.140)
a new evolution of flight.
Manolis Kellis (54:53.260)
Some populated the oceans with dolphins and whales
Lex Fridman (54:57.460)
going off to swim, et cetera.
Lex Fridman (54:58.700)
But we all are fundamentally mammals.
Lex Fridman (55:01.380)
So you can take the genomes of all these species
Lex Fridman (55:04.220)
and align them on top of each other
Lex Fridman (55:06.260)
and basically create nucleotide resolution correspondences.
Lex Fridman (55:11.860)
What my PhD work showed is that when you do that,
Lex Fridman (55:14.220)
when you line up species on top of each other,
Manolis Kellis (55:17.180)
you can see that within protein coding genes,
Lex Fridman (55:19.860)
there's a particular pattern of evolution
Manolis Kellis (55:21.980)
that is dictated by the level at which
Lex Fridman (55:25.780)
evolutionary selection acts.
Manolis Kellis (55:27.700)
If I'm coding for a protein and I change
Lex Fridman (55:30.700)
the third codon position of a triplet
Manolis Kellis (55:34.340)
that codes for that amino acid,
Lex Fridman (55:36.580)
the same amino acid will be encoded.
Lex Fridman (55:38.980)
So that basically means that any kind of mutation
Lex Fridman (55:42.020)
that preserves that translation that is invariant
Manolis Kellis (55:46.140)
to that ultimate functional assessment
Lex Fridman (55:49.580)
that evolution will give is tolerated.
Lex Fridman (55:52.420)
So for any function that you're trying to achieve,
Lex Fridman (55:55.100)
there's a set of sequences that encode it.
Manolis Kellis (55:57.820)
You can now look at the mapping,
Lex Fridman (56:00.380)
the graph isomorphism, if you wish,
Manolis Kellis (56:04.460)
between all of the possible DNA encodings
Lex Fridman (56:07.460)
of a particular function and that function.
Lex Fridman (56:09.780)
And instead of having just that exact sequence
Lex Fridman (56:12.420)
at the protein level, you can think of the set
Manolis Kellis (56:15.020)
of protein sequences that all fulfill the same function.
Lex Fridman (56:18.020)
What's evolution doing?
Manolis Kellis (56:19.420)
Evolution has two components.
Lex Fridman (56:20.820)
One component is random, blind, and stupid mutation.
Manolis Kellis (56:25.300)
The other component is super smart, ruthless selection.
Lex Fridman (56:32.020)
That's my mom calling from Greece.
Manolis Kellis (56:35.220)
Yes, I might be a fully grown man, but I am a Greek.
Lex Fridman (56:40.060)
Did you just cancel the call?
Manolis Kellis (56:41.540)
Wow, you're in trouble.
Lex Fridman (56:42.380)
I know, I'm in trouble.
Manolis Kellis (56:43.220)
No, she's gonna be calling the cops.
Lex Fridman (56:44.860)
Honey, are you okay?
Manolis Kellis (56:45.700)
I'm gonna edit this clip out and send it to her.
Lex Fridman (56:47.700)
Sure.
Lex Fridman (56:51.660)
So there's a lot of encoding
Lex Fridman (56:53.060)
for the same kind of function.
Manolis Kellis (56:54.300)
Yeah, so you now have this mapping
Lex Fridman (56:56.620)
between all of the set of functions
Manolis Kellis (56:58.740)
that could all encode the same,
Lex Fridman (57:00.900)
all of the set of sequences
Manolis Kellis (57:02.180)
that can all encode the same function.
Lex Fridman (57:04.260)
What evolutionary signatures does
Manolis Kellis (57:06.580)
is that it basically looks at the shape
Lex Fridman (57:08.980)
of that distribution of sequences
Manolis Kellis (57:11.180)
that all encode the same thing.
Lex Fridman (57:13.060)
And based on that shape, you can basically say,
Manolis Kellis (57:15.220)
ooh, proteins have a very different shape
Lex Fridman (57:17.940)
than RNA structures, than regulatory motifs, et cetera.
Lex Fridman (57:21.340)
So just by scanning a sequence, ignoring the sequence
Lex Fridman (57:24.500)
and just looking at the patterns of change,
Manolis Kellis (57:26.740)
I'm like, wow, this thing is evolving like a protein
Lex Fridman (57:29.380)
and that thing is evolving like a motif
Lex Fridman (57:31.700)
and that thing is evolving.
Lex Fridman (57:33.180)
So that's exactly what we just did for COVID.
