Ayanna Howard: Human-Robot Interaction and Ethics of Safety-Critical Systems
AI 与机器学习技术与编程心理与人性音乐与艺术商业与创业
📋 章节目录
暂无章节信息
🔑 关键词
donhumanrobotrobotsroboticstrustdataspacegoinginteractionbetterpersonhumansbiassaidworkingtermsalgorithmshardhaving
💬 精彩语录
暂无语录
🎙️ 完整对话（2134 条）
Lex Fridman (00:00.000)
The following is a conversation with Ayana Howard.
以下是与阿亚娜·霍华德的对话。
Lex Fridman (00:03.380)
She's a roboticist, professor Georgia Tech,
她是一位机器人专家，佐治亚理工学院的教授，
Lex Fridman (00:06.180)
and director of the Human Automation Systems Lab,
人类自动化系统实验室主任
Lex Fridman (00:09.820)
with research interests in human robot interaction,
对人机交互的研究兴趣，
Lex Fridman (00:12.780)
assisted robots in the home, therapy gaming apps,
家庭辅助机器人、治疗游戏应用程序、
Lex Fridman (00:15.980)
and remote robotic exploration of extreme environments.
以及极端环境的远程机器人探索。
Lex Fridman (00:20.260)
Like me, in her work, she cares a lot
和我一样，她在工作中也很关心
Ayanna Howard (00:23.420)
about both robots and human beings,
关于机器人和人类，
Lex Fridman (00:26.340)
and so I really enjoyed this conversation.
所以我真的很喜欢这次谈话。
Ayanna Howard (00:29.540)
This is the Artificial Intelligence Podcast.
这是人工智能播客。
Lex Fridman (00:32.580)
If you enjoy it, subscribe on YouTube,
如果您喜欢，请在 YouTube 上订阅，
Ayanna Howard (00:34.940)
give it five stars on Apple Podcast,
在 Apple Podcast 上给它五颗星，
Lex Fridman (00:36.940)
follow on Spotify, support it on Patreon,
在 Spotify 上关注，在 Patreon 上支持，
Ayanna Howard (00:39.580)
or simply connect with me on Twitter
或者直接在 Twitter 上与我联系
Lex Fridman (00:41.700)
at Lex Friedman, spelled F R I D M A N.
在 Lex Friedman，拼写为 F R I D M A N。
Ayanna Howard (00:45.640)
I recently started doing ads
我最近开始做广告
Lex Fridman (00:47.140)
at the end of the introduction.
在介绍的最后。
Ayanna Howard (00:48.700)
I'll do one or two minutes after introducing the episode,
我会在介绍这一集后做一两分钟，
Lex Fridman (00:51.660)
and never any ads in the middle
中间没有任何广告
Ayanna Howard (00:53.180)
that can break the flow of the conversation.
这可能会破坏谈话的流畅性。
Lex Fridman (00:55.500)
I hope that works for you
Lex Fridman (00:56.860)
and doesn't hurt the listening experience.
Lex Fridman (01:00.140)
This show is presented by Cash App,
Ayanna Howard (01:02.260)
the number one finance app in the App Store.
Lex Fridman (01:04.740)
I personally use Cash App to send money to friends,
Lex Fridman (01:07.540)
but you can also use it to buy, sell,
Lex Fridman (01:09.300)
and deposit Bitcoin in just seconds.
Ayanna Howard (01:11.700)
Cash App also has a new investing feature.
Lex Fridman (01:14.580)
You can buy fractions of a stock, say $1 worth,
Ayanna Howard (01:17.520)
no matter what the stock price is.
Lex Fridman (01:19.640)
Broker services are provided by Cash App Investing,
Ayanna Howard (01:22.560)
a subsidiary of Square and Member SIPC.
Lex Fridman (01:25.840)
I'm excited to be working with Cash App
Ayanna Howard (01:28.140)
to support one of my favorite organizations called First,
Lex Fridman (01:31.540)
best known for their FIRST Robotics and Lego competitions.
Ayanna Howard (01:35.060)
They educate and inspire hundreds of thousands of students
Lex Fridman (01:38.340)
in over 110 countries,
Lex Fridman (01:40.060)
and have a perfect rating at Charity Navigator,
Lex Fridman (01:42.820)
which means that donated money
Ayanna Howard (01:44.100)
is used to maximum effectiveness.
Lex Fridman (01:46.860)
When you get Cash App from the App Store or Google Play
Lex Fridman (01:49.580)
and use code LEXPODCAST, you'll get $10,
Lex Fridman (01:53.500)
and Cash App will also donate $10 to FIRST,
Ayanna Howard (01:56.420)
which again, is an organization
Lex Fridman (01:58.260)
that I've personally seen inspire girls and boys
Ayanna Howard (02:01.060)
to dream of engineering a better world.
Lex Fridman (02:04.260)
And now, here's my conversation with Ayanna Howard.
Lex Fridman (02:09.420)
What or who is the most amazing robot you've ever met,
Lex Fridman (02:13.620)
or perhaps had the biggest impact on your career?
Ayanna Howard (02:16.700)
I haven't met her, but I grew up with her,
Lex Fridman (02:21.060)
but of course, Rosie.
Ayanna Howard (02:22.740)
So, and I think it's because also.
Lex Fridman (02:25.220)
Who's Rosie?
Ayanna Howard (02:26.120)
Rosie from the Jetsons.
Lex Fridman (02:27.780)
She is all things to all people, right?
Ayanna Howard (02:30.940)
Think about it.
Lex Fridman (02:31.780)
Like anything you wanted, it was like magic, it happened.
Lex Fridman (02:35.060)
So people not only anthropomorphize,
Lex Fridman (02:37.860)
but project whatever they wish for the robot to be onto.
Ayanna Howard (02:41.940)
Onto Rosie.
Lex Fridman (02:42.920)
But also, I mean, think about it.
Ayanna Howard (02:44.580)
She was socially engaging.
Lex Fridman (02:46.780)
She every so often had an attitude, right?
Ayanna Howard (02:50.020)
She kept us honest.
Lex Fridman (02:51.940)
She would push back sometimes
Ayanna Howard (02:53.740)
when George was doing some weird stuff.
Lex Fridman (02:56.980)
But she cared about people, especially the kids.
Ayanna Howard (03:01.180)
She was like the perfect robot.
Lex Fridman (03:03.980)
And you've said that people don't want
Ayanna Howard (03:06.460)
their robots to be perfect.
Lex Fridman (03:09.740)
Can you elaborate that?
Lex Fridman (03:11.140)
What do you think that is?
Lex Fridman (03:11.980)
Just like you said, Rosie pushed back a little bit
Ayanna Howard (03:14.780)
every once in a while.
Lex Fridman (03:15.720)
Yeah, so I think it's that.
Lex Fridman (03:18.260)
So if you think about robotics in general,
Lex Fridman (03:19.860)
we want them because they enhance our quality of life.
Lex Fridman (03:23.900)
And usually that's linked to something that's functional.
Lex Fridman (03:27.000)
Even if you think of self driving cars,
Lex Fridman (03:28.640)
why is there a fascination?
Lex Fridman (03:29.980)
Because people really do hate to drive.
Ayanna Howard (03:31.500)
Like there's the like Saturday driving
Lex Fridman (03:34.140)
where I can just speed,
Lex Fridman (03:35.300)
but then there's the I have to go to work every day
Lex Fridman (03:37.500)
and I'm in traffic for an hour.
Ayanna Howard (03:38.980)
I mean, people really hate that.
Lex Fridman (03:40.380)
And so robots are designed to basically enhance
Ayanna Howard (03:45.380)
our ability to increase our quality of life.
Lex Fridman (03:49.740)
And so the perfection comes from this aspect of interaction.
Ayanna Howard (03:55.460)
If I think about how we drive, if we drove perfectly,
Lex Fridman (04:00.020)
we would never get anywhere, right?
Lex Fridman (04:02.140)
So think about how many times you had to run past the light
Lex Fridman (04:07.140)
because you see the car behind you
Ayanna Howard (04:09.020)
is about to crash into you.
Lex Fridman (04:10.380)
Or that little kid kind of runs into the street
Lex Fridman (04:15.320)
and so you have to cross on the other side
Lex Fridman (04:17.300)
because there's no cars, right?
Ayanna Howard (04:18.460)
Like if you think about it, we are not perfect drivers.
Lex Fridman (04:21.220)
Some of it is because it's our world.
Lex Fridman (04:23.580)
And so if you have a robot that is perfect
Lex Fridman (04:26.780)
in that sense of the word,
Ayanna Howard (04:28.740)
they wouldn't really be able to function with us.
Lex Fridman (04:31.180)
Can you linger a little bit on the word perfection?
Lex Fridman (04:34.520)
So from the robotics perspective,
Lex Fridman (04:37.380)
what does that word mean
Lex Fridman (04:39.460)
and how is sort of the optimal behavior
Lex Fridman (04:42.900)
as you're describing different
Lex Fridman (04:44.460)
than what we think is perfection?
Lex Fridman (04:46.620)
Yeah, so perfection, if you think about it
Ayanna Howard (04:49.460)
in the more theoretical point of view,
Lex Fridman (04:51.980)
it's really tied to accuracy, right?
Lex Fridman (04:54.060)
So if I have a function,
Lex Fridman (04:55.620)
can I complete it at 100% accuracy with zero errors?
Lex Fridman (05:00.660)
And so that's kind of, if you think about perfection
Lex Fridman (05:04.180)
in the sense of the word.
Lex Fridman (05:05.220)
And in the self driving car realm,
Lex Fridman (05:07.500)
do you think from a robotics perspective,
Ayanna Howard (05:10.460)
we kind of think that perfection means
Lex Fridman (05:13.940)
following the rules perfectly,
Ayanna Howard (05:15.580)
sort of defining, staying in the lane, changing lanes.
Lex Fridman (05:19.580)
When there's a green light, you go.
Ayanna Howard (05:20.900)
When there's a red light, you stop.
Lex Fridman (05:22.300)
And that's the, and be able to perfectly see
Ayanna Howard (05:26.660)
all the entities in the scene.
Lex Fridman (05:29.140)
That's the limit of what we think of as perfection.
Lex Fridman (05:31.980)
And I think that's where the problem comes
Lex Fridman (05:33.740)
is that when people think about perfection for robotics,
Ayanna Howard (05:38.340)
the ones that are the most successful
Lex Fridman (05:40.820)
are the ones that are quote unquote perfect.
Ayanna Howard (05:43.260)
Like I said, Rosie is perfect,
Lex Fridman (05:44.660)
but she actually wasn't perfect in terms of accuracy,
Lex Fridman (05:47.380)
but she was perfect in terms of how she interacted
Lex Fridman (05:50.380)
and how she adapted.
Lex Fridman (05:51.540)
And I think that's some of the disconnect
Lex Fridman (05:53.300)
is that we really want perfection
Ayanna Howard (05:56.460)
with respect to its ability to adapt to us.
Lex Fridman (05:59.980)
We don't really want perfection with respect to 100% accuracy
Lex Fridman (06:03.500)
with respect to the rules that we just made up anyway, right?
Lex Fridman (06:06.780)
And so I think there's this disconnect sometimes
Ayanna Howard (06:09.500)
between what we really want and what happens.
Lex Fridman (06:13.260)
And we see this all the time, like in my research, right?
Ayanna Howard (06:15.940)
Like the optimal, quote unquote optimal interactions
Lex Fridman (06:20.340)
are when the robot is adapting based on the person,
Ayanna Howard (06:24.300)
not 100% following what's optimal based on the rules.
Lex Fridman (06:29.540)
Just to link on autonomous vehicles for a second,
Ayanna Howard (06:32.580)
just your thoughts, maybe off the top of the head,
Lex Fridman (06:36.180)
how hard is that problem do you think
Lex Fridman (06:37.940)
based on what we just talked about?
Lex Fridman (06:40.100)
There's a lot of folks in the automotive industry,
Ayanna Howard (06:42.900)
they're very confident from Elon Musk to Waymo
Lex Fridman (06:45.900)
to all these companies.
Lex Fridman (06:47.620)
How hard is it to solve that last piece?
Lex Fridman (06:50.420)
The last mile.
Ayanna Howard (06:51.340)
The gap between the perfection and the human definition
Lex Fridman (06:57.500)
of how you actually function in this world.
Ayanna Howard (06:59.460)
Yeah, so this is a moving target.
Lex Fridman (07:00.580)
So I remember when all the big companies
Ayanna Howard (07:04.460)
started to heavily invest in this
Lex Fridman (07:06.780)
and there was a number of even roboticists
Ayanna Howard (07:09.860)
as well as folks who were putting in the VCs
Lex Fridman (07:13.180)
and corporations, Elon Musk being one of them that said,
Ayanna Howard (07:16.660)
self driving cars on the road with people
Lex Fridman (07:19.460)
within five years, that was a little while ago.
Lex Fridman (07:24.180)
And now people are saying five years, 10 years, 20 years,
Lex Fridman (07:29.780)
some are saying never, right?
Ayanna Howard (07:31.500)
I think if you look at some of the things
Lex Fridman (07:33.700)
that are being successful is these
Ayanna Howard (07:39.420)
basically fixed environments
Lex Fridman (07:41.140)
where you still have some anomalies, right?
Ayanna Howard (07:43.980)
You still have people walking, you still have stores,
Lex Fridman (07:46.460)
but you don't have other drivers, right?
Ayanna Howard (07:50.060)
Like other human drivers are,
Lex Fridman (07:51.700)
it's a dedicated space for the cars.
Ayanna Howard (07:55.580)
Because if you think about robotics in general,
Lex Fridman (07:57.140)
where has always been successful?
Ayanna Howard (07:59.020)
I mean, you can say manufacturing,
Lex Fridman (08:00.580)
like way back in the day, right?
Ayanna Howard (08:02.260)
It was a fixed environment, humans were not part
Lex Fridman (08:04.340)
of the equation, we're a lot better than that.
Lex Fridman (08:07.180)
But like when we can carve out scenarios
Lex Fridman (08:10.940)
that are closer to that space,
Ayanna Howard (08:13.780)
then I think that it's where we are.
Lex Fridman (08:16.660)
So a closed campus where you don't have self driving cars
Lex Fridman (08:20.540)
and maybe some protection so that the students
Lex Fridman (08:23.780)
don't jet in front just because they wanna see what happens.
Ayanna Howard (08:27.220)
Like having a little bit, I think that's where
Lex Fridman (08:29.940)
we're gonna see the most success in the near future.
Lex Fridman (08:32.300)
And be slow moving.
Lex Fridman (08:33.660)
Right, not 55, 60, 70 miles an hour,
Lex Fridman (08:37.900)
but the speed of a golf cart, right?
Lex Fridman (08:42.100)
So that said, the most successful
Ayanna Howard (08:45.220)
in the automotive industry robots operating today
Lex Fridman (08:47.900)
in the hands of real people are ones that are traveling
Ayanna Howard (08:51.600)
over 55 miles an hour and in unconstrained environments,
Lex Fridman (08:55.540)
which is Tesla vehicles, so Tesla autopilot.
Lex Fridman (08:58.880)
So I would love to hear sort of your,
Lex Fridman (09:01.720)
just thoughts of two things.
Lex Fridman (09:04.300)
So one, I don't know if you've gotten to see,
Lex Fridman (09:07.020)
you've heard about something called smart summon
Ayanna Howard (09:10.020)
where Tesla system, autopilot system,
Lex Fridman (09:13.520)
where the car drives zero occupancy, no driver
Ayanna Howard (09:17.140)
in the parking lot slowly sort of tries to navigate
Lex Fridman (09:19.980)
the parking lot to find itself to you.
Lex Fridman (09:22.720)
And there's some incredible amounts of videos
Lex Fridman (09:25.900)
and just hilarity that happens as it awkwardly tries
Ayanna Howard (09:28.860)
to navigate this environment, but it's a beautiful
Lex Fridman (09:32.340)
nonverbal communication between machine and human
Ayanna Howard (09:35.180)
that I think is a, it's like, it's some of the work
Lex Fridman (09:38.780)
that you do in this kind of interesting
Ayanna Howard (09:40.660)
human robot interaction space.
Lex Fridman (09:42.060)
So what are your thoughts in general about it?
Lex Fridman (09:43.780)
So I do have that feature.
Lex Fridman (09:46.980)
Do you drive a Tesla?
Lex Fridman (09:47.820)
I do, mainly because I'm a gadget freak, right?
Lex Fridman (09:52.100)
So I say it's a gadget that happens to have some wheels.
Lex Fridman (09:55.620)
And yeah, I've seen some of the videos.
Lex Fridman (09:58.220)
But what's your experience like?
Ayanna Howard (09:59.420)
I mean, you're a human robot interaction roboticist,
Lex Fridman (10:02.700)
you're a legit sort of expert in the field.
Lex Fridman (10:05.580)
So what does it feel for a machine to come to you?
Lex Fridman (10:08.260)
It's one of these very fascinating things,
Lex Fridman (10:11.900)
but also I am hyper, hyper alert, right?
Lex Fridman (10:16.100)
Like I'm hyper alert, like my butt, my thumb is like,
Ayanna Howard (10:20.540)
oh, okay, I'm ready to take over.
Lex Fridman (10:23.220)
Even when I'm in my car or I'm doing things like automated
Ayanna Howard (10:27.080)
backing into, so there's like a feature where you can do
Lex Fridman (10:30.420)
this automating backing into a parking space,
Ayanna Howard (10:33.140)
or bring the car out of your garage,
Lex Fridman (10:35.660)
or even, you know, pseudo autopilot on the freeway, right?
Ayanna Howard (10:40.260)
I am hypersensitive.
Lex Fridman (10:42.220)
I can feel like as I'm navigating,
Ayanna Howard (10:44.720)
like, yeah, that's an error right there.
Lex Fridman (10:46.900)
Like I am very aware of it, but I'm also fascinated by it.
Lex Fridman (10:52.260)
And it does get better.
Lex Fridman (10:54.300)
Like I look and see it's learning from all of these people
Ayanna Howard (10:58.980)
who are cutting it on, like every time I cut it on,
Lex Fridman (11:02.700)
it's getting better, right?
Lex Fridman (11:04.120)
And so I think that's what's amazing about it is that.
Lex Fridman (11:07.100)
This nice dance of you're still hyper vigilant.
Lex Fridman (11:10.340)
So you're still not trusting it at all.
Lex Fridman (11:12.780)
Yeah.
Lex Fridman (11:13.600)
And yet you're using it.
Lex Fridman (11:14.580)
On the highway, if I were to, like what,
Ayanna Howard (11:17.580)
as a roboticist, we'll talk about trust a little bit.
Lex Fridman (11:22.640)
How do you explain that?
Ayanna Howard (11:23.640)
You still use it.
Lex Fridman (11:25.020)
Is it the gadget freak part?
Lex Fridman (11:26.460)
Like where you just enjoy exploring technology?
Lex Fridman (11:30.700)
Or is that the right actually balance
Ayanna Howard (11:33.680)
between robotics and humans is where you use it,
Lex Fridman (11:36.860)
but don't trust it.
Lex Fridman (11:38.340)
And somehow there's this dance
Lex Fridman (11:40.100)
that ultimately is a positive.
Ayanna Howard (11:42.100)
Yeah, so I think I'm,
Lex Fridman (11:44.620)
I just don't necessarily trust technology,
Lex Fridman (11:48.080)
but I'm an early adopter, right?
Lex Fridman (11:50.140)
So when it first comes out,
Ayanna Howard (11:51.960)
I will use everything,
Lex Fridman (11:54.260)
but I will be very, very cautious of how I use it.
Lex Fridman (11:57.420)
Do you read about it or do you explore it by just try it?
Lex Fridman (12:01.020)
Do you like crudely, to put it crudely,
Lex Fridman (12:04.980)
do you read the manual or do you learn through exploration?
Lex Fridman (12:07.960)
I'm an explorer.
Ayanna Howard (12:08.800)
If I have to read the manual, then I do design.
Lex Fridman (12:12.320)
Then it's a bad user interface.
Ayanna Howard (12:14.180)
It's a failure.
Lex Fridman (12:16.460)
Elon Musk is very confident that you kind of take it
Ayanna Howard (12:19.540)
from where it is now to full autonomy.
Lex Fridman (12:21.780)
So from this human robot interaction,
Ayanna Howard (12:24.500)
where you don't really trust and then you try
Lex Fridman (12:26.700)
and then you catch it when it fails to,
Ayanna Howard (12:29.180)
it's going to incrementally improve itself
Lex Fridman (12:32.300)
into full where you don't need to participate.
Lex Fridman (12:36.500)
What's your sense of that trajectory?
Lex Fridman (12:39.860)
Is it feasible?
Lex Fridman (12:41.040)
So the promise there is by the end of next year,
Lex Fridman (12:44.580)
by the end of 2020 is the current promise.
Lex Fridman (12:47.180)
What's your sense about that journey that Tesla's on?
Lex Fridman (12:53.620)
So there's kind of three things going on though.
Ayanna Howard (12:56.580)
I think in terms of will people go like as a user,
Lex Fridman (13:03.260)
as a adopter, will you trust going to that point?
Lex Fridman (13:08.460)
I think so, right?
Lex Fridman (13:10.080)
Like there are some users and it's because what happens is
Ayanna Howard (13:13.020)
when you're hypersensitive at the beginning
Lex Fridman (13:16.700)
and then the technology tends to work,
Ayanna Howard (13:19.300)
your apprehension slowly goes away.
Lex Fridman (13:23.820)
And as people, we tend to swing to the other extreme, right?
Ayanna Howard (13:28.260)
Because it's like, oh, I was like hyper, hyper fearful
Lex Fridman (13:30.900)
or hypersensitive and it was awesome.
Lex Fridman (13:33.940)
And we just tend to swing.
Lex Fridman (13:35.600)
That's just human nature.
Lex Fridman (13:37.380)
And so you will have, I mean, and I...
Lex Fridman (13:38.860)
That's a scary notion because most people
Ayanna Howard (13:41.520)
are now extremely untrusting of autopilot.
Lex Fridman (13:44.980)
They use it, but they don't trust it.
Lex Fridman (13:46.460)
And it's a scary notion that there's a certain point
Lex Fridman (13:48.900)
where you allow yourself to look at the smartphone
Ayanna Howard (13:51.340)
for like 20 seconds.
Lex Fridman (13:53.100)
And then there'll be this phase shift
Ayanna Howard (13:55.300)
where it'll be like 20 seconds, 30 seconds,
Lex Fridman (13:57.580)
one minute, two minutes.
Ayanna Howard (13:59.980)
It's a scary proposition.
Lex Fridman (14:02.020)
But that's people, right?
Ayanna Howard (14:03.460)
That's just, that's humans.
Lex Fridman (14:05.560)
I mean, I think of even our use of,
Lex Fridman (14:09.980)
I mean, just everything on the internet, right?
Lex Fridman (14:12.380)
Like think about how reliant we are on certain apps
Lex Fridman (14:16.860)
and certain engines, right?
Lex Fridman (14:20.260)
20 years ago, people have been like, oh yeah, that's stupid.
Ayanna Howard (14:22.680)
Like that makes no sense.
Lex Fridman (14:23.940)
Like, of course that's false.
Ayanna Howard (14:25.900)
Like now it's just like, oh, of course I've been using it.
Lex Fridman (14:29.100)
It's been correct all this time.
Ayanna Howard (14:30.740)
Of course aliens, I didn't think they existed,
Lex Fridman (14:34.340)
but now it says they do, obviously.
Ayanna Howard (14:37.620)
100%, earth is flat.
Lex Fridman (14:39.500)
So, okay, but you said three things.
Lex Fridman (14:43.860)
So one is the human.
Lex Fridman (14:44.700)
Okay, so one is the human.
Lex Fridman (14:45.820)
And I think there will be a group of individuals
Lex Fridman (14:47.820)
that will swing, right?
Ayanna Howard (14:49.580)
I just.
Lex Fridman (14:50.420)
Teenagers.
Ayanna Howard (14:51.260)
Teenage, I mean, it'll be, it'll be adults.
Lex Fridman (14:54.380)
There's actually an age demographic
Ayanna Howard (14:56.400)
that's optimal for technology adoption.
Lex Fridman (15:00.140)
And you can actually find them.
Lex Fridman (15:02.260)
And they're actually pretty easy to find.
Lex Fridman (15:03.940)
Just based on their habits, based on,
Lex Fridman (15:06.100)
so if someone like me who wasn't a roboticist
Lex Fridman (15:10.420)
would probably be the optimal kind of person, right?
Ayanna Howard (15:13.580)
Early adopter, okay with technology,
Lex Fridman (15:15.660)
very comfortable and not hypersensitive, right?
Ayanna Howard (15:20.020)
I'm just hypersensitive cause I designed this stuff.
Lex Fridman (15:23.580)
So there is a target demographic that will swing.
Ayanna Howard (15:25.940)
The other one though,
Lex Fridman (15:26.820)
is you still have these humans that are on the road.
Ayanna Howard (15:31.380)
That one is a harder, harder thing to do.
Lex Fridman (15:35.100)
And as long as we have people that are on the same streets,
Ayanna Howard (15:40.660)
that's gonna be the big issue.
Lex Fridman (15:42.480)
And it's just because you can't possibly,
Ayanna Howard (15:45.260)
I wanna say you can't possibly map the,
Lex Fridman (15:48.020)
some of the silliness of human drivers, right?
Ayanna Howard (15:51.380)
Like as an example, when you're next to that car
Lex Fridman (15:56.240)
that has that big sticker called student driver, right?
Ayanna Howard (15:59.780)
Like you are like, oh, either I'm going to like go around.
Lex Fridman (16:04.580)
Like we are, we know that that person
Lex Fridman (16:06.740)
is just gonna make mistakes that make no sense, right?
Lex Fridman (16:09.260)
How do you map that information?
Ayanna Howard (16:11.860)
Or if I am in a car and I look over
Lex Fridman (16:14.300)
and I see two fairly young looking individuals
Lex Fridman (16:19.220)
and there's no student driver bumper
Lex Fridman (16:21.100)
and I see them chit chatting to each other,
Lex Fridman (16:22.820)
I'm like, oh, that's an issue, right?
Lex Fridman (16:26.140)
So how do you get that kind of information
Lex Fridman (16:28.420)
and that experience into basically an autopilot?
Lex Fridman (16:35.660)
And there's millions of cases like that
Ayanna Howard (16:37.260)
where we take little hints to establish context.
Lex Fridman (16:41.220)
I mean, you said kind of beautifully poetic human things,
Lex Fridman (16:44.360)
but there's probably subtle things about the environment
Lex Fridman (16:47.120)
about it being maybe time for commuters
Ayanna Howard (16:52.900)
to start going home from work
Lex Fridman (16:55.220)
and therefore you can make some kind of judgment
Ayanna Howard (16:57.140)
about the group behavior of pedestrians, blah, blah, blah,
Lex Fridman (17:00.060)
and so on and so on.
Lex Fridman (17:01.180)
Or even cities, right?
Lex Fridman (17:02.660)
Like if you're in Boston, how people cross the street,
Ayanna Howard (17:07.100)
like lights are not an issue versus other places
Lex Fridman (17:10.660)
where people will actually wait for the crosswalk.
Ayanna Howard (17:15.580)
Seattle or somewhere peaceful.
Lex Fridman (17:18.940)
But what I've also seen sort of just even in Boston
Ayanna Howard (17:22.540)
that intersection to intersection is different.
Lex Fridman (17:25.500)
So every intersection has a personality of its own.
Lex Fridman (17:28.940)
So certain neighborhoods of Boston are different.
Lex Fridman (17:30.860)
So we kind of, and based on different timing of day,
Ayanna Howard (17:35.220)
at night, it's all, there's a dynamic to human behavior
Lex Fridman (17:40.320)
that we kind of figure out ourselves.
Ayanna Howard (17:42.420)
We're not able to introspect and figure it out,
Lex Fridman (17:46.100)
but somehow our brain learns it.
Ayanna Howard (17:49.340)
We do.
Lex Fridman (17:50.340)
And so you're saying, is there a shortcut?
Lex Fridman (17:54.860)
Is there a shortcut, though, for a robot?
Lex Fridman (17:56.420)
Is there something that could be done, you think,
Ayanna Howard (17:59.060)
that, you know, that's what we humans do.
Lex Fridman (18:02.660)
It's just like bird flight, right?
Ayanna Howard (18:04.660)
That's the example they give for flight.
Lex Fridman (18:06.500)
Do you necessarily need to build a bird that flies
Lex Fridman (18:09.260)
or can you do an airplane?
Lex Fridman (18:11.860)
Is there a shortcut to it?
Lex Fridman (18:13.020)
So I think the shortcut is, and I kind of,
Lex Fridman (18:16.700)
I talk about it as a fixed space,
Ayanna Howard (18:19.340)
where, so imagine that there's a neighborhood
Lex Fridman (18:23.280)
that's a new smart city or a new neighborhood
Lex Fridman (18:26.500)
that says, you know what?
Lex Fridman (18:27.540)
We are going to design this new city
Ayanna Howard (18:31.460)
based on supporting self driving cars.
Lex Fridman (18:33.660)
And then doing things, knowing that there's anomalies,
Lex Fridman (18:37.660)
knowing that people are like this, right?
Lex Fridman (18:39.620)
And designing it based on that assumption
Ayanna Howard (18:42.080)
that like, we're gonna have this.
Lex Fridman (18:43.940)
That would be an example of a shortcut.
Lex Fridman (18:45.540)
So you still have people,
Lex Fridman (18:47.140)
but you do very specific things
Ayanna Howard (18:49.260)
to try to minimize the noise a little bit
Lex Fridman (18:51.740)
as an example.
Lex Fridman (18:53.820)
And the people themselves become accepting of the notion
Lex Fridman (18:56.180)
that there's autonomous cars, right?
Ayanna Howard (18:57.740)
Right, like they move into,
Lex Fridman (18:59.700)
so right now you have like a,
Lex Fridman (19:01.420)
you will have a self selection bias, right?
Lex Fridman (19:03.580)
Like individuals will move into this neighborhood
Ayanna Howard (19:06.180)
knowing like this is part of like the real estate pitch,
Lex Fridman (19:09.420)
right?
Lex Fridman (19:10.620)
And so I think that's a way to do a shortcut.
Lex Fridman (19:14.140)
One, it allows you to deploy.
Ayanna Howard (19:17.540)
It allows you to collect then data with these variances
Lex Fridman (19:21.900)
and anomalies, cause people are still people,
Lex Fridman (19:24.020)
but it's a safer space and it's more of an accepting space.
Lex Fridman (19:28.820)
I.e. when something in that space might happen
Ayanna Howard (19:31.900)
because things do,
Lex Fridman (19:34.100)
because you already have the self selection,
Ayanna Howard (19:36.060)
like people would be, I think a little more forgiving
Lex Fridman (19:39.220)
than other places.
Lex Fridman (19:40.700)
And you said three things, did we cover all of them?
Lex Fridman (19:43.100)
The third is legal law, liability,
Ayanna Howard (19:46.340)
which I don't really want to touch,
Lex Fridman (19:47.820)
but it's still of concern.
Lex Fridman (19:50.900)
And the mishmash with like with policy as well,
Lex Fridman (19:53.260)
sort of government, all that whole.
Ayanna Howard (19:55.740)
That big ball of stuff.
Lex Fridman (19:57.740)
Yeah, gotcha.
Lex Fridman (19:59.100)
So that's, so we're out of time now.
Lex Fridman (1:00:05.680)
under the lens of robotics,
Ayanna Howard (1:00:09.560)
having hardware, interacting with people.
Lex Fridman (1:00:12.100)
So you're a world class expert researcher in robotics,
Lex Fridman (1:00:17.840)
and yet others, you know, there's a few,
Lex Fridman (1:00:21.120)
it's a small but fierce community of people,
Lex Fridman (1:00:24.160)
but most of them don't take the journey
Lex Fridman (1:00:26.600)
into the H of HRI, into the human.
Lex Fridman (1:00:29.440)
So why did you brave into the interaction with humans?
Lex Fridman (1:00:34.440)
It seems like a really hard problem.
Ayanna Howard (1:00:36.880)
It's a hard problem, and it's very risky as an academic.
Lex Fridman (1:00:41.080)
And I knew that when I started down that journey,
Ayanna Howard (1:00:46.200)
that it was very risky as an academic
Lex Fridman (1:00:49.880)
in this world that was nuance, it was just developing.
Ayanna Howard (1:00:53.440)
We didn't even have a conference, right, at the time.
Lex Fridman (1:00:56.720)
Because it was the interesting problems.
Ayanna Howard (1:01:00.120)
That was what drove me.
Lex Fridman (1:01:01.560)
It was the fact that I looked at what interests me
Ayanna Howard (1:01:06.920)
in terms of the application space and the problems.
Lex Fridman (1:01:10.400)
And that pushed me into trying to figure out
Lex Fridman (1:01:14.900)
what people were and what humans were
Lex Fridman (1:01:16.840)
and how to adapt to them.
Ayanna Howard (1:01:19.040)
If those problems weren't so interesting,
Lex Fridman (1:01:21.280)
I'd probably still be sending rovers to glaciers, right?
Lex Fridman (1:01:26.280)
But the problems were interesting.
Lex Fridman (1:01:28.080)
And the other thing was that they were hard, right?
Lex Fridman (1:01:30.600)
So it's, I like having to go into a room
Lex Fridman (1:01:34.560)
and being like, I don't know what to do.
Lex Fridman (1:01:37.000)
And then going back and saying, okay,
Lex Fridman (1:01:38.280)
I'm gonna figure this out.
Ayanna Howard (1:01:39.800)
I do not, I'm not driven when I go in like,
Lex Fridman (1:01:42.320)
oh, there are no surprises.
Ayanna Howard (1:01:44.040)
Like, I don't find that satisfying.
Lex Fridman (1:01:47.320)
If that was the case,
Lex Fridman (1:01:48.160)
I'd go someplace and make a lot more money, right?
Lex Fridman (1:01:51.020)
I think I stay in academic because and choose to do this
Ayanna Howard (1:01:55.000)
because I can go into a room and like, that's hard.
Lex Fridman (1:01:58.280)
Yeah, I think just from my perspective,
Ayanna Howard (1:02:01.720)
maybe you can correct me on it,
Lex Fridman (1:02:03.200)
but if I just look at the field of AI broadly,
Ayanna Howard (1:02:06.720)
it seems that human robot interaction has the most,
Lex Fridman (1:02:12.020)
one of the most number of open problems.
Ayanna Howard (1:02:16.540)
Like people, especially relative to how many people
Lex Fridman (1:02:20.280)
are willing to acknowledge that there are this,
Ayanna Howard (1:02:23.920)
because most people are just afraid of the humans
Lex Fridman (1:02:26.160)
so they don't even acknowledge
Lex Fridman (1:02:27.240)
how many open problems there are.
Lex Fridman (1:02:28.200)
But it's in terms of difficult problems
Ayanna Howard (1:02:30.440)
to solve exciting spaces,
Lex Fridman (1:02:32.400)
it seems to be incredible for that.
Ayanna Howard (1:02:35.840)
It is, and it's exciting.
Lex Fridman (1:02:38.680)
You've mentioned trust before.
Lex Fridman (1:02:40.040)
What role does trust from interacting with autopilot
Lex Fridman (1:02:46.860)
to in the medical context,
Lex Fridman (1:02:48.480)
what role does trust play in the human robot interactions?
Lex Fridman (1:02:51.320)
So some of the things I study in this domain
Ayanna Howard (1:02:53.920)
is not just trust, but it really is over trust.
Lex Fridman (1:02:56.920)
How do you think about over trust?
Ayanna Howard (1:02:58.160)
Like what is, first of all, what is trust
Lex Fridman (1:03:02.280)
and what is over trust?
Ayanna Howard (1:03:03.360)
Basically, the way I look at it is,
Lex Fridman (1:03:05.780)
trust is not what you click on a survey,
Ayanna Howard (1:03:08.040)
trust is about your behavior.
Lex Fridman (1:03:09.560)
So if you interact with the technology
Ayanna Howard (1:03:13.460)
based on the decision or the actions of the technology
Lex Fridman (1:03:17.280)
as if you trust that decision, then you're trusting.
Lex Fridman (1:03:22.360)
And even in my group, we've done surveys
Lex Fridman (1:03:25.560)
that on the thing, do you trust robots?
Ayanna Howard (1:03:28.240)
Of course not.
Lex Fridman (1:03:29.080)
Would you follow this robot in a burdening building?
Ayanna Howard (1:03:31.640)
Of course not.
Lex Fridman (1:03:32.920)
And then you look at their actions and you're like,
Ayanna Howard (1:03:35.480)
clearly your behavior does not match what you think
Lex Fridman (1:03:39.640)
or what you think you would like to think.
Lex Fridman (1:03:42.000)
And so I'm really concerned about the behavior
Lex Fridman (1:03:44.040)
because that's really at the end of the day,
Ayanna Howard (1:03:45.800)
when you're in the world,
Lex Fridman (1:03:47.340)
that's what will impact others around you.
Ayanna Howard (1:03:50.500)
It's not whether before you went onto the street,
Lex Fridman (1:03:52.920)
you clicked on like, I don't trust self driving cars.
Ayanna Howard (1:03:55.640)
Yeah, that from an outsider perspective,
Lex Fridman (1:03:58.680)
it's always frustrating to me.
Ayanna Howard (1:04:00.600)
Well, I read a lot, so I'm insider
Lex Fridman (1:04:02.480)
in a certain philosophical sense.