Lex Fridman (57:35.620)
So our paper that we posted in bioRxiv about coronavirus
Lex Fridman (57:39.020)
basically took this concept of evolutionary signatures
Lex Fridman (57:42.020)
and applied it on the SARS CoV2 genome
Lex Fridman (57:45.740)
that is responsible for the COVID 19 pandemic.
Lex Fridman (57:48.540)
And comparing it to?
Lex Fridman (57:50.540)
To 44 serbicovirus species.
Lex Fridman (57:52.460)
So this is the beta.
Lex Fridman (57:53.700)
What word did you just use, serbicovirus?
Manolis Kellis (57:56.220)
Serbicovirus, so SARS related beta coronavirus.
Lex Fridman (58:00.460)
It's a portmanteau of a bunch.
Lex Fridman (58:01.460)
So that whole family of viruses.
Lex Fridman (58:03.060)
Yeah, so.
Lex Fridman (58:03.900)
How big is that family by the way?
Lex Fridman (58:05.100)
We have 44 species that, or I mean.
Lex Fridman (58:07.420)
There's 44 species in the family?
Lex Fridman (58:09.340)
Yeah. Virus is a clever bunch.
Manolis Kellis (58:11.100)
No, no, but there's just 44.
Lex Fridman (58:12.900)
And again, we don't call them species in viruses.
Manolis Kellis (58:15.660)
We call them strains.
Lex Fridman (58:16.500)
But anyway, there's 44 strains.
Lex Fridman (58:18.260)
And that's a tiny little subset of maybe another 50 strains
Lex Fridman (58:22.300)
that are just far too distantly related.
Manolis Kellis (58:24.460)
Most of those only infect bats as the host
Lex Fridman (58:29.060)
and a subset of only four or five have ever infected humans.
Lex Fridman (58:34.020)
And we basically took all of those
Lex Fridman (58:35.660)
and we aligned them in the same exact way
Manolis Kellis (58:37.740)
that we've aligned mammals.
Lex Fridman (58:39.020)
And then we looked at what proteins are,
Manolis Kellis (58:42.340)
which of the currently hypothesized genes
Lex Fridman (58:44.980)
for the coronavirus genome
Manolis Kellis (58:47.420)
are in fact evolving like proteins and which ones are not.
Lex Fridman (58:50.260)
And what we found is that ORF10,
Manolis Kellis (58:52.940)
the last little open reading frame,
Lex Fridman (58:54.620)
the last little gene in the genome is bogus.
Manolis Kellis (58:56.980)
That's not a protein at all.
Lex Fridman (58:58.700)
What is it?
Manolis Kellis (58:59.900)
It's an RNA structure.
Lex Fridman (59:01.820)
That doesn't have a.
Manolis Kellis (59:03.540)
It doesn't get translated into amino acids.
Lex Fridman (59:05.700)
And that, so it's important to narrow down
Manolis Kellis (59:08.260)
to basically discover what's useful and what's not.
Lex Fridman (59:10.780)
Exactly.
Lex Fridman (59:11.620)
Basically, what is even the set of genes?
Lex Fridman (59:13.580)
The other thing that these evolutionary signatures showed
Manolis Kellis (59:15.500)
is that within ORF3A lies a tiny little additional gene
Lex Fridman (59:20.660)
encoded within the other gene.
Lex Fridman (59:22.700)
So you can translate a DNA sequence
Lex Fridman (59:24.540)
in three different reading frames.
Manolis Kellis (59:26.820)
If you start in the first one, it's ATG, et cetera.
Lex Fridman (59:30.100)
If you start on the second one, it's TGC, et cetera.
Lex Fridman (59:32.940)
And there's a gene within a gene.
Lex Fridman (59:36.620)
So there's a whole other protein
Manolis Kellis (59:37.860)
that we didn't know about that might be super important.
Lex Fridman (59:41.180)
So we don't even know the building blocks of SARS COVID 2.
Lex Fridman (59:45.620)
So if we want to understand coronavirus biology
Lex Fridman (59:48.300)
and eventually find it successfully,
Manolis Kellis (59:50.500)
we need to even have the set of genes
Lex Fridman (59:51.940)
and these evolutionary signatures
Manolis Kellis (59:53.660)
that I developed in my PhD work.
Lex Fridman (59:55.540)
Are you really useful here?
Manolis Kellis (59:56.380)
We just recently used.
Lex Fridman (59:57.420)
You know what, let's run with that tangent
Manolis Kellis (59:59.500)
for a little bit, if it's okay.
🔗 相关节目