Ayanna Howard (1:04:06.040)
It's frustrating to me how often trust is used in surveys
Lex Fridman (1:04:10.680)
and how people say, make claims out of any kind of finding
Ayanna Howard (1:04:15.680)
they make while somebody clicking on answer.
Lex Fridman (1:04:18.680)
You just trust is a, yeah, behavior just,
Ayanna Howard (1:04:23.700)
you said it beautifully.
Lex Fridman (1:04:24.580)
I mean, the action, your own behavior is what trust is.
Ayanna Howard (1:04:28.080)
I mean, that everything else is not even close.
Lex Fridman (1:04:30.740)
It's almost like absurd comedic poetry
Ayanna Howard (1:04:36.040)
that you weave around your actual behavior.
Lex Fridman (1:04:38.500)
So some people can say their trust,
Ayanna Howard (1:04:41.780)
you know, I trust my wife, husband or not,
Lex Fridman (1:04:45.620)
whatever, but the actions is what speaks volumes.
Ayanna Howard (1:04:48.260)
You bug their car, you probably don't trust them.
Lex Fridman (1:04:52.260)
I trust them, I'm just making sure.
Ayanna Howard (1:04:53.820)
No, no, that's, yeah.
Lex Fridman (1:04:55.620)
Like even if you think about cars,
Ayanna Howard (1:04:57.260)
I think it's a beautiful case.
Lex Fridman (1:04:58.580)
I came here at some point, I'm sure,
Lex Fridman (1:05:01.260)
on either Uber or Lyft, right?
Lex Fridman (1:05:03.580)
I remember when it first came out, right?
Ayanna Howard (1:05:06.020)
I bet if they had had a survey,
Lex Fridman (1:05:08.020)
would you get in the car with a stranger and pay them?
Ayanna Howard (1:05:11.420)
Yes.
Lex Fridman (1:05:12.660)
How many people do you think would have said,
Lex Fridman (1:05:15.300)
like, really?
Lex Fridman (1:05:16.620)
Wait, even worse, would you get in the car
Ayanna Howard (1:05:18.660)
with a stranger at 1 a.m. in the morning
Lex Fridman (1:05:21.900)
to have them drop you home as a single female?
Ayanna Howard (1:05:24.780)
Yeah.
Lex Fridman (1:05:25.620)
Like how many people would say, that's stupid.
Ayanna Howard (1:05:29.280)
Yeah.
Lex Fridman (1:05:30.120)
And now look at where we are.
Lex Fridman (1:05:31.540)
I mean, people put kids, right?
Lex Fridman (1:05:33.940)
Like, oh yeah, my child has to go to school
Lex Fridman (1:05:37.660)
and yeah, I'm gonna put my kid in this car with a stranger.
Lex Fridman (1:05:42.300)
I mean, it's just fascinating how, like,
Lex Fridman (1:05:45.580)
what we think we think is not necessarily
Lex Fridman (1:05:48.260)
matching our behavior.
Ayanna Howard (1:05:49.620)
Yeah, and certainly with robots, with autonomous vehicles
Lex Fridman (1:05:52.260)
and all the kinds of robots you work with,
Ayanna Howard (1:05:54.620)
that's, it's, yeah, it's, the way you answer it,
Lex Fridman (1:06:00.340)
especially if you've never interacted with that robot before,
Ayanna Howard (1:06:04.300)
if you haven't had the experience,
Lex Fridman (1:06:05.620)
you being able to respond correctly on a survey is impossible.
Lex Fridman (1:06:09.540)
But what do you, what role does trust play
Lex Fridman (1:06:12.460)
in the interaction, do you think?
Lex Fridman (1:06:14.220)
Like, is it good to, is it good to trust a robot?
Lex Fridman (1:06:19.380)
What does over trust mean?
Ayanna Howard (1:06:21.620)
Or is it, is it good to kind of how you feel
Lex Fridman (1:06:23.980)
about autopilot currently, which is like,
Ayanna Howard (1:06:26.460)
from a roboticist's perspective, is like,
Lex Fridman (1:06:29.380)
oh, still very cautious?
Ayanna Howard (1:06:31.460)
Yeah, so this is still an open area of research,
Lex Fridman (1:06:34.860)
but basically what I would like in a perfect world
Ayanna Howard (1:06:40.700)
is that people trust the technology when it's working 100%,
Lex Fridman (1:06:44.900)
and people will be hypersensitive
Lex Fridman (1:06:47.260)
and identify when it's not.
Lex Fridman (1:06:49.060)
But of course we're not there.
Ayanna Howard (1:06:50.940)
That's the ideal world.
Lex Fridman (1:06:53.620)
And, but we find is that people swing, right?
Ayanna Howard (1:06:56.460)
They tend to swing, which means that if my first,
Lex Fridman (1:07:01.300)
and like, we have some papers,
Lex Fridman (1:07:02.900)
like first impressions is everything, right?
Lex Fridman (1:07:05.260)
If my first instance with technology,
Ayanna Howard (1:07:07.620)
with robotics is positive, it mitigates any risk,
Lex Fridman (1:07:12.700)
it correlates with like best outcomes,
Ayanna Howard (1:07:16.860)
it means that I'm more likely to either not see it
Lex Fridman (1:07:21.460)
when it makes some mistakes or faults,
Ayanna Howard (1:07:24.180)
or I'm more likely to forgive it.
Lex Fridman (1:07:28.660)
And so this is a problem
Lex Fridman (1:07:30.340)
because technology is not 100% accurate, right?
Lex Fridman (1:07:32.620)
It's not 100% accurate, although it may be perfect.
Lex Fridman (1:07:35.100)
How do you get that first moment right, do you think?
Lex Fridman (1:07:37.700)
There's also an education about the capabilities
Lex Fridman (1:07:40.740)
and limitations of the system.
Lex Fridman (1:07:42.500)
Do you have a sense of how do you educate people correctly
Lex Fridman (1:07:45.740)
in that first interaction?
Lex Fridman (1:07:47.140)
Again, this is an open ended problem.
Lex Fridman (1:07:50.260)
So one of the study that actually has given me some hope
Lex Fridman (1:07:55.020)
that I were trying to figure out how to put in robotics.
Lex Fridman (1:07:57.660)
So there was a research study
Lex Fridman (1:08:01.300)
that it showed for medical AI systems,
Ayanna Howard (1:08:03.460)
giving information to radiologists about,
Lex Fridman (1:08:07.820)
here you need to look at these areas on the X ray.
Lex Fridman (1:08:13.980)
What they found was that when the system provided
Lex Fridman (1:08:18.900)
one choice, there was this aspect of either no trust
Lex Fridman (1:08:25.340)
or over trust, right?
Lex Fridman (1:08:26.860)
Like I don't believe it at all,
Ayanna Howard (1:08:29.820)
or a yes, yes, yes, yes.
Lex Fridman (1:08:33.580)
And they would miss things, right?
Ayanna Howard (1:08:36.380)
Instead, when the system gave them multiple choices,
Lex Fridman (1:08:40.580)
like here are the three, even if it knew like,
Ayanna Howard (1:08:43.260)
it had estimated that the top area you need to look at
Lex Fridman (1:08:45.940)
was some place on the X ray.
Ayanna Howard (1:08:49.780)
If it gave like one plus others,
Lex Fridman (1:08:54.060)
the trust was maintained and the accuracy of the entire
Lex Fridman (1:09:00.420)
population increased, right?
Lex Fridman (1:09:03.580)
So basically it was a, you're still trusting the system,
Lex Fridman (1:09:07.500)
but you're also putting in a little bit of like,
Lex Fridman (1:09:09.580)
your human expertise, like your human decision processing
Ayanna Howard (1:09:13.660)
into the equation.
Lex Fridman (1:09:15.540)
So it helps to mitigate that over trust risk.
Ayanna Howard (1:09:18.540)
Yeah, so there's a fascinating balance that the strike.
Lex Fridman (1:09:21.580)
Haven't figured out again, robotics is still an open research.
Ayanna Howard (1:09:24.420)
This is exciting open area research, exactly.
Lex Fridman (1:09:26.740)
So what are some exciting applications
Lex Fridman (1:09:28.940)
of human robot interaction?
Lex Fridman (1:09:30.180)
You started a company, maybe you can talk about
Ayanna Howard (1:09:33.060)
the exciting efforts there, but in general also
Lex Fridman (1:09:36.740)
what other space can robots interact with humans and help?
Ayanna Howard (1:09:41.020)
Yeah, so besides healthcare,
Lex Fridman (1:09:42.340)
cause you know, that's my bias lens.
Ayanna Howard (1:09:44.540)
My other bias lens is education.
Lex Fridman (1:09:47.100)
I think that, well, one, we definitely,
Ayanna Howard (1:09:51.260)
we in the US, you know, we're doing okay with teachers,
Lex Fridman (1:09:54.780)
but there's a lot of school districts
Ayanna Howard (1:09:56.860)
that don't have enough teachers.
Lex Fridman (1:09:58.300)
If you think about the teacher student ratio
Ayanna Howard (1:10:01.940)
for at least public education in some districts, it's crazy.
Lex Fridman (1:10:06.700)
It's like, how can you have learning in that classroom,
Lex Fridman (1:10:10.020)
right?
Lex Fridman (1:10:10.860)
Because you just don't have the human capital.
Lex Fridman (1:10:12.980)
And so if you think about robotics,
Lex Fridman (1:10:15.500)
bringing that in to classrooms,
Ayanna Howard (1:10:18.460)
as well as the afterschool space,
Lex Fridman (1:10:20.340)
where they offset some of this lack of resources
Ayanna Howard (1:10:25.100)
in certain communities, I think that's a good place.
Lex Fridman (1:10:28.460)
And then turning on the other end
Ayanna Howard (1:10:30.900)
is using these systems then for workforce retraining
Lex Fridman (1:10:35.260)
and dealing with some of the things
Ayanna Howard (1:10:38.940)
that are going to come out later on of job loss,
Lex Fridman (1:10:43.020)
like thinking about robots and in AI systems
Ayanna Howard (1:10:45.900)
for retraining and workforce development.
Lex Fridman (1:10:48.340)
I think that's exciting areas that can be pushed even more,
Lex Fridman (1:10:53.220)
and it would have a huge, huge impact.
Lex Fridman (1:10:56.780)
What would you say are some of the open problems
Ayanna Howard (1:10:59.620)
in education, sort of, it's exciting.
Lex Fridman (1:11:03.220)
So young kids and the older folks
Ayanna Howard (1:11:08.740)
or just folks of all ages who need to be retrained,
Lex Fridman (1:11:12.580)
who need to sort of open themselves up
Ayanna Howard (1:11:14.260)
to a whole nother area of work.
Lex Fridman (1:11:17.700)
What are the problems to be solved there?
Lex Fridman (1:11:20.060)
How do you think robots can help?
Lex Fridman (1:11:22.460)
We have the engagement aspect, right?
Lex Fridman (1:11:24.820)
So we can figure out the engagement.
Lex Fridman (1:11:26.460)
That's not a...
Lex Fridman (1:11:27.300)
What do you mean by engagement?
Lex Fridman (1:11:28.900)
So identifying whether a person is focused,
Ayanna Howard (1:11:34.940)
is like that we can figure out.
Lex Fridman (1:11:38.740)
What we can figure out and there's some positive results
Ayanna Howard (1:11:43.900)
in this is that personalized adaptation
Lex Fridman (1:11:47.180)
based on any concepts, right?
Lex Fridman (1:11:49.660)
So imagine I think about, I have an agent
Lex Fridman (1:11:54.580)
and I'm working with a kid learning, I don't know,
Ayanna Howard (1:11:59.620)
algebra two, can that same agent then switch
Lex Fridman (1:12:03.820)
and teach some type of new coding skill
Lex Fridman (1:12:07.980)
to a displaced mechanic?
Lex Fridman (1:12:11.420)
Like, what does that actually look like, right?
Ayanna Howard (1:12:14.500)
Like hardware might be the same, content is different,
Lex Fridman (1:12:19.540)
two different target demographics of engagement.
Lex Fridman (1:12:22.700)
Like how do you do that?
Lex Fridman (1:12:24.580)
How important do you think personalization
Lex Fridman (1:12:26.820)
is in human robot interaction?
Lex Fridman (1:12:28.580)
And not just a mechanic or student,
Lex Fridman (1:12:31.980)
but like literally to the individual human being.
Lex Fridman (1:12:35.340)
I think personalization is really important,
Lex Fridman (1:12:37.540)
but a caveat is that I think we'd be okay
Lex Fridman (1:12:42.140)
if we can personalize to the group, right?
Lex Fridman (1:12:44.700)
And so if I can label you
Lex Fridman (1:12:49.700)
as along some certain dimensions,
Ayanna Howard (1:12:52.780)
then even though it may not be you specifically,
Lex Fridman (1:12:56.500)
I can put you in this group.
Lex Fridman (1:12:58.220)
So the sample size, this is how they best learn,
Lex Fridman (1:13:00.500)
this is how they best engage.
Ayanna Howard (1:13:03.220)
Even at that level, it's really important.
Lex Fridman (1:13:06.780)
And it's because, I mean, it's one of the reasons
Lex Fridman (1:13:09.620)
why educating in large classrooms is so hard, right?
Lex Fridman (1:13:13.340)
You teach to the median,
Lex Fridman (1:13:15.980)
but there's these individuals that are struggling
Lex Fridman (1:13:19.780)
and then you have highly intelligent individuals
Lex Fridman (1:13:22.340)
and those are the ones that are usually kind of left out.
Lex Fridman (1:13:26.340)
So highly intelligent individuals may be disruptive
Lex Fridman (1:13:28.900)
and those who are struggling might be disruptive
Lex Fridman (1:13:30.860)
because they're both bored.
Ayanna Howard (1:13:32.980)
Yeah, and if you narrow the definition of the group
Lex Fridman (1:13:35.580)
or in the size of the group enough,
Ayanna Howard (1:13:37.900)
you'll be able to address their individual,
Lex Fridman (1:13:40.380)
it's not individual needs, but really the most important
Lex Fridman (1:13:44.580)
group needs, right?
Lex Fridman (1:13:45.980)
And that's kind of what a lot of successful
Ayanna Howard (1:13:47.780)
recommender systems do with Spotify and so on.
Lex Fridman (1:13:50.980)
So it's sad to believe, but as a music listener,
Ayanna Howard (1:13:53.820)
probably in some sort of large group,
Lex Fridman (1:13:55.940)
it's very sadly predictable.
Ayanna Howard (1:13:58.300)
You have been labeled.
Lex Fridman (1:13:59.260)
Yeah, I've been labeled and successfully so
Ayanna Howard (1:14:02.100)
because they're able to recommend stuff that I like.
Lex Fridman (1:14:04.820)
Yeah, but applying that to education, right?
Ayanna Howard (1:14:07.740)
There's no reason why it can't be done.
Lex Fridman (1:14:09.780)
Do you have a hope for our education system?
Ayanna Howard (1:14:13.060)
I have more hope for workforce development.
Lex Fridman (1:14:16.180)
And that's because I'm seeing investments.
Ayanna Howard (1:14:19.660)
Even if you look at VC investments in education,
Lex Fridman (1:14:23.300)
the majority of it has lately been going
Lex Fridman (1:14:26.140)
to workforce retraining, right?
Lex Fridman (1:14:28.540)
And so I think that government investments is increasing.
Lex Fridman (1:14:32.860)
There's like a claim and some of it's based on fear, right?
Lex Fridman (1:14:36.060)
Like AI is gonna come and take over all these jobs.
Lex Fridman (1:14:37.980)
What are we gonna do with all these nonpaying taxes
Lex Fridman (1:14:41.500)
that aren't coming to us by our citizens?
Lex Fridman (1:14:44.340)
And so I think I'm more hopeful for that.
Lex Fridman (1:14:48.060)
Not so hopeful for early education
Ayanna Howard (1:14:51.780)
because it's still a who's gonna pay for it.
Lex Fridman (1:14:56.380)
And you won't see the results for like 16 to 18 years.
Ayanna Howard (1:15:01.380)
It's hard for people to wrap their heads around that.
Lex Fridman (1:15:07.180)
But on the retraining part, what are your thoughts?
Ayanna Howard (1:15:10.580)
There's a candidate, Andrew Yang running for president
Lex Fridman (1:15:13.860)
and saying that sort of AI, automation, robots.
Ayanna Howard (1:15:18.940)
Universal basic income.
Lex Fridman (1:15:20.940)
Universal basic income in order to support us
Ayanna Howard (1:15:23.900)
as we kind of automation takes people's jobs
Lex Fridman (1:15:26.740)
and allows you to explore and find other means.
Ayanna Howard (1:15:30.180)
Like do you have a concern of society
Lex Fridman (1:15:35.660)
transforming effects of automation and robots and so on?
Ayanna Howard (1:15:40.500)
I do.
Lex Fridman (1:15:41.340)
I do know that AI robotics will displace workers.
Ayanna Howard (1:15:46.180)
Like we do know that.
Lex Fridman (1:15:47.980)
But there'll be other workers
Ayanna Howard (1:15:49.500)
that will be defined new jobs.
Lex Fridman (1:15:54.980)
What I worry about is, that's not what I worry about.
Lex Fridman (1:15:57.460)
Like will all the jobs go away?
Lex Fridman (1:15:59.500)
What I worry about is the type of jobs that will come out.
Ayanna Howard (1:16:02.460)
Like people who graduate from Georgia Tech will be okay.
Lex Fridman (1:16:06.340)
We give them the skills,
Ayanna Howard (1:16:07.660)
they will adapt even if their current job goes away.
Lex Fridman (1:16:10.660)
I do worry about those
Ayanna Howard (1:16:12.620)
that don't have that quality of an education.
Lex Fridman (1:16:15.460)
Will they have the ability,
Lex Fridman (1:16:18.300)
the background to adapt to those new jobs?
Lex Fridman (1:16:21.700)
That I don't know.
Ayanna Howard (1:16:22.980)
That I worry about,
Lex Fridman (1:16:24.220)
which will create even more polarization
Ayanna Howard (1:16:27.220)
in our society, internationally and everywhere.
Lex Fridman (1:16:31.220)
I worry about that.
Ayanna Howard (1:16:32.940)
I also worry about not having equal access
Lex Fridman (1:16:36.820)
to all these wonderful things that AI can do
Lex Fridman (1:16:39.540)
and robotics can do.
Lex Fridman (1:16:41.100)
I worry about that.
Ayanna Howard (1:16:43.620)
People like me from Georgia Tech from say MIT
Lex Fridman (1:16:48.860)
will be okay, right?
Lex Fridman (1:16:50.340)
But that's such a small part of the population
Lex Fridman (1:16:53.340)
that we need to think much more globally
Ayanna Howard (1:16:55.940)
of having access to the beautiful things,
Lex Fridman (1:16:58.500)
whether it's AI in healthcare, AI in education,
Lex Fridman (1:17:01.580)
AI in politics, right?
Lex Fridman (1:17:05.140)
I worry about that.
Lex Fridman (1:17:05.980)
And that's part of the thing that you were talking about
Lex Fridman (1:17:08.140)
is people that build the technology
Ayanna Howard (1:17:09.660)
have to be thinking about ethics,
Lex Fridman (1:17:12.420)
have to be thinking about access and all those things.
Lex Fridman (1:17:15.220)
And not just a small subset.
Lex Fridman (1:17:17.900)
Let me ask some philosophical,
Ayanna Howard (1:17:20.300)
slightly romantic questions.
Lex Fridman (1:17:22.460)
People that listen to this will be like,
Ayanna Howard (1:17:24.900)
here he goes again.
Lex Fridman (1:17:26.180)
Okay, do you think one day we'll build an AI system
Ayanna Howard (1:17:31.940)
that a person can fall in love with
Lex Fridman (1:17:35.500)
and it would love them back?
Ayanna Howard (1:17:37.900)
Like in the movie, Her, for example.
Lex Fridman (1:17:39.780)
Yeah, although she kind of didn't fall in love with him
Ayanna Howard (1:17:43.260)
or she fell in love with like a million other people,
Lex Fridman (1:17:45.500)
something like that.
Ayanna Howard (1:17:47.060)
You're the jealous type, I see.
Lex Fridman (1:17:48.460)
We humans are the jealous type.
Ayanna Howard (1:17:50.820)
Yes, so I do believe that we can design systems
Lex Fridman (1:17:55.060)
where people would fall in love with their robot,
Ayanna Howard (1:17:59.420)
with their AI partner.
Lex Fridman (1:18:03.220)
That I do believe.
Ayanna Howard (1:18:05.100)
Because it's actually,
Lex Fridman (1:18:06.300)
and I don't like to use the word manipulate,
Lex Fridman (1:18:08.900)
but as we see, there are certain individuals
Lex Fridman (1:18:12.300)
that can be manipulated
Lex Fridman (1:18:13.340)
if you understand the cognitive science about it, right?
Lex Fridman (1:18:16.260)
Right, so I mean, if you could think of all close
Ayanna Howard (1:18:19.620)
relationship and love in general
Lex Fridman (1:18:21.380)
as a kind of mutual manipulation,
Ayanna Howard (1:18:24.700)
that dance, the human dance.
Lex Fridman (1:18:27.100)
I mean, manipulation is a negative connotation.
Lex Fridman (1:18:30.180)
And that's why I don't like to use that word particularly.
Lex Fridman (1:18:32.820)
I guess another way to phrase it is,
Ayanna Howard (1:18:34.220)
you're getting at is it could be algorithmatized
Lex Fridman (1:18:36.900)
or something, it could be a.
Ayanna Howard (1:18:38.380)
The relationship building part can be.
Lex Fridman (1:18:40.620)
I mean, just think about it.
Ayanna Howard (1:18:41.820)
We have, and I don't use dating sites,
Lex Fridman (1:18:44.780)
but from what I heard, there are some individuals
Lex Fridman (1:18:48.940)
that have been dating that have never saw each other, right?
Lex Fridman (1:18:52.780)
In fact, there's a show I think
Ayanna Howard (1:18:54.100)
that tries to like weed out fake people.
Lex Fridman (1:18:57.540)
Like there's a show that comes out, right?
Ayanna Howard (1:18:59.460)
Because like people start faking.
Lex Fridman (1:19:01.940)
Like, what's the difference of that person
Lex Fridman (1:19:05.140)
on the other end being an AI agent, right?
Lex Fridman (1:19:08.020)
And having a communication
Lex Fridman (1:19:09.340)
and you building a relationship remotely,
Lex Fridman (1:19:12.180)
like there's no reason why that can't happen.
Ayanna Howard (1:19:15.660)
In terms of human robot interaction,
Lex Fridman (1:19:17.620)
so what role, you've kind of mentioned
Ayanna Howard (1:19:19.660)
with data emotion being, can be problematic
Lex Fridman (1:19:23.940)
if not implemented well, I suppose.
Lex Fridman (1:19:26.220)
What role does emotion and some other human like things,
Lex Fridman (1:19:30.500)
the imperfect things come into play here
Lex Fridman (1:19:32.820)
for good human robot interaction and something like love?
Lex Fridman (1:19:37.300)
Yeah, so in this case, and you had asked,
Lex Fridman (1:19:39.780)
can an AI agent love a human back?
Lex Fridman (1:19:43.700)
I think they can emulate love back, right?
Lex Fridman (1:19:47.340)
And so what does that actually mean?
Lex Fridman (1:19:48.980)
It just means that if you think about their programming,
Ayanna Howard (1:19:52.260)
they might put the other person's needs
Lex Fridman (1:19:55.220)
in front of theirs in certain situations, right?
Ayanna Howard (1:19:57.980)
You look at, think about it as a return on investment.
Lex Fridman (1:20:00.380)
Like, what's my return on investment?
Ayanna Howard (1:20:01.740)
As part of that equation, that person's happiness,
Lex Fridman (1:20:04.540)
has some type of algorithm waiting to it.
Lex Fridman (1:20:07.940)
And the reason why is because I care about them, right?
Lex Fridman (1:20:11.380)
That's the only reason, right?
Lex Fridman (1:20:13.700)
But if I care about them and I show that,
Lex Fridman (1:20:15.540)
then my final objective function
Lex Fridman (1:20:18.300)
is length of time of the engagement, right?
Lex Fridman (1:20:20.580)
So you can think of how to do this actually quite easily.
Lex Fridman (1:20:24.020)
And so.
Lex Fridman (1:20:24.860)
But that's not love?
Ayanna Howard (1:20:27.420)
Well, so that's the thing.
Lex Fridman (1:20:29.940)
I think it emulates love
Ayanna Howard (1:20:32.580)
because we don't have a classical definition of love.
Lex Fridman (1:20:38.540)
Right, but, and we don't have the ability
Ayanna Howard (1:20:41.660)
to look into each other's minds to see the algorithm.
Lex Fridman (1:20:45.500)
And I mean, I guess what I'm getting at is,
Ayanna Howard (1:20:48.740)
is it possible that, especially if that's learned,
Lex Fridman (1:20:51.020)
especially if there's some mystery
Lex Fridman (1:20:52.580)
and black box nature to the system,
Lex Fridman (1:20:55.220)
how is that, you know?
Lex Fridman (1:20:57.660)
How is it any different?
Lex Fridman (1:20:58.580)
How is it any different in terms of sort of
Ayanna Howard (1:21:00.660)
if the system says, I'm conscious, I'm afraid of death,
Lex Fridman (1:21:05.180)
and it does indicate that it loves you.
Ayanna Howard (1:21:10.860)
Another way to sort of phrase it,
Lex Fridman (1:21:12.780)
be curious to see what you think.
Lex Fridman (1:21:14.180)
Do you think there'll be a time
Lex Fridman (1:21:16.700)
when robots should have rights?
Ayanna Howard (1:21:20.140)
You've kind of phrased the robot in a very roboticist way
Lex Fridman (1:21:23.420)
and just a really good way, but saying, okay,
Ayanna Howard (1:21:25.940)
well, there's an objective function
Lex Fridman (1:21:27.940)
and I could see how you can create
Ayanna Howard (1:21:30.620)
a compelling human robot interaction experience
Lex Fridman (1:21:33.380)
that makes you believe that the robot cares for your needs
Lex Fridman (1:21:36.300)
and even something like loves you.
Lex Fridman (1:21:38.940)
But what if the robot says, please don't turn me off?
Lex Fridman (1:21:43.740)
What if the robot starts making you feel
Lex Fridman (1:21:46.460)
like there's an entity, a being, a soul there, right?
Lex Fridman (1:21:50.060)
Do you think there'll be a future,
Lex Fridman (1:21:53.420)
hopefully you won't laugh too much at this,
Lex Fridman (1:21:55.700)
but where they do ask for rights?
Lex Fridman (1:22:00.020)
So I can see a future
Ayanna Howard (1:22:03.980)
if we don't address it in the near term
Lex Fridman (1:22:08.500)
where these agents, as they adapt and learn,
Ayanna Howard (1:22:11.820)
could say, hey, this should be something that's fundamental.
Lex Fridman (1:22:15.820)
I hopefully think that we would address it
Ayanna Howard (1:22:18.860)
before it gets to that point.
Lex Fridman (1:22:20.580)
So you think that's a bad future?
Ayanna Howard (1:22:22.140)
Is that a negative thing where they ask
Lex Fridman (1:22:25.340)
we're being discriminated against?
Ayanna Howard (1:22:27.740)
I guess it depends on what role
Lex Fridman (1:22:31.100)
have they attained at that point, right?
Lex Fridman (1:22:34.340)
And so if I think about now.
Lex Fridman (1:22:35.820)
Careful what you say because the robots 50 years from now
Ayanna Howard (1:22:39.220)
I'll be listening to this and you'll be on TV saying,
Lex Fridman (1:22:42.180)
this is what roboticists used to believe.
Lex Fridman (1:22:44.420)
Well, right?
Lex Fridman (1:22:45.260)
And so this is my, and as I said, I have a bias lens
Lex Fridman (1:22:48.700)
and my robot friends will understand that.
Lex Fridman (1:22:52.780)
So if you think about it, and I actually put this
Ayanna Howard (1:22:55.180)
in kind of the, as a roboticist,
Lex Fridman (1:22:59.660)
you don't necessarily think of robots as human
Ayanna Howard (1:23:02.500)
with human rights, but you could think of them
Lex Fridman (1:23:05.020)
either in the category of property,
Lex Fridman (1:23:09.180)
or you can think of them in the category of animals, right?
Lex Fridman (1:23:14.340)
And so both of those have different types of rights.
Lex Fridman (1:23:18.340)
So animals have their own rights as a living being,
Lex Fridman (1:23:22.740)
but they can't vote, they can't write,
Ayanna Howard (1:23:25.060)
they can be euthanized, but as humans,
Lex Fridman (1:23:29.700)
if we abuse them, we go to jail, right?
Lex Fridman (1:23:32.980)
So they do have some rights that protect them,
Lex Fridman (1:23:35.980)
but don't give them the rights of like citizenship.
Lex Fridman (1:23:40.140)
And then if you think about property,
Lex Fridman (1:23:42.260)
property, the rights are associated with the person, right?
Lex Fridman (1:23:45.700)
So if someone vandalizes your property
Lex Fridman (1:23:49.500)
or steals your property, like there are some rights,
Lex Fridman (1:23:53.820)
but it's associated with the person who owns that.
Lex Fridman (1:23:58.660)
If you think about it back in the day,
Lex Fridman (1:24:01.500)
and if you remember, we talked about
Lex Fridman (1:24:03.380)
how society has changed, women were property, right?
Ayanna Howard (1:24:08.180)
They were not thought of as having rights.
Lex Fridman (1:24:11.860)
They were thought of as property of, like their...
Ayanna Howard (1:24:15.740)
Yeah, assaulting a woman meant
Lex Fridman (1:24:17.620)
assaulting the property of somebody else.
Ayanna Howard (1:24:20.060)
Exactly, and so what I envision is,
Lex Fridman (1:24:22.900)
is that we will establish some type of norm at some point,
Lex Fridman (1:24:27.820)
but that it might evolve, right?
Lex Fridman (1:24:29.580)
Like if you look at women's rights now,
Ayanna Howard (1:24:31.460)
like there are still some countries that don't have,
Lex Fridman (1:24:35.380)
and the rest of the world is like,
Lex Fridman (1:24:36.700)
why that makes no sense, right?
Lex Fridman (1:24:39.260)
And so I do see a world where we do establish
Ayanna Howard (1:24:42.100)
some type of grounding.
Lex Fridman (1:24:44.140)
It might be based on property rights,
Ayanna Howard (1:24:45.700)
it might be based on animal rights.
Lex Fridman (1:24:47.620)
And if it evolves that way,
Ayanna Howard (1:24:50.700)
I think we will have this conversation at that time,
Lex Fridman (1:24:54.460)
because that's the way our society traditionally has evolved.
Ayanna Howard (1:24:58.500)
Beautifully put, just out of curiosity,
Lex Fridman (1:25:01.860)
Anki, Jibo, Mayflower Robotics,
Ayanna Howard (1:25:05.460)
with their robot Curie, SciFiWorks, WeThink Robotics,
Lex Fridman (1:25:08.380)
were all these amazing robotics companies
Ayanna Howard (1:25:10.580)
led, created by incredible roboticists,
Lex Fridman (1:25:14.300)
and they've all went out of business recently.
Lex Fridman (1:25:19.580)
Why do you think they didn't last long?
Lex Fridman (1:25:21.660)
Why is it so hard to run a robotics company,
Ayanna Howard (1:25:25.140)
especially one like these, which are fundamentally
Lex Fridman (1:25:29.300)
HRI human robot interaction robots?
Lex Fridman (1:25:34.380)
Or personal robots?
Lex Fridman (1:25:35.700)
Each one has a story,
Ayanna Howard (1:25:37.100)
only one of them I don't understand, and that was Anki.
Lex Fridman (1:25:41.180)
That's actually the only one I don't understand.
Ayanna Howard (1:25:43.340)
I don't understand it either.
Lex Fridman (1:25:44.660)
No, no, I mean, I look like from the outside,
Ayanna Howard (1:25:47.020)
I've looked at their sheets, I've looked at the data that's.
Lex Fridman (1:25:50.500)
Oh, you mean like business wise,
Ayanna Howard (1:25:51.740)
you don't understand, I got you.
Lex Fridman (1:25:52.900)
Yeah.
Ayanna Howard (1:25:53.740)
Yeah, and like I look at all, I look at that data,
Lex Fridman (1:25:59.180)
and I'm like, they seem to have like product market fit.
Ayanna Howard (1:26:02.660)
Like, so that's the only one I don't understand.
Lex Fridman (1:26:05.660)
The rest of it was product market fit.
Lex Fridman (1:26:08.260)
What's product market fit?
Lex Fridman (1:26:09.860)
Just that of, like how do you think about it?
Lex Fridman (1:26:11.940)
Yeah, so although WeThink Robotics was getting there, right?
Lex Fridman (1:26:15.620)
But I think it's just the timing,
Ayanna Howard (1:26:17.420)
it just, their clock just timed out.
Lex Fridman (1:26:20.340)
I think if they'd been given a couple more years,
Ayanna Howard (1:26:23.100)
they would have been okay.
Lex Fridman (1:26:25.060)
But the other ones were still fairly early
Ayanna Howard (1:26:28.620)
by the time they got into the market.
Lex Fridman (1:26:30.100)
And so product market fit is,
Ayanna Howard (1:26:32.740)
I have a product that I wanna sell at a certain price.
Lex Fridman (1:26:37.140)
Are there enough people out there, the market,
Ayanna Howard (1:26:40.060)
that are willing to buy the product at that market price
Lex Fridman (1:26:42.780)
for me to be a functional viable profit bearing company?
Lex Fridman (1:26:47.820)
Right?
Lex Fridman (1:26:48.940)
So product market fit.
Ayanna Howard (1:26:50.420)
If it costs you a thousand dollars
Lex Fridman (1:26:53.300)
and everyone wants it and only is willing to pay a dollar,
Ayanna Howard (1:26:57.340)
you have no product market fit.
Lex Fridman (1:26:59.260)
Even if you could sell it for, you know,
Ayanna Howard (1:27:01.940)
it's enough for a dollar, cause you can't.
Lex Fridman (1:27:03.660)
So how hard is it for robots?
Ayanna Howard (1:27:05.380)
Sort of maybe if you look at iRobot,
Lex Fridman (1:27:07.580)
the company that makes Roombas, vacuum cleaners,
Lex Fridman (1:27:10.740)
can you comment on, did they find the right product,
Lex Fridman (1:27:14.100)
market product fit?
Ayanna Howard (1:27:15.940)
Like, are people willing to pay for robots
Lex Fridman (1:27:18.540)
is also another kind of question underlying all this.
Lex Fridman (1:27:20.540)
So if you think about iRobot and their story, right?
Lex Fridman (1:27:23.700)
Like when they first, they had enough of a runway, right?
Ayanna Howard (1:27:28.660)
When they first started,
Lex Fridman (1:27:29.780)
they weren't doing vacuum cleaners, right?
Ayanna Howard (1:27:31.340)
They were contracts primarily, government contracts,
Lex Fridman (1:27:36.540)
designing robots.
Ayanna Howard (1:27:37.380)
Or military robots.
Lex Fridman (1:27:38.220)
Yeah, I mean, that's what they were.
Lex Fridman (1:27:39.380)
That's how they started, right?
Lex Fridman (1:27:40.820)
And then.
Ayanna Howard (1:27:41.660)
They still do a lot of incredible work there.
Lex Fridman (1:27:42.740)
But yeah, that was the initial thing
Ayanna Howard (1:27:44.660)
that gave them enough funding to.
Lex Fridman (1:27:46.620)
To then try to, the vacuum cleaner is what I've been told
Ayanna Howard (1:27:50.740)
was not like their first rendezvous
Lex Fridman (1:27:53.900)
in terms of designing a product, right?
Lex Fridman (1:27:56.500)
And so they were able to survive
Lex Fridman (1:27:59.300)
until they got to the point
Lex Fridman (1:28:00.620)
that they found a product price market, right?
Lex Fridman (1:28:05.540)
And even with, if you look at the Roomba,
Ayanna Howard (1:28:09.100)
the price point now is different
Lex Fridman (1:28:10.540)
than when it was first released, right?
Ayanna Howard (1:28:12.260)
It was an early adopter price,
Lex Fridman (1:28:13.460)
but they found enough people
Ayanna Howard (1:28:14.540)
who were willing to fund it.
Lex Fridman (1:28:16.700)
And I mean, I forgot what their loss profile was
Ayanna Howard (1:28:20.340)
for the first couple of years,
Lex Fridman (1:28:22.180)
but they became profitable in sufficient time
Ayanna Howard (1:28:25.860)
that they didn't have to close their doors.
Lex Fridman (1:28:28.140)
So they found the right,
Ayanna Howard (1:28:29.140)
there's still people willing to pay
Lex Fridman (1:28:31.860)
a large amount of money,
Lex Fridman (1:28:32.700)
so over $1,000 for a vacuum cleaner.
Lex Fridman (1:28:35.940)
Unfortunately for them,
Ayanna Howard (1:28:37.780)
now that they've proved everything out,
Lex Fridman (1:28:39.180)
figured it all out,
Ayanna Howard (1:28:40.020)
now there's competitors.
Lex Fridman (1:28:40.860)
Yeah, and so that's the next thing, right?
Ayanna Howard (1:28:43.500)
The competition,
Lex Fridman (1:28:44.660)
and they have quite a number, even internationally.
Ayanna Howard (1:28:47.500)
Like there's some products out there,
Lex Fridman (1:28:50.180)
you can go to Europe and be like,
Ayanna Howard (1:28:52.420)
oh, I didn't even know this one existed.
Lex Fridman (1:28:55.020)
So this is the thing though,
Ayanna Howard (1:28:56.780)
like with any market,
Lex Fridman (1:28:59.300)
I would, this is not a bad time,
Ayanna Howard (1:29:03.580)
although as a roboticist, it's kind of depressing,
Lex Fridman (1:29:06.500)
but I actually think about things like with,
Ayanna Howard (1:29:11.340)
I would say that all of the companies
Lex Fridman (1:29:13.060)
that are now in the top five or six,
Lex Fridman (1:29:15.780)
they weren't the first to the stage, right?
Lex Fridman (1:29:19.620)
Like Google was not the first search engine,
Lex Fridman (1:29:22.780)
sorry, Altavista, right?
Lex Fridman (1:29:24.780)
Facebook was not the first, sorry, MySpace, right?
Ayanna Howard (1:29:28.340)
Like think about it,
Lex Fridman (1:29:29.180)
they were not the first players.
Ayanna Howard (1:29:31.100)
Those first players,
Lex Fridman (1:29:32.980)
like they're not in the top five, 10 of Fortune 500 companies,
Lex Fridman (1:29:38.580)
right?
Lex Fridman (1:29:39.420)
They proved, they started to prove out the market,
Ayanna Howard (1:29:43.940)
they started to get people interested,
Lex Fridman (1:29:46.340)
they started the buzz,
Lex Fridman (1:29:48.300)
but they didn't make it to that next level.
Lex Fridman (1:29:50.060)
But the second batch, right?
Ayanna Howard (1:29:52.300)
The second batch, I think might make it to the next level.
Lex Fridman (1:29:57.540)
When do you think the Facebook of robotics?
Ayanna Howard (1:30:02.380)
The Facebook of robotics.
Lex Fridman (1:30:04.740)
Sorry, I take that phrase back because people deeply,
Ayanna Howard (1:30:08.500)
for some reason, well, I know why,
Lex Fridman (1:30:10.340)
but it's, I think, exaggerated distrust Facebook
Ayanna Howard (1:30:13.700)
because of the privacy concerns and so on.
Lex Fridman (1:30:15.500)
And with robotics, one of the things you have to make sure
Ayanna Howard (1:30:18.420)
is all the things we talked about is to be transparent
Lex Fridman (1:30:21.340)
and have people deeply trust you
Ayanna Howard (1:30:22.980)
to let a robot into their lives, into their home.
Lex Fridman (1:30:25.780)
When do you think the second batch of robots will come?
Ayanna Howard (1:30:28.620)
Is it five, 10 years, 20 years
Lex Fridman (1:30:32.140)
that we'll have robots in our homes
Lex Fridman (1:30:34.700)
and robots in our hearts?
Lex Fridman (1:30:36.540)
So if I think about, and because I try to follow
Ayanna Howard (1:30:38.900)
the VC kind of space in terms of robotic investments,
Lex Fridman (1:30:43.180)
and right now, and I don't know
Ayanna Howard (1:30:44.900)
if they're gonna be successful,
Lex Fridman (1:30:45.900)
I don't know if this is the second batch,
Lex Fridman (1:30:49.220)
but there's only one batch that's focused
Lex Fridman (1:30:50.980)
on like the first batch, right?
Lex Fridman (1:30:52.900)
And then there's all these self driving Xs, right?
Lex Fridman (1:30:56.260)
And so I don't know if they're a first batch of something
Ayanna Howard (1:30:59.540)
or if like, I don't know quite where they fit in,
Lex Fridman (1:31:03.060)
but there's a number of companies,
Ayanna Howard (1:31:05.540)
the co robot, I call them co robots
Lex Fridman (1:31:08.500)
that are still getting VC investments.
Ayanna Howard (1:31:13.060)
Some of them have some of the flavor
Lex Fridman (1:31:14.500)
of like Rethink Robotics.
Ayanna Howard (1:31:15.740)
Some of them have some of the flavor of like Curie.
Lex Fridman (1:31:18.980)
What's a co robot?
Lex Fridman (1:31:20.740)
So basically a robot and human working in the same space.
Lex Fridman (1:31:26.060)
So some of the companies are focused on manufacturing.
Lex Fridman (1:31:30.500)
So having a robot and human working together
Lex Fridman (1:31:34.220)
in a factory, some of these co robots
Ayanna Howard (1:31:37.580)
are robots and humans working in the home,
Lex Fridman (1:31:41.220)
working in clinics, like there's different versions
Ayanna Howard (1:31:43.180)
of these companies in terms of their products,
Lex Fridman (1:31:45.380)
but they're all, so we think robotics would be
Ayanna Howard (1:31:48.660)
like one of the first, at least well known companies
Lex Fridman (1:31:52.660)
focused on this space.
Lex Fridman (1:31:54.580)
So I don't know if this is a second batch
Lex Fridman (1:31:56.700)
or if this is still part of the first batch,
Ayanna Howard (1:32:00.940)
that I don't know.
Lex Fridman (1:32:01.980)
And then you have all these other companies
Ayanna Howard (1:32:03.740)
in this self driving space.
Lex Fridman (1:32:06.860)
And I don't know if that's a first batch
Ayanna Howard (1:32:09.380)
or again, a second batch.
Lex Fridman (1:32:11.140)
Yeah.
Lex Fridman (1:32:11.980)
So there's a lot of mystery about this now.
Lex Fridman (1:32:13.860)
Of course, it's hard to say that this is the second batch
Lex Fridman (1:32:16.380)
until it proves out, right?
Lex Fridman (1:32:18.460)
Correct.
Ayanna Howard (1:32:19.300)
Yeah, we need a unicorn.
Lex Fridman (1:32:20.540)
Yeah, exactly.
Lex Fridman (1:32:23.460)
Why do you think people are so afraid,
Lex Fridman (1:32:27.700)
at least in popular culture of legged robots
Ayanna Howard (1:32:30.460)
like those worked in Boston Dynamics
Lex Fridman (1:32:32.340)
or just robotics in general,
Ayanna Howard (1:32:34.140)
if you were to psychoanalyze that fear,
Lex Fridman (1:32:36.300)
what do you make of it?
Lex Fridman (1:32:37.980)
And should they be afraid, sorry?
Lex Fridman (1:32:39.780)
So should people be afraid?
Ayanna Howard (1:32:41.420)
I don't think people should be afraid.
Lex Fridman (1:32:43.860)
But with a caveat, I don't think people should be afraid
Ayanna Howard (1:32:47.060)
given that most of us in this world
Lex Fridman (1:32:51.460)
understand that we need to change something, right?
Lex Fridman (1:32:55.660)
So given that.
Lex Fridman (1:32:58.100)
Now, if things don't change, be very afraid.
Lex Fridman (1:33:01.500)
Which is the dimension of change that's needed?
Lex Fridman (1:33:04.380)
So changing, thinking about the ramifications,
Ayanna Howard (1:33:07.740)
thinking about like the ethics,
Lex Fridman (1:33:09.420)
thinking about like the conversation is going on, right?
Ayanna Howard (1:33:12.740)
It's no longer a we're gonna deploy it
Lex Fridman (1:33:15.860)
and forget that this is a car that can kill pedestrians
Lex Fridman (1:33:20.300)
that are walking across the street, right?
Lex Fridman (1:33:22.500)
We're not in that stage.
Ayanna Howard (1:33:23.340)
We're putting these roads out.
Lex Fridman (1:33:25.820)
There are people out there.
Ayanna Howard (1:33:27.500)
A car could be a weapon.
Lex Fridman (1:33:28.700)
Like people are now, solutions aren't there yet,
Lex Fridman (1:33:33.140)
but people are thinking about this
Lex Fridman (1:33:35.300)
as we need to be ethically responsible
Ayanna Howard (1:33:38.460)
as we send these systems out,
Lex Fridman (1:33:40.820)
robotics, medical, self driving.
Lex Fridman (1:33:43.060)
And military too.
Lex Fridman (1:33:43.940)
And military.
Ayanna Howard (1:33:45.260)
Which is not as often talked about,
Lex Fridman (1:33:46.980)
but it's really where probably these robots
Ayanna Howard (1:33:50.260)
will have a significant impact as well.
Lex Fridman (1:33:51.900)
Correct, correct.
Ayanna Howard (1:33:52.820)
Right, making sure that they can think rationally,
Lex Fridman (1:33:57.340)
even having the conversations,
Lex Fridman (1:33:58.700)
who should pull the trigger, right?
Lex Fridman (1:34:01.260)
But overall you're saying if we start to think
Ayanna Howard (1:34:03.380)
more and more as a community about these ethical issues,
Lex Fridman (1:34:05.740)
people should not be afraid.
Ayanna Howard (1:34:06.980)
Yeah, I don't think people should be afraid.
Lex Fridman (1:34:08.660)
I think that the return on investment,
Ayanna Howard (1:34:10.540)
the impact, positive impact will outweigh
Lex Fridman (1:34:14.060)
any of the potentially negative impacts.
Lex Fridman (1:34:17.500)
Do you have worries of existential threats
Lex Fridman (1:34:20.540)
of robots or AI that some people kind of talk about
Lex Fridman (1:34:25.540)
and romanticize about in the next decade,
Lex Fridman (1:34:28.620)
the next few decades?
Ayanna Howard (1:34:29.980)
No, I don't.
Lex Fridman (1:34:31.340)
Singularity would be an example.
Lex Fridman (1:34:33.700)
So my concept is that, so remember,
Lex Fridman (1:34:36.380)
robots, AI, is designed by people.
Ayanna Howard (1:34:39.580)
It has our values.
Lex Fridman (1:34:41.260)
And I always correlate this with a parent and a child.
Lex Fridman (1:34:45.060)
So think about it, as a parent, what do we want?
Lex Fridman (1:34:47.100)
We want our kids to have a better life than us.
Ayanna Howard (1:34:49.860)
We want them to expand.
Lex Fridman (1:34:52.300)
We want them to experience the world.
Lex Fridman (1:34:55.780)
And then as we grow older, our kids think and know
Lex Fridman (1:34:59.740)
they're smarter and better and more intelligent
Lex Fridman (1:35:03.020)
and have better opportunities.
Lex Fridman (1:35:04.780)
And they may even stop listening to us.
Lex Fridman (1:35:08.220)
They don't go out and then kill us, right?
Lex Fridman (1:35:10.500)
Like, think about it.
Ayanna Howard (1:35:11.340)
It's because we, it's instilled in them values.
Lex Fridman (1:35:14.180)
We instilled in them this whole aspect of community.
Lex Fridman (1:35:17.420)
And yes, even though you're maybe smarter
Lex Fridman (1:35:19.780)
and have more money and dah, dah, dah,
Ayanna Howard (1:35:22.460)
it's still about this love, caring relationship.
Lex Fridman (1:35:26.780)
And so that's what I believe.
Lex Fridman (1:35:27.740)
So even if like, you know,
Lex Fridman (1:35:29.020)
we've created the singularity in some archaic system
Ayanna Howard (1:35:32.140)
back in like 1980 that suddenly evolves,
Lex Fridman (1:35:35.340)
the fact is it might say, I am smarter, I am sentient.
Ayanna Howard (1:35:40.180)
These humans are really stupid,
Lex Fridman (1:35:43.220)
but I think it'll be like, yeah,
Lex Fridman (1:35:46.060)
but I just can't destroy them.
Lex Fridman (1:35:47.620)
Yeah, for sentimental value.
Ayanna Howard (1:35:49.660)
It's still just to come back for Thanksgiving dinner
Lex Fridman (1:35:53.140)
every once in a while.
Ayanna Howard (1:35:53.980)
Exactly.
Lex Fridman (1:35:54.820)
That's such, that's so beautifully put.
Ayanna Howard (1:35:57.460)
You've also said that The Matrix may be
Lex Fridman (1:36:00.580)
one of your more favorite AI related movies.
Lex Fridman (1:36:03.660)
Can you elaborate why?
Lex Fridman (1:36:05.580)
Yeah, it is one of my favorite movies.
Lex Fridman (1:36:07.860)
And it's because it represents
Lex Fridman (1:36:11.180)
kind of all the things I think about.
Lex Fridman (1:36:14.060)
So there's a symbiotic relationship
Lex Fridman (1:36:16.100)
between robots and humans, right?
Ayanna Howard (1:36:20.140)
That symbiotic relationship is that they don't destroy us,
Lex Fridman (1:36:22.500)
they enslave us, right?
Lex Fridman (1:36:24.620)
But think about it,
Lex Fridman (1:36:28.060)
even though they enslaved us,
Lex Fridman (1:36:30.260)
they needed us to be happy, right?
Lex Fridman (1:36:32.820)
And in order to be happy,
Ayanna Howard (1:36:33.860)
they had to create this cruddy world
Lex Fridman (1:36:35.420)
that they then had to live in, right?
Ayanna Howard (1:36:36.980)
That's the whole premise.
Lex Fridman (1:36:39.460)
But then there were humans that had a choice, right?
Ayanna Howard (1:36:44.380)
Like you had a choice to stay in this horrific,
Lex Fridman (1:36:47.660)
horrific world where it was your fantasy life
Ayanna Howard (1:36:51.220)
with all of the anomalies, perfection, but not accurate.
Lex Fridman (1:36:54.740)
Or you can choose to be on your own
Lex Fridman (1:36:57.940)
and like have maybe no food for a couple of days,
Lex Fridman (1:37:02.500)
but you were totally autonomous.
Lex Fridman (1:37:05.180)
And so I think of that as, and that's why.
Lex Fridman (1:37:07.980)
So it's not necessarily us being enslaved,
Lex Fridman (1:37:09.700)
but I think about us having the symbiotic relationship.
Lex Fridman (1:37:13.060)
Robots and AI, even if they become sentient,
Ayanna Howard (1:37:15.780)
they're still part of our society
Lex Fridman (1:37:17.100)
and they will suffer just as much as we.
Lex Fridman (1:37:20.700)
And there will be some kind of equilibrium
Lex Fridman (1:37:23.820)
that we'll have to find some symbiotic relationship.
Ayanna Howard (1:37:26.700)
Right, and then you have the ethicists,
Lex Fridman (1:37:28.220)
the robotics folks that are like,
Ayanna Howard (1:37:30.180)
no, this has got to stop, I will take the other pill
Lex Fridman (1:37:34.500)
in order to make a difference.
Lex Fridman (1:37:36.300)
So if you could hang out for a day with a robot,
Lex Fridman (1:37:40.380)
real or from science fiction, movies, books, safely,
Lex Fridman (1:37:44.740)
and get to pick his or her, their brain,
Lex Fridman (1:37:48.780)
who would you pick?
Ayanna Howard (1:37:55.980)
Gotta say it's Data.
Lex Fridman (1:37:57.620)
Data.
Ayanna Howard (1:37:58.740)
I was gonna say Rosie,
Lex Fridman (1:38:00.460)
but I'm not really interested in her brain.
Ayanna Howard (1:38:03.660)
I'm interested in Data's brain.
Lex Fridman (1:38:05.820)
Data pre or post emotion chip?
Ayanna Howard (1:38:08.460)
Pre.
Lex Fridman (1:38:10.460)
But don't you think it'd be a more interesting conversation
Lex Fridman (1:38:15.100)
post emotion chip?
Lex Fridman (1:38:16.180)
Yeah, it would be drama.
Lex Fridman (1:38:17.740)
And I'm human, I deal with drama all the time.
Lex Fridman (1:38:22.860)
But the reason why I wanna pick Data's brain
Ayanna Howard (1:38:24.860)
is because I could have a conversation with him
Lex Fridman (1:38:29.540)
and ask, for example, how can we fix this ethics problem?
Lex Fridman (1:38:34.540)
And he could go through like the rational thinking
Lex Fridman (1:38:38.300)
and through that, he could also help me
Ayanna Howard (1:38:40.780)
think through it as well.
Lex Fridman (1:38:42.220)
And so there's like these fundamental questions
Ayanna Howard (1:38:44.860)
I think I could ask him
Lex Fridman (1:38:46.420)
that he would help me also learn from.
Lex Fridman (1:38:49.980)
And that fascinates me.
Lex Fridman (1:38:52.860)
I don't think there's a better place to end it.
Ayanna Howard (1:38:55.140)
Ayana, thank you so much for talking to us, it was an honor.
Lex Fridman (1:38:57.300)
Thank you, thank you.
Ayanna Howard (1:38:58.140)
This was fun.
Lex Fridman (1:39:00.300)
Thanks for listening to this conversation
Lex Fridman (1:39:02.420)
and thank you to our presenting sponsor, Cash App.
Lex Fridman (1:39:05.900)
Download it, use code LexPodcast,
Ayanna Howard (1:39:08.540)
you'll get $10 and $10 will go to FIRST,
Lex Fridman (1:39:11.340)
a STEM education nonprofit that inspires
Ayanna Howard (1:39:13.620)
hundreds of thousands of young minds
Lex Fridman (1:39:15.820)
to become future leaders and innovators.
Ayanna Howard (1:39:18.740)
If you enjoy this podcast, subscribe on YouTube,
Lex Fridman (1:39:21.540)
give it five stars on Apple Podcast,
Ayanna Howard (1:39:23.540)
follow on Spotify, support on Patreon
Lex Fridman (1:39:26.220)
or simply connect with me on Twitter.
Lex Fridman (1:39:29.300)
And now let me leave you with some words of wisdom
Lex Fridman (1:39:31.860)
from Arthur C. Clarke.
Ayanna Howard (1:39:35.180)
Whether we are based on carbon or on silicon
Lex Fridman (1:39:38.580)
makes no fundamental difference.
Ayanna Howard (1:39:40.620)
We should each be treated with appropriate respect.
Lex Fridman (1:39:43.660)
Thank you for listening and hope to see you next time.
Lex Fridman (20:03.540)
Do you think from a robotics perspective,
Lex Fridman (20:07.180)
you know, if you're kind of honest of what cars do,
Ayanna Howard (20:09.820)
they kind of threaten each other's life all the time.
Lex Fridman (20:14.860)
So cars are various.
Ayanna Howard (20:17.340)
I mean, in order to navigate intersections,
Lex Fridman (20:19.300)
there's an assertiveness, there's a risk taking.
Lex Fridman (20:22.300)
And if you were to reduce it to an objective function,
Lex Fridman (20:25.300)
there's a probability of murder in that function,
Ayanna Howard (20:28.740)
meaning you killing another human being
Lex Fridman (20:31.900)
and you're using that.
Ayanna Howard (20:33.580)
First of all, it has to be low enough
Lex Fridman (20:36.940)
to be acceptable to you on an ethical level
Ayanna Howard (20:39.700)
as an individual human being,
Lex Fridman (20:41.300)
but it has to be high enough for people to respect you
Ayanna Howard (20:45.300)
to not sort of take advantage of you completely
Lex Fridman (20:47.540)
and jaywalk in front of you and so on.
Ayanna Howard (20:49.620)
So, I mean, I don't think there's a right answer here,
Lex Fridman (20:53.100)
but what's, how do we solve that?
Lex Fridman (20:56.100)
How do we solve that from a robotics perspective
Lex Fridman (20:57.940)
when danger and human life is at stake?
Ayanna Howard (21:00.140)
Yeah, as they say, cars don't kill people,
Lex Fridman (21:01.980)
people kill people.
Ayanna Howard (21:02.940)
People kill people.
Lex Fridman (21:05.100)
Right.
Lex Fridman (21:07.100)
So I think.
Lex Fridman (21:08.620)
And now robotic algorithms would be killing people.
Ayanna Howard (21:10.780)
Right, so it will be robotics algorithms that are pro,
Lex Fridman (21:14.380)
no, it will be robotic algorithms don't kill people.
Lex Fridman (21:16.980)
Developers of robotic algorithms kill people, right?
Lex Fridman (21:19.740)
I mean, one of the things is people are still in the loop
Lex Fridman (21:22.940)
and at least in the near and midterm,
Lex Fridman (21:26.540)
I think people will still be in the loop at some point,
Ayanna Howard (21:29.420)
even if it's a developer.
Lex Fridman (21:30.300)
Like we're not necessarily at the stage
Ayanna Howard (21:31.860)
where robots are programming autonomous robots
Lex Fridman (21:36.740)
with different behaviors quite yet.
Ayanna Howard (21:39.980)
It's a scary notion, sorry to interrupt,
Lex Fridman (21:42.260)
that a developer has some responsibility
Ayanna Howard (21:47.420)
in the death of a human being.
Lex Fridman (21:49.700)
That's a heavy burden.
Ayanna Howard (21:50.620)
I mean, I think that's why the whole aspect of ethics
Lex Fridman (21:55.460)
in our community is so, so important, right?
Ayanna Howard (21:58.500)
Like, because it's true.
Lex Fridman (22:00.060)
If you think about it, you can basically say,
Lex Fridman (22:04.820)
I'm not going to work on weaponized AI, right?
Lex Fridman (22:07.460)
Like people can say, that's not what I'm gonna do.
Lex Fridman (22:09.860)
But yet you are programming algorithms
Lex Fridman (22:12.740)
that might be used in healthcare algorithms
Ayanna Howard (22:15.620)
that might decide whether this person
Lex Fridman (22:17.260)
should get this medication or not.
Lex Fridman (22:18.980)
And they don't and they die.
Lex Fridman (22:21.420)
Okay, so that is your responsibility, right?
Lex Fridman (22:25.100)
And if you're not conscious and aware
Lex Fridman (22:27.340)
that you do have that power when you're coding
Lex Fridman (22:30.020)
and things like that, I think that's just not a good thing.
Lex Fridman (22:35.020)
Like we need to think about this responsibility
Ayanna Howard (22:38.020)
as we program robots and computing devices
Lex Fridman (22:41.820)
much more than we are.
Ayanna Howard (22:44.340)
Yeah, so it's not an option to not think about ethics.
Lex Fridman (22:46.980)
I think it's a majority, I would say, of computer science.
Ayanna Howard (22:51.340)
Sort of, it's kind of a hot topic now,
Lex Fridman (22:53.860)
I think about bias and so on, but it's,
Lex Fridman (22:56.620)
and we'll talk about it, but usually it's kind of,
Lex Fridman (23:00.380)
it's like a very particular group of people
Ayanna Howard (23:02.700)
that work on that.
Lex Fridman (23:04.260)
And then people who do like robotics are like,
Ayanna Howard (23:06.940)
well, I don't have to think about that.
Lex Fridman (23:09.380)
There's other smart people thinking about it.
Ayanna Howard (23:11.180)
It seems that everybody has to think about it.
Lex Fridman (23:14.580)
It's not, you can't escape the ethics,
Ayanna Howard (23:17.060)
whether it's bias or just every aspect of ethics
Lex Fridman (23:21.140)
that has to do with human beings.
Ayanna Howard (23:22.700)
Everyone.
Lex Fridman (23:23.540)
So think about, I'm gonna age myself,
Lex Fridman (23:25.700)
but I remember when we didn't have like testers, right?
Lex Fridman (23:30.140)
And so what did you do?
Lex Fridman (23:31.100)
As a developer, you had to test your own code, right?
Lex Fridman (23:33.580)
Like you had to go through all the cases and figure it out
Lex Fridman (23:36.140)
and then they realized that,
Lex Fridman (23:39.140)
we probably need to have testing
Ayanna Howard (23:40.620)
because we're not getting all the things.
Lex Fridman (23:42.460)
And so from there, what happens is like most developers,
Ayanna Howard (23:45.540)
they do a little bit of testing, but it's usually like,
Lex Fridman (23:48.100)
okay, did my compiler bug out?
Ayanna Howard (23:49.780)
Let me look at the warnings.
Lex Fridman (23:51.140)
Okay, is that acceptable or not, right?
Ayanna Howard (23:53.260)
Like that's how you typically think about as a developer
Lex Fridman (23:55.820)
and you'll just assume that it's going to go
Ayanna Howard (23:58.220)
to another process and they're gonna test it out.
Lex Fridman (24:01.100)
But I think we need to go back to those early days
Ayanna Howard (24:04.340)
when you're a developer, you're developing,
Lex Fridman (24:07.540)
there should be like the say,
Ayanna Howard (24:09.500)
okay, let me look at the ethical outcomes of this
Lex Fridman (24:12.180)
because there isn't a second like testing ethical testers,
Ayanna Howard (24:16.020)
right, it's you.
Lex Fridman (24:18.060)
We did it back in the early coding days.
Ayanna Howard (24:21.180)
I think that's where we are with respect to ethics.
Lex Fridman (24:23.300)
Like let's go back to what was good practices
Lex Fridman (24:26.300)
and only because we were just developing the field.
Lex Fridman (24:30.060)
Yeah, and it's a really heavy burden.
Ayanna Howard (24:33.980)
I've had to feel it recently in the last few months,
Lex Fridman (24:37.500)
but I think it's a good one to feel like
Ayanna Howard (24:39.420)
I've gotten a message, more than one from people.
Lex Fridman (24:43.380)
You know, I've unfortunately gotten some attention recently
Lex Fridman (24:47.420)
and I've gotten messages that say that
Lex Fridman (24:50.380)
I have blood on my hands
Ayanna Howard (24:52.300)
because of working on semi autonomous vehicles.
Lex Fridman (24:56.260)
So the idea that you have semi autonomy means
Ayanna Howard (24:59.220)
people will become, will lose vigilance and so on.
Lex Fridman (25:02.020)
That's actually be humans, as we described.
Lex Fridman (25:05.140)
And because of that, because of this idea
Lex Fridman (25:08.100)
that we're creating automation,
Ayanna Howard (25:10.060)
there'll be people be hurt because of it.
Lex Fridman (25:12.780)
And I think that's a beautiful thing.
Ayanna Howard (25:14.540)
I mean, it's, you know, there's many nights
Lex Fridman (25:16.220)
where I wasn't able to sleep because of this notion.
Ayanna Howard (25:18.820)
You know, you really do think about people that might die
Lex Fridman (25:22.380)
because of this technology.
Ayanna Howard (25:23.860)
Of course, you can then start rationalizing saying,
Lex Fridman (25:26.580)
well, you know what, 40,000 people die in the United States
Ayanna Howard (25:29.100)
every year and we're trying to ultimately try to save lives.
Lex Fridman (25:32.380)
But the reality is your code you've written
Ayanna Howard (25:35.780)
might kill somebody.
Lex Fridman (25:36.700)
And that's an important burden to carry with you
Ayanna Howard (25:38.900)
as you design the code.
Lex Fridman (25:41.180)
I don't even think of it as a burden
Ayanna Howard (25:43.820)
if we train this concept correctly from the beginning.
Lex Fridman (25:47.540)
And I use, and not to say that coding is like
Ayanna Howard (25:50.300)
being a medical doctor, but think about it.
Lex Fridman (25:52.420)
Medical doctors, if they've been in situations
Lex Fridman (25:56.100)
where their patient didn't survive, right?
Lex Fridman (25:58.300)
Do they give up and go away?
Ayanna Howard (26:00.820)
No, every time they come in,
Lex Fridman (26:02.540)
they know that there might be a possibility
Ayanna Howard (26:05.460)
that this patient might not survive.
Lex Fridman (26:07.260)
And so when they approach every decision,
Ayanna Howard (26:10.140)
like that's in the back of their head.
Lex Fridman (26:11.980)
And so why isn't that we aren't teaching,
Lex Fridman (26:15.860)
and those are tools though, right?
Lex Fridman (26:17.220)
They are given some of the tools to address that
Lex Fridman (26:19.740)
so that they don't go crazy.
Lex Fridman (26:21.500)
But we don't give those tools
Lex Fridman (26:24.220)
so that it does feel like a burden
Lex Fridman (26:26.180)
versus something of I have a great gift
Lex Fridman (26:28.700)
and I can do great, awesome good,
Lex Fridman (26:31.100)
but with it comes great responsibility.
Ayanna Howard (26:33.340)
I mean, that's what we teach in terms of
Lex Fridman (26:35.820)
if you think about the medical schools, right?
Ayanna Howard (26:37.420)
Great gift, great responsibility.
Lex Fridman (26:39.540)
I think if we just change the messaging a little,
Ayanna Howard (26:42.140)
great gift, being a developer, great responsibility.
Lex Fridman (26:45.580)
And this is how you combine those.
Lex Fridman (26:48.340)
But do you think, I mean, this is really interesting.
Lex Fridman (26:52.180)
It's outside, I actually have no friends
Ayanna Howard (26:54.300)
who are sort of surgeons or doctors.
Lex Fridman (26:58.260)
I mean, what does it feel like
Ayanna Howard (27:00.020)
to make a mistake in a surgery and somebody to die
Lex Fridman (27:03.780)
because of that?
Ayanna Howard (27:04.780)
Like, is that something you could be taught
Lex Fridman (27:07.020)
in medical school, sort of how to be accepting of that risk?
Ayanna Howard (27:10.580)
So, because I do a lot of work with healthcare robotics,
Lex Fridman (27:14.940)
I have not lost a patient, for example.
Lex Fridman (27:18.460)
The first one's always the hardest, right?
Lex Fridman (27:20.900)
But they really teach the value, right?
Ayanna Howard (27:27.300)
So, they teach responsibility,
Lex Fridman (27:28.740)
but they also teach the value.
Ayanna Howard (27:30.780)
Like, you're saving 40,000,
Lex Fridman (27:34.700)
but in order to really feel good about that,
Ayanna Howard (27:38.260)
when you come to a decision,
Lex Fridman (27:40.100)
you have to be able to say at the end,
Lex Fridman (27:42.220)
I did all that I could possibly do, right?
Lex Fridman (27:45.300)
Versus a, well, I just picked the first widget, right?
Ayanna Howard (27:49.100)
Like, so every decision is actually thought through.
Lex Fridman (27:52.220)
It's not a habit, it's not a,
Ayanna Howard (27:53.780)
let me just take the best algorithm
Lex Fridman (27:55.340)
that my friend gave me, right?
Lex Fridman (27:57.060)
It's a, is this it, is this the best?
Lex Fridman (27:59.540)
Have I done my best to do good, right?
Lex Fridman (28:03.100)
And so...
Lex Fridman (28:03.940)
You're right, and I think burden is the wrong word.
Ayanna Howard (28:06.500)
It's a gift, but you have to treat it extremely seriously.
Lex Fridman (28:10.740)
Correct.
Ayanna Howard (28:13.260)
So, on a slightly related note,
Lex Fridman (28:15.500)
in a recent paper,
Ayanna Howard (28:16.420)
The Ugly Truth About Ourselves and Our Robot Creations,
Lex Fridman (28:20.140)
you discuss, you highlight some biases
Ayanna Howard (28:24.300)
that may affect the function of various robotic systems.
Lex Fridman (28:27.100)
Can you talk through, if you remember, examples of some?
Ayanna Howard (28:30.100)
There's a lot of examples.
Lex Fridman (28:31.300)
I usually... What is bias, first of all?
Ayanna Howard (28:33.060)
Yeah, so bias is this,
Lex Fridman (28:37.060)
and so bias, which is different than prejudice.
Ayanna Howard (28:38.820)
So, bias is that we all have these preconceived notions
Lex Fridman (28:41.860)
about particular, everything from particular groups
Lex Fridman (28:45.940)
to habits to identity, right?
Lex Fridman (28:49.700)
So, we have these predispositions,
Lex Fridman (28:51.420)
and so when we address a problem,
Lex Fridman (28:54.100)
we look at a problem and make a decision,
Ayanna Howard (28:56.020)
those preconceived notions might affect our outputs,
Lex Fridman (29:01.340)
our outcomes.
Ayanna Howard (29:02.220)
So, there the bias can be positive and negative,
Lex Fridman (29:04.700)
and then is prejudice the negative kind of bias?
Lex Fridman (29:07.980)
Prejudice is the negative, right?
Lex Fridman (29:09.180)
So, prejudice is that not only are you aware of your bias,
Lex Fridman (29:13.540)
but you are then take it and have a negative outcome,
Lex Fridman (29:18.820)
even though you're aware, like...
Lex Fridman (29:20.660)
And there could be gray areas too.
Lex Fridman (29:22.980)
There's always gray areas.
Ayanna Howard (29:24.620)
That's the challenging aspect of all ethical questions.
Lex Fridman (29:27.580)
So, I always like...
Ayanna Howard (29:28.620)
So, there's a funny one,
Lex Fridman (29:30.020)
and in fact, I think it might be in the paper,
Ayanna Howard (29:31.740)
because I think I talk about self driving cars,
Lex Fridman (29:34.180)
but think about this.
Lex Fridman (29:35.460)
We, for teenagers, right?
Lex Fridman (29:39.500)
Typically, insurance companies charge quite a bit of money
Ayanna Howard (29:44.540)
if you have a teenage driver.
Lex Fridman (29:46.740)
So, you could say that's an age bias, right?
Lex Fridman (29:50.860)
But no one will claim...
Lex Fridman (29:52.380)
I mean, parents will be grumpy,
Lex Fridman (29:54.060)
but no one really says that that's not fair.
Lex Fridman (29:58.660)
That's interesting.
Ayanna Howard (29:59.500)
We don't...
Lex Fridman (30:00.340)
That's right, that's right.
Ayanna Howard (30:01.580)
It's everybody in human factors and safety research almost...
Lex Fridman (30:06.580)
I mean, it's quite ruthlessly critical of teenagers.
Lex Fridman (30:12.780)
And we don't question, is that okay?
Lex Fridman (30:15.020)
Is that okay to be ageist in this kind of way?
Lex Fridman (30:17.140)
It is, and it is ageist, right?
Lex Fridman (30:18.580)
It's definitely ageist, there's no question about it.
Lex Fridman (30:20.780)
And so, this is the gray area, right?
Lex Fridman (30:24.940)
Because you know that teenagers are more likely
Ayanna Howard (30:29.820)
to be in accidents,
Lex Fridman (30:30.860)
and so, there's actually some data to it.
Lex Fridman (30:33.060)
But then, if you take that same example,
Lex Fridman (30:34.980)
and you say, well, I'm going to make the insurance higher
Ayanna Howard (30:39.380)
for an area of Boston,
Lex Fridman (30:43.380)
because there's a lot of accidents.
Lex Fridman (30:45.020)
And then, they find out that that's correlated
Lex Fridman (30:48.260)
with socioeconomics.
Lex Fridman (30:50.220)
Well, then it becomes a problem, right?
Lex Fridman (30:52.420)
Like, that is not acceptable,
Lex Fridman (30:55.180)
but yet, the teenager, which is age...
Lex Fridman (30:58.940)
It's against age, is, right?
Ayanna Howard (31:01.820)
We figure that out as a society by having conversations,
Lex Fridman (31:05.260)
by having discourse.
Ayanna Howard (31:06.180)
I mean, throughout history,
Lex Fridman (31:07.540)
the definition of what is ethical or not has changed,
Lex Fridman (31:11.340)
and hopefully, always for the better.
Lex Fridman (31:14.300)
Correct, correct.
Ayanna Howard (31:15.420)
So, in terms of bias or prejudice in algorithms,
Lex Fridman (31:22.300)
what examples do you sometimes think about?
Ayanna Howard (31:25.540)
So, I think about quite a bit the medical domain,
Lex Fridman (31:28.940)
just because historically, right?
Ayanna Howard (31:31.260)
The healthcare domain has had these biases,
Lex Fridman (31:34.500)
typically based on gender and ethnicity, primarily.
Ayanna Howard (31:40.220)
A little in age, but not so much.
Lex Fridman (31:43.660)
Historically, if you think about FDA and drug trials,
Ayanna Howard (31:49.260)
it's harder to find a woman that aren't childbearing,
Lex Fridman (31:54.540)
and so you may not test on drugs at the same level.
Ayanna Howard (31:56.900)
Right, so there's these things.
Lex Fridman (31:58.940)
And so, if you think about robotics, right?
Ayanna Howard (32:02.900)
Something as simple as,
Lex Fridman (32:04.860)
I'd like to design an exoskeleton, right?
Lex Fridman (32:07.740)
What should the material be?
Lex Fridman (32:09.180)
What should the weight be?
Lex Fridman (32:10.140)
What should the form factor be?
Lex Fridman (32:14.260)
Who are you gonna design it around?
Ayanna Howard (32:16.940)
I will say that in the US,
Lex Fridman (32:19.620)
women average height and weight
Ayanna Howard (32:21.620)
is slightly different than guys.
Lex Fridman (32:23.380)
So, who are you gonna choose?
Ayanna Howard (32:25.820)
Like, if you're not thinking about it from the beginning,
Lex Fridman (32:28.900)
as, okay, when I design this and I look at the algorithms
Lex Fridman (32:33.420)
and I design the control system and the forces
Lex Fridman (32:35.540)
and the torques, if you're not thinking about,
Ayanna Howard (32:38.060)
well, you have different types of body structure,
Lex Fridman (32:41.500)
you're gonna design to what you're used to.
Lex Fridman (32:44.380)
Oh, this fits all the folks in my lab, right?
Lex Fridman (32:48.060)
So, think about it from the very beginning is important.
Lex Fridman (32:51.300)
What about sort of algorithms that train on data
Lex Fridman (32:54.500)
kind of thing?
Ayanna Howard (32:55.940)
Sadly, our society already has a lot of negative bias.
Lex Fridman (33:01.140)
And so, if we collect a lot of data,
Ayanna Howard (33:04.540)
even if it's a balanced way,
Lex Fridman (33:06.100)
that's going to contain the same bias
Ayanna Howard (33:07.620)
that our society contains.
Lex Fridman (33:08.820)
And so, yeah, is there things there that bother you?
Ayanna Howard (33:13.540)
Yeah, so you actually said something.
Lex Fridman (33:15.420)
You had said how we have biases,
Lex Fridman (33:19.740)
but hopefully we learn from them and we become better, right?
Lex Fridman (33:22.940)
And so, that's where we are now, right?
Ayanna Howard (33:24.940)
So, the data that we're collecting is historic.
Lex Fridman (33:28.420)
So, it's based on these things
Ayanna Howard (33:29.940)
when we knew it was bad to discriminate,
Lex Fridman (33:32.420)
but that's the data we have and we're trying to fix it now,
Lex Fridman (33:35.900)
but we're fixing it based on the data
Lex Fridman (33:37.660)
that was used in the first place.
Ayanna Howard (33:39.260)
Fix it in post.
Lex Fridman (33:40.460)
Right, and so the decisions,
Lex Fridman (33:43.580)
and you can look at everything from the whole aspect
Lex Fridman (33:46.700)
of predictive policing, criminal recidivism.
Ayanna Howard (33:51.220)
There was a recent paper that had the healthcare algorithms,
Lex Fridman (33:54.100)
which had a kind of a sensational titles.
Ayanna Howard (33:58.020)
I'm not pro sensationalism in titles,
Lex Fridman (34:00.980)
but again, you read it, right?
Ayanna Howard (34:03.540)
So, it makes you read it,
Lex Fridman (34:05.540)
but I'm like, really?
Ayanna Howard (34:06.780)
Like, ugh, you could have.
Lex Fridman (34:08.740)
What's the topic of the sensationalism?
Lex Fridman (34:10.580)
I mean, what's underneath it?
Lex Fridman (34:13.100)
What's, if you could sort of educate me
Ayanna Howard (34:16.100)
on what kind of bias creeps into the healthcare space.
Lex Fridman (34:18.940)
Yeah, so.
Ayanna Howard (34:19.780)
I mean, you already kind of mentioned.
Lex Fridman (34:21.260)
Yeah, so this one was the headline was
Ayanna Howard (34:24.820)
racist AI algorithms.
Lex Fridman (34:27.300)
Okay, like, okay, that's totally a clickbait title.
Lex Fridman (34:30.700)
And so you looked at it and so there was data
Lex Fridman (34:34.060)
that these researchers had collected.
Ayanna Howard (34:36.460)
I believe, I wanna say it was either Science or Nature.
Lex Fridman (34:39.220)
It just was just published,
Lex Fridman (34:40.460)
but they didn't have a sensational title.
Lex Fridman (34:42.420)
It was like the media.
Lex Fridman (34:44.700)
And so they had looked at demographics,
Lex Fridman (34:47.300)
I believe, between black and white women, right?
Lex Fridman (34:51.940)
And they showed that there was a discrepancy
Lex Fridman (34:56.660)
in the outcomes, right?
Lex Fridman (34:58.980)
And so, and it was tied to ethnicity, tied to race.
Lex Fridman (35:02.220)
The piece that the researchers did
Ayanna Howard (35:04.620)
actually went through the whole analysis, but of course.
Lex Fridman (35:08.620)
I mean, the journalists with AI are problematic
Ayanna Howard (35:11.900)
across the board, let's say.
Lex Fridman (35:14.140)
And so this is a problem, right?
Lex Fridman (35:15.980)
And so there's this thing about,
Lex Fridman (35:18.100)
oh, AI, it has all these problems.
Ayanna Howard (35:20.420)
We're doing it on historical data
Lex Fridman (35:22.740)
and the outcomes are uneven based on gender
Ayanna Howard (35:25.900)
or ethnicity or age.
Lex Fridman (35:27.940)
But I am always saying is like, yes,
Lex Fridman (35:30.660)
we need to do better, right?
Lex Fridman (35:32.340)
We need to do better.
Ayanna Howard (35:33.460)
It is our duty to do better.
Lex Fridman (35:36.620)
But the worst AI is still better than us.
Ayanna Howard (35:39.700)
Like, you take the best of us
Lex Fridman (35:41.820)
and we're still worse than the worst AI,
Ayanna Howard (35:44.020)
at least in terms of these things.
Lex Fridman (35:45.500)
And that's actually not discussed, right?
Lex Fridman (35:47.820)
And so I think, and that's why the sensational title, right?
Lex Fridman (35:51.780)
And so it's like, so then you can have individuals go like,
Ayanna Howard (35:54.180)
oh, we don't need to use this AI.
Lex Fridman (35:55.340)
I'm like, oh, no, no, no, no.
Ayanna Howard (35:56.620)
I want the AI instead of the doctors
Lex Fridman (36:00.780)
that provided that data,
Lex Fridman (36:01.860)
because it's still better than that, right?
Lex Fridman (36:04.060)
I think that's really important to linger on,
Ayanna Howard (36:06.660)
is the idea that this AI is racist.
Lex Fridman (36:10.300)
It's like, well, compared to what?
Ayanna Howard (36:14.020)
Sort of, I think we set, unfortunately,
Lex Fridman (36:20.100)
way too high of a bar for AI algorithms.
Lex Fridman (36:23.220)
And in the ethical space where perfect is,
Lex Fridman (36:25.940)
I would argue, probably impossible.
Ayanna Howard (36:28.940)
Then if we set the bar of perfection, essentially,
Lex Fridman (36:33.020)
of it has to be perfectly fair, whatever that means,
Ayanna Howard (36:37.500)
it means we're setting it up for failure.
Lex Fridman (36:39.580)
But that's really important to say what you just said,
Ayanna Howard (36:41.940)
which is, well, it's still better than it is.
Lex Fridman (36:44.900)
And one of the things I think
Ayanna Howard (36:46.860)
that we don't get enough credit for,
Lex Fridman (36:50.260)
just in terms of as developers,
Lex Fridman (36:52.140)
is that you can now poke at it, right?
Lex Fridman (36:55.820)
So it's harder to say, is this hospital,
Lex Fridman (36:58.820)
is this city doing something, right?
Lex Fridman (37:01.020)
Until someone brings in a civil case, right?
Ayanna Howard (37:04.380)
Well, with AI, it can process through all this data
Lex Fridman (37:07.100)
and say, hey, yes, there was an issue here,
Lex Fridman (37:12.500)
but here it is, we've identified it,
Lex Fridman (37:14.460)
and then the next step is to fix it.
Ayanna Howard (37:16.140)
I mean, that's a nice feedback loop
Lex Fridman (37:18.060)
versus waiting for someone to sue someone else
Lex Fridman (37:21.300)
before it's fixed, right?
Lex Fridman (37:22.740)
And so I think that power,
Lex Fridman (37:25.060)
we need to capitalize on a little bit more, right?
Lex Fridman (37:27.580)
Instead of having the sensational titles,
Ayanna Howard (37:29.660)
have the, okay, this is a problem,
Lex Fridman (37:33.300)
and this is how we're fixing it,
Lex Fridman (37:34.540)
and people are putting money to fix it
Lex Fridman (37:36.500)
because we can make it better.
Ayanna Howard (37:38.580)
I look at like facial recognition,
Lex Fridman (37:40.340)
how Joy, she basically called out a couple of companies
Lex Fridman (37:45.460)
and said, hey, and most of them were like,
Lex Fridman (37:48.220)
oh, embarrassment, and the next time it had been fixed,
Lex Fridman (37:53.020)
right, it had been fixed better, right?
Lex Fridman (37:54.860)
And then it was like, oh, here's some more issues.
Lex Fridman (37:56.740)
And I think that conversation then moves that needle
Lex Fridman (38:01.740)
to having much more fair and unbiased and ethical aspects,
Ayanna Howard (38:07.540)
as long as both sides, the developers are willing to say,
Lex Fridman (38:10.580)
okay, I hear you, yes, we are going to improve,
Lex Fridman (38:14.020)
and you have other developers who are like,
Lex Fridman (38:16.100)
hey, AI, it's wrong, but I love it, right?
Ayanna Howard (38:19.620)
Yes, so speaking of this really nice notion
Lex Fridman (38:23.020)
that AI is maybe flawed but better than humans,
Lex Fridman (38:26.980)
so just made me think of it,
Lex Fridman (38:29.140)
one example of flawed humans is our political system.
Lex Fridman (38:34.100)
Do you think, or you said judicial as well,
Lex Fridman (38:38.700)
do you have a hope for AI sort of being elected
Ayanna Howard (38:46.140)
for president or running our Congress
Lex Fridman (38:49.780)
or being able to be a powerful representative of the people?
Lex Fridman (38:53.940)
So I mentioned, and I truly believe that this whole world
Lex Fridman (38:58.940)
of AI is in partnerships with people.
Lex Fridman (39:01.340)
And so what does that mean?
Lex Fridman (39:02.420)
I don't believe, or maybe I just don't,
Ayanna Howard (39:07.620)
I don't believe that we should have an AI for president,
Lex Fridman (39:11.420)
but I do believe that a president
Lex Fridman (39:13.540)
should use AI as an advisor, right?
Lex Fridman (39:15.900)
Like, if you think about it,
Ayanna Howard (39:17.420)
every president has a cabinet of individuals
Lex Fridman (39:21.900)
that have different expertise
Lex Fridman (39:23.660)
that they should listen to, right?
Lex Fridman (39:26.060)
Like, that's kind of what we do.
Lex Fridman (39:27.980)
And you put smart people with smart expertise
Lex Fridman (39:31.100)
around certain issues, and you listen.
Ayanna Howard (39:33.420)
I don't see why AI can't function
Lex Fridman (39:35.700)
as one of those smart individuals giving input.
Lex Fridman (39:39.260)
So maybe there's an AI on healthcare,
Lex Fridman (39:41.020)
maybe there's an AI on education and right,
Lex Fridman (39:43.820)
like all of these things that a human is processing, right?
Lex Fridman (39:48.780)
Because at the end of the day,
Ayanna Howard (39:51.380)
there's people that are human
Lex Fridman (39:53.540)
that are going to be at the end of the decision.
Lex Fridman (39:55.500)
And I don't think as a world, as a culture, as a society,
Lex Fridman (39:59.260)
that we would totally, and this is us,
Ayanna Howard (40:02.980)
like this is some fallacy about us,
Lex Fridman (40:05.260)
but we need to see that leader, that person as human.
Lex Fridman (40:11.780)
And most people don't realize
Lex Fridman (40:13.180)
that like leaders have a whole lot of advice, right?
Ayanna Howard (40:16.940)
Like when they say something, it's not that they woke up,
Lex Fridman (40:19.500)
well, usually they don't wake up in the morning
Lex Fridman (40:21.780)
and be like, I have a brilliant idea, right?
Lex Fridman (40:24.340)
It's usually a, okay, let me listen.
Ayanna Howard (40:26.620)
I have a brilliant idea,
Lex Fridman (40:27.460)
but let me get a little bit of feedback on this.
Ayanna Howard (40:29.780)
Like, okay.
Lex Fridman (40:30.900)
And then it's a, yeah, that was an awesome idea
Ayanna Howard (40:33.020)
or it's like, yeah, let me go back.
Lex Fridman (40:35.780)
We already talked through a bunch of them,
Lex Fridman (40:37.300)
but are there some possible solutions
Lex Fridman (40:41.380)
to the bias that's present in our algorithms
Lex Fridman (40:45.100)
beyond what we just talked about?
Lex Fridman (40:46.540)
So I think there's two paths.
Ayanna Howard (40:49.180)
One is to figure out how to systematically
Lex Fridman (40:53.620)
do the feedback and corrections.
Lex Fridman (40:56.380)
So right now it's ad hoc, right?
Lex Fridman (40:57.980)
It's a researcher identify some outcomes
Lex Fridman (41:02.300)
that are not, don't seem to be fair, right?
Lex Fridman (41:05.260)
They publish it, they write about it.
Lex Fridman (41:07.780)
And the, either the developer or the companies
Lex Fridman (41:11.260)
that have adopted the algorithms may try to fix it, right?
Lex Fridman (41:14.100)
And so it's really ad hoc and it's not systematic.
Lex Fridman (41:18.700)
There's, it's just, it's kind of like,
Ayanna Howard (41:21.260)
I'm a researcher, that seems like an interesting problem,
Lex Fridman (41:24.460)
which means that there's a whole lot out there
Lex Fridman (41:26.340)
that's not being looked at, right?
Lex Fridman (41:28.900)
Cause it's kind of researcher driven.
Lex Fridman (41:32.740)
And I don't necessarily have a solution,
Lex Fridman (41:35.460)
but that process I think could be done a little bit better.
Ayanna Howard (41:41.020)
One way is I'm going to poke a little bit
Lex Fridman (41:44.820)
at some of the corporations, right?
Ayanna Howard (41:48.060)
Like maybe the corporations when they think
Lex Fridman (41:50.660)
about a product, they should, instead of,
Ayanna Howard (41:53.660)
in addition to hiring these, you know, bug,
Lex Fridman (41:57.780)
they give these.
Ayanna Howard (41:59.660)
Oh yeah, yeah, yeah.
Lex Fridman (42:01.420)
Like awards when you find a bug.
Ayanna Howard (42:02.780)
Yeah, security bug, you know, let's put it
Lex Fridman (42:06.620)
like we will give the, whatever the award is
Ayanna Howard (42:09.580)
that we give for the people who find these security holes,
Lex Fridman (42:12.460)
find an ethics hole, right?
Ayanna Howard (42:13.820)
Like find an unfairness hole
Lex Fridman (42:15.220)
and we will pay you X for each one you find.
Lex Fridman (42:17.660)
I mean, why can't they do that?
Lex Fridman (42:19.620)
One is a win win.
Ayanna Howard (42:20.900)
They show that they're concerned about it,
Lex Fridman (42:22.940)
that this is important and they don't have
Ayanna Howard (42:24.980)
to necessarily dedicate it their own like internal resources.
Lex Fridman (42:28.660)
And it also means that everyone who has
Ayanna Howard (42:30.780)
like their own bias lens, like I'm interested in age.
Lex Fridman (42:34.460)
And so I'll find the ones based on age
Lex Fridman (42:36.420)
and I'm interested in gender and right,
Lex Fridman (42:38.260)
which means that you get like all
Ayanna Howard (42:39.860)
of these different perspectives.
Lex Fridman (42:41.420)
But you think of it in a data driven way.
Lex Fridman (42:43.220)
So like sort of, if we look at a company like Twitter,
Lex Fridman (42:48.220)
it gets, it's under a lot of fire
Ayanna Howard (42:51.660)
for discriminating against certain political beliefs.
Lex Fridman (42:54.820)
Correct.
Lex Fridman (42:55.880)
And sort of, there's a lot of people,
Lex Fridman (42:58.060)
this is the sad thing,
Ayanna Howard (42:59.260)
cause I know how hard the problem is
Lex Fridman (43:00.700)
and I know the Twitter folks are working really hard at it.
Ayanna Howard (43:03.060)
Even Facebook that everyone seems to hate
Lex Fridman (43:04.980)
are working really hard at this.
Ayanna Howard (43:06.860)
You know, the kind of evidence that people bring
Lex Fridman (43:09.320)
is basically anecdotal evidence.
Ayanna Howard (43:11.240)
Well, me or my friend, all we said is X
Lex Fridman (43:15.020)
and for that we got banned.
Lex Fridman (43:17.100)
And that's kind of a discussion of saying,
Lex Fridman (43:20.980)
well, look, that's usually, first of all,
Ayanna Howard (43:23.260)
the whole thing is taken out of context.
Lex Fridman (43:25.500)
So they present sort of anecdotal evidence.
Lex Fridman (43:28.660)
And how are you supposed to, as a company,
Lex Fridman (43:31.140)
in a healthy way, have a discourse
Lex Fridman (43:33.080)
about what is and isn't ethical?
Lex Fridman (43:35.980)
How do we make algorithms ethical
Lex Fridman (43:38.060)
when people are just blowing everything?
Lex Fridman (43:40.780)
Like they're outraged about a particular
Ayanna Howard (43:45.140)
anecdotal piece of evidence that's very difficult
Lex Fridman (43:48.220)
to sort of contextualize in the big data driven way.
Lex Fridman (43:52.660)
Do you have a hope for companies like Twitter and Facebook?
Lex Fridman (43:55.900)
Yeah, so I think there's a couple of things going on, right?
Ayanna Howard (43:59.820)
First off, remember this whole aspect
Lex Fridman (44:04.860)
of we are becoming reliant on technology.
Ayanna Howard (44:09.420)
We're also becoming reliant on a lot of these,
Lex Fridman (44:14.380)
the apps and the resources that are provided, right?
Lex Fridman (44:17.980)
So some of it is kind of anger, like I need you, right?
Lex Fridman (44:21.660)
And you're not working for me, right?
Ayanna Howard (44:23.220)
Not working for me, all right.
Lex Fridman (44:24.660)
But I think, and so some of it,
Lex Fridman (44:27.300)
and I wish that there was a little bit
Lex Fridman (44:31.380)
of change of rethinking.
Lex Fridman (44:32.860)
So some of it is like, oh, we'll fix it in house.
Lex Fridman (44:35.560)
No, that's like, okay, I'm a fox
Lex Fridman (44:38.980)
and I'm going to watch these hens
Lex Fridman (44:40.940)
because I think it's a problem that foxes eat hens.
Lex Fridman (44:44.060)
No, right?
Lex Fridman (44:45.180)
Like be good citizens and say, look, we have a problem.
Lex Fridman (44:50.860)
And we are willing to open ourselves up
Lex Fridman (44:54.820)
for others to come in and look at it
Lex Fridman (44:57.060)
and not try to fix it in house.
Lex Fridman (44:58.740)
Because if you fix it in house,
Ayanna Howard (45:00.460)
there's conflict of interest.
Lex Fridman (45:01.940)
If I find something, I'm probably going to want to fix it
Lex Fridman (45:04.440)
and hopefully the media won't pick it up, right?
Lex Fridman (45:07.300)
And that then causes distrust
Ayanna Howard (45:09.320)
because someone inside is going to be mad at you
Lex Fridman (45:11.880)
and go out and talk about how,
Lex Fridman (45:13.580)
yeah, they canned the resume survey because it, right?
Lex Fridman (45:17.780)
Like be nice people.
Ayanna Howard (45:19.320)
Like just say, look, we have this issue.
Lex Fridman (45:22.760)
Community, help us fix it.
Lex Fridman (45:24.420)
And we will give you like, you know,
Lex Fridman (45:25.780)
the bug finder fee if you do.
Ayanna Howard (45:28.100)
Did you ever hope that the community,
Lex Fridman (45:31.260)
us as a human civilization on the whole is good
Lex Fridman (45:35.340)
and can be trusted to guide the future of our civilization
Lex Fridman (45:39.500)
into a positive direction?
Ayanna Howard (45:40.940)
I think so.
Lex Fridman (45:41.880)
So I'm an optimist, right?
Ayanna Howard (45:44.100)
And, you know, there were some dark times in history always.
Lex Fridman (45:49.980)
I think now we're in one of those dark times.
Ayanna Howard (45:52.900)
I truly do.
Lex Fridman (45:53.740)
In which aspect?
Ayanna Howard (45:54.620)
The polarization.
Lex Fridman (45:56.260)
And it's not just US, right?
Lex Fridman (45:57.560)
So if it was just US, I'd be like, yeah, it's a US thing,
Lex Fridman (46:00.020)
but we're seeing it like worldwide, this polarization.
Lex Fridman (46:04.380)
And so I worry about that.
Lex Fridman (46:06.540)
But I do fundamentally believe that at the end of the day,
Lex Fridman (46:11.980)
people are good, right?
Lex Fridman (46:13.420)
And why do I say that?
Ayanna Howard (46:14.780)
Because anytime there's a scenario
Lex Fridman (46:17.700)
where people are in danger and I will use,
Lex Fridman (46:20.820)
so Atlanta, we had a snowmageddon
Lex Fridman (46:24.260)
and people can laugh about that.
Ayanna Howard (46:26.620)
People at the time, so the city closed for, you know,
Lex Fridman (46:30.460)
little snow, but it was ice and the city closed down.
Lex Fridman (46:33.420)
But you had people opening up their homes and saying,
Lex Fridman (46:35.720)
hey, you have nowhere to go, come to my house, right?
Ayanna Howard (46:39.060)
Hotels were just saying like, sleep on the floor.
Lex Fridman (46:41.820)
Like places like, you know, the grocery stores were like,
Ayanna Howard (46:44.420)
hey, here's food.
Lex Fridman (46:45.940)
There was no like, oh, how much are you gonna pay me?
Ayanna Howard (46:47.940)
It was like this, such a community.
Lex Fridman (46:50.500)
And like people who didn't know each other,
Lex Fridman (46:52.140)
strangers were just like, can I give you a ride home?
Lex Fridman (46:55.540)
And that was a point I was like, you know what, like.
Ayanna Howard (46:59.420)
That reveals that the deeper thing is,
Lex Fridman (47:03.100)
there's a compassionate love that we all have within us.
Ayanna Howard (47:06.940)
It's just that when all of that is taken care of
Lex Fridman (47:09.500)
and get bored, we love drama.
Lex Fridman (47:11.300)
And that's, I think almost like the division
Lex Fridman (47:14.820)
is a sign of the times being good,
Ayanna Howard (47:17.100)
is that it's just entertaining
Lex Fridman (47:19.060)
on some unpleasant mammalian level to watch,
Ayanna Howard (47:24.220)
to disagree with others.
Lex Fridman (47:26.140)
And Twitter and Facebook are actually taking advantage
Ayanna Howard (47:30.260)
of that in a sense because it brings you back
Lex Fridman (47:33.220)
to the platform and they're advertiser driven,
Lex Fridman (47:36.180)
so they make a lot of money.
Lex Fridman (47:37.620)
So you go back and you click.
Ayanna Howard (47:39.300)
Love doesn't sell quite as well in terms of advertisement.
Lex Fridman (47:43.700)
It doesn't.
Lex Fridman (47:44.940)
So you've started your career
Lex Fridman (47:46.980)
at NASA Jet Propulsion Laboratory,
Lex Fridman (47:49.100)
but before I ask a few questions there,
Lex Fridman (47:51.980)
have you happened to have ever seen Space Odyssey,
Lex Fridman (47:54.460)
2001 Space Odyssey?
Lex Fridman (47:57.220)
Yes.
Ayanna Howard (47:58.060)
Okay, do you think HAL 9000,
Lex Fridman (48:01.420)
so we're talking about ethics.
Lex Fridman (48:03.420)
Do you think HAL did the right thing
Lex Fridman (48:06.700)
by taking the priority of the mission
Lex Fridman (48:08.580)
over the lives of the astronauts?
Lex Fridman (48:10.260)
Do you think HAL is good or evil?
Ayanna Howard (48:15.900)
Easy questions.
Lex Fridman (48:16.900)
Yeah.
Ayanna Howard (48:19.420)
HAL was misguided.
Lex Fridman (48:21.380)
You're one of the people that would be in charge
Ayanna Howard (48:24.060)
of an algorithm like HAL.
Lex Fridman (48:26.140)
Yeah.
Lex Fridman (48:26.980)
What would you do better?
Lex Fridman (48:28.340)
If you think about what happened
Lex Fridman (48:31.180)
was there was no fail safe, right?
Lex Fridman (48:35.380)
So perfection, right?
Lex Fridman (48:37.780)
Like what is that?
Lex Fridman (48:38.620)
I'm gonna make something that I think is perfect,
Lex Fridman (48:40.840)
but if my assumptions are wrong,
Lex Fridman (48:44.620)
it'll be perfect based on the wrong assumptions, right?
Ayanna Howard (48:47.560)
That's something that you don't know until you deploy
Lex Fridman (48:51.700)
and then you're like, oh yeah, messed up.
Lex Fridman (48:53.820)
But what that means is that when we design software,
Lex Fridman (48:58.340)
such as in Space Odyssey,
Ayanna Howard (49:00.300)
when we put things out,
Lex Fridman (49:02.100)
that there has to be a fail safe.
Ayanna Howard (49:04.000)
There has to be the ability that once it's out there,
Lex Fridman (49:07.700)
we can grade it as an F and it fails
Lex Fridman (49:11.360)
and it doesn't continue, right?
Lex Fridman (49:13.060)
There's some way that it can be brought in
Lex Fridman (49:16.020)
and removed in that aspect.
Lex Fridman (49:19.620)
Because that's what happened with HAL.
Ayanna Howard (49:21.060)
It was like assumptions were wrong.
Lex Fridman (49:23.740)
It was perfectly correct based on those assumptions
Lex Fridman (49:27.820)
and there was no way to change it,
Lex Fridman (49:31.020)
change the assumptions at all.
Lex Fridman (49:34.020)
And the change to fall back would be to a human.
Lex Fridman (49:37.020)
So you ultimately think like human should be,
Ayanna Howard (49:42.340)
it's not turtles or AI all the way down.
Lex Fridman (49:45.580)
It's at some point, there's a human that actually.
Ayanna Howard (49:47.820)
I still think that,
Lex Fridman (49:48.860)
and again, because I do human robot interaction,
Ayanna Howard (49:51.420)
I still think the human needs to be part of the equation
Lex Fridman (49:54.980)
at some point.
Lex Fridman (49:56.440)
So what, just looking back,
Lex Fridman (49:58.460)
what are some fascinating things in robotic space
Lex Fridman (50:01.900)
that NASA was working at the time?
Lex Fridman (50:03.460)
Or just in general, what have you gotten to play with
Lex Fridman (50:07.700)
and what are your memories from working at NASA?
Lex Fridman (50:10.060)
Yeah, so one of my first memories
Ayanna Howard (50:13.580)
was they were working on a surgical robot system
Lex Fridman (50:18.580)
that could do eye surgery, right?
Lex Fridman (50:21.880)
And this was back in, oh my gosh, it must've been,
Lex Fridman (50:25.700)
oh, maybe 92, 93, 94.
Lex Fridman (50:30.580)
So it's like almost like a remote operation.
Lex Fridman (50:32.880)
Yeah, it was remote operation.
Ayanna Howard (50:34.720)
In fact, you can even find some old tech reports on it.
Lex Fridman (50:38.400)
So think of it, like now we have DaVinci, right?
Lex Fridman (50:41.620)
Like think of it, but these were like the late 90s, right?
Lex Fridman (50:45.880)
And I remember going into the lab one day
Lex Fridman (50:48.240)
and I was like, what's that, right?
Lex Fridman (50:51.000)
And of course it wasn't pretty, right?
Ayanna Howard (50:53.960)
Because the technology, but it was like functional
Lex Fridman (50:56.640)
and you had this individual that could use
Ayanna Howard (50:59.240)
a version of haptics to actually do the surgery
Lex Fridman (51:01.960)
and they had this mockup of a human face
Lex Fridman (51:04.360)
and like the eyeballs and you can see this little drill.
Lex Fridman (51:08.480)
And I was like, oh, that is so cool.
Ayanna Howard (51:11.680)
That one I vividly remember
Lex Fridman (51:13.720)
because it was so outside of my like possible thoughts
Ayanna Howard (51:18.640)
of what could be done.
Lex Fridman (51:20.040)
It's the kind of precision
Lex Fridman (51:21.360)
and I mean, what's the most amazing of a thing like that?
Lex Fridman (51:26.120)
I think it was the precision.
Ayanna Howard (51:28.240)
It was the kind of first time
Lex Fridman (51:31.960)
that I had physically seen
Lex Fridman (51:34.880)
this robot machine human interface, right?
Lex Fridman (51:39.640)
Versus, cause manufacturing had been,
Lex Fridman (51:42.400)
you saw those kind of big robots, right?
Lex Fridman (51:44.520)
But this was like, oh, this is in a person.
Ayanna Howard (51:48.040)
There's a person and a robot like in the same space.
Lex Fridman (51:51.400)
I'm meeting them in person.
Ayanna Howard (51:53.000)
Like for me, it was a magical moment
Lex Fridman (51:55.440)
that I can't, it was life transforming
Ayanna Howard (51:57.900)
that I recently met Spot Mini from Boston Dynamics.
Lex Fridman (52:00.560)
Oh, see.
Ayanna Howard (52:01.400)
I don't know why, but on the human robot interaction
Lex Fridman (52:04.680)
for some reason I realized how easy it is to anthropomorphize
Lex Fridman (52:09.680)
and it was, I don't know, it was almost
Lex Fridman (52:12.580)
like falling in love, this feeling of meeting.
Lex Fridman (52:14.700)
And I've obviously seen these robots a lot
Lex Fridman (52:17.300)
on video and so on, but meeting in person,
Ayanna Howard (52:19.180)
just having that one on one time is different.
Lex Fridman (52:22.340)
So have you had a robot like that in your life
Lex Fridman (52:25.020)
that made you maybe fall in love with robotics?
Lex Fridman (52:28.300)
Sort of like meeting in person.
Ayanna Howard (52:32.140)
I mean, I loved robotics since, yeah.
Lex Fridman (52:35.860)
So I was a 12 year old.
Ayanna Howard (52:37.900)
Like I'm gonna be a roboticist, actually was,
Lex Fridman (52:40.020)
I called it cybernetics.
Lex Fridman (52:41.180)
But so my motivation was Bionic Woman.
Lex Fridman (52:44.700)
I don't know if you know that.
Lex Fridman (52:46.260)
And so, I mean, that was like a seminal moment,
Lex Fridman (52:49.500)
but I didn't meet, like that was TV, right?
Ayanna Howard (52:52.340)
Like it wasn't like I was in the same space and I met
Lex Fridman (52:54.500)
and I was like, oh my gosh, you're like real.
Ayanna Howard (52:56.540)
Just linking on Bionic Woman, which by the way,
Lex Fridman (52:58.820)
because I read that about you.
Ayanna Howard (53:01.100)
I watched bits of it and it's just so,
Lex Fridman (53:04.340)
no offense, terrible.
Ayanna Howard (53:05.520)
It's cheesy if you look at it now.
Lex Fridman (53:08.500)
It's cheesy, no.
Ayanna Howard (53:09.340)
I've seen a couple of reruns lately.
Lex Fridman (53:10.900)
But it's, but of course at the time it's probably
Ayanna Howard (53:15.100)
captured the imagination.
Lex Fridman (53:16.740)
But the sound effects.
Ayanna Howard (53:18.100)
Especially when you're younger, it just catch you.
Lex Fridman (53:23.100)
But which aspect, did you think of it,
Ayanna Howard (53:24.720)
you mentioned cybernetics, did you think of it as robotics
Lex Fridman (53:27.700)
or did you think of it as almost constructing
Lex Fridman (53:30.140)
artificial beings?
Lex Fridman (53:31.620)
Like, is it the intelligent part that captured
Lex Fridman (53:36.200)
your fascination or was it the whole thing?
Lex Fridman (53:38.060)
Like even just the limbs and just the.
Lex Fridman (53:39.820)
So for me, it would have, in another world,
Lex Fridman (53:42.900)
I probably would have been more of a biomedical engineer
Ayanna Howard (53:46.820)
because what fascinated me was the parts,
Lex Fridman (53:50.040)
like the bionic parts, the limbs, those aspects of it.
Lex Fridman (53:55.060)
Are you especially drawn to humanoid or humanlike robots?
Lex Fridman (53:59.620)
I would say humanlike, not humanoid, right?
Lex Fridman (54:03.060)
And when I say humanlike, I think it's this aspect
Lex Fridman (54:05.900)
of that interaction, whether it's social
Lex Fridman (54:09.140)
and it's like a dog, right?
Lex Fridman (54:10.660)
Like that's humanlike because it understand us,
Ayanna Howard (54:14.100)
it interacts with us at that very social level
Lex Fridman (54:18.500)
to, you know, humanoids are part of that,
Lex Fridman (54:21.860)
but only if they interact with us as if we are human.
Lex Fridman (54:26.860)
Okay, but just to linger on NASA for a little bit,
Lex Fridman (54:30.980)
what do you think, maybe if you have other memories,
Lex Fridman (54:34.100)
but also what do you think is the future of robots in space?
Ayanna Howard (54:38.580)
We'll mention how, but there's incredible robots
Lex Fridman (54:41.900)
that NASA's working on in general thinking about
Ayanna Howard (54:44.100)
in our, as we venture out, human civilization ventures out
Lex Fridman (54:49.820)
into space, what do you think the future of robots is there?
Ayanna Howard (54:52.260)
Yeah, so I mean, there's the near term.
Lex Fridman (54:53.700)
For example, they just announced the rover
Ayanna Howard (54:57.300)
that's going to the moon, which, you know,
Lex Fridman (55:00.780)
that's kind of exciting, but that's like near term.
Ayanna Howard (55:06.100)
You know, my favorite, favorite, favorite series
Lex Fridman (55:11.180)
is Star Trek, right?
Ayanna Howard (55:13.340)
You know, I really hope, and even Star Trek,
Lex Fridman (55:17.200)
like if I calculate the years, I wouldn't be alive,
Lex Fridman (55:20.100)
but I would really, really love to be in that world.
Lex Fridman (55:26.700)
Like, even if it's just at the beginning,
Ayanna Howard (55:28.460)
like, you know, like voyage, like adventure one.
Lex Fridman (55:33.180)
So basically living in space.
Ayanna Howard (55:35.740)
Yeah.
Lex Fridman (55:36.580)
With, what robots, what are robots?
Ayanna Howard (55:39.740)
With data.
Lex Fridman (55:40.580)
What role?
Ayanna Howard (55:41.400)
The data would have to be, even though that wasn't,
Lex Fridman (55:42.820)
you know, that was like later, but.
Lex Fridman (55:44.740)
So data is a robot that has human like qualities.
Lex Fridman (55:49.160)
Right, without the emotion chip.
Ayanna Howard (55:50.500)
Yeah.
Lex Fridman (55:51.340)
You don't like emotion.
Ayanna Howard (55:52.220)
Well, so data with the emotion chip
Lex Fridman (55:54.220)
was kind of a mess, right?
Ayanna Howard (55:58.580)
It took a while for that thing to adapt,
Lex Fridman (56:04.660)
but, and so why was that an issue?
Ayanna Howard (56:08.580)
The issue is that emotions make us irrational agents.
Lex Fridman (56:14.240)
That's the problem.
Lex Fridman (56:15.240)
And yet he could think through things,
Lex Fridman (56:20.040)
even if it was based on an emotional scenario, right?
Ayanna Howard (56:23.440)
Based on pros and cons.
Lex Fridman (56:25.080)
But as soon as you made him emotional,
Ayanna Howard (56:28.520)
one of the metrics he used for evaluation
Lex Fridman (56:31.160)
was his own emotions, not people around him, right?
Ayanna Howard (56:35.480)
Like, and so.
Lex Fridman (56:37.280)
We do that as children, right?
Lex Fridman (56:39.000)
So we're very egocentric when we're young.
Lex Fridman (56:40.920)
We are very egocentric.
Lex Fridman (56:42.320)
And so isn't that just an early version of the emotion chip
Lex Fridman (56:45.800)
then, I haven't watched much Star Trek.
Lex Fridman (56:48.280)
Except I have also met adults, right?
Lex Fridman (56:52.460)
And so that is a developmental process.
Lex Fridman (56:54.600)
And I'm sure there's a bunch of psychologists
Lex Fridman (56:57.600)
that can go through, like you can have a 60 year old adult
Lex Fridman (57:00.640)
who has the emotional maturity of a 10 year old, right?
Lex Fridman (57:04.640)
And so there's various phases that people should go through
Ayanna Howard (57:08.880)
in order to evolve and sometimes you don't.
Lex Fridman (57:11.480)
So how much psychology do you think,
Ayanna Howard (57:14.840)
a topic that's rarely mentioned in robotics,
Lex Fridman (57:17.600)
but how much does psychology come to play
Lex Fridman (57:19.700)
when you're talking about HRI, human robot interaction?
Lex Fridman (57:23.600)
When you have to have robots
Ayanna Howard (57:25.000)
that actually interact with humans.
Lex Fridman (57:26.120)
Tons.
Lex Fridman (57:26.960)
So we, like my group, as well as I read a lot
Lex Fridman (57:31.360)
in the cognitive science literature,
Ayanna Howard (57:33.280)
as well as the psychology literature.
Lex Fridman (57:36.160)
Because they understand a lot about human, human relations
Lex Fridman (57:42.720)
and developmental milestones and things like that.
Lex Fridman (57:45.920)
And so we tend to look to see what's been done out there.
Ayanna Howard (57:53.120)
Sometimes what we'll do is we'll try to match that to see,
Lex Fridman (57:56.500)
is that human, human relationship the same as human robot?
Ayanna Howard (58:00.980)
Sometimes it is, and sometimes it's different.
Lex Fridman (58:03.080)
And then when it's different, we have to,
Ayanna Howard (58:04.740)
we try to figure out, okay,
Lex Fridman (58:06.440)
why is it different in this scenario?
Lex Fridman (58:09.040)
But it's the same in the other scenario, right?
Lex Fridman (58:11.900)
And so we try to do that quite a bit.
Ayanna Howard (58:15.320)
Would you say that's, if we're looking at the future
Lex Fridman (58:17.800)
of human robot interaction,
Lex Fridman (58:19.140)
would you say the psychology piece is the hardest?
Lex Fridman (58:22.040)
Like if, I mean, it's a funny notion for you as,
Ayanna Howard (58:25.640)
I don't know if you consider, yeah.
Lex Fridman (58:27.360)
I mean, one way to ask it,
Lex Fridman (58:28.400)
do you consider yourself a roboticist or a psychologist?
Lex Fridman (58:32.000)
Oh, I consider myself a roboticist
Ayanna Howard (58:33.600)
that plays the act of a psychologist.
Lex Fridman (58:36.240)
But if you were to look at yourself sort of,
Ayanna Howard (58:40.120)
20, 30 years from now,
Lex Fridman (58:42.360)
do you see yourself more and more
Lex Fridman (58:43.880)
wearing the psychology hat?
Lex Fridman (58:47.560)
Another way to put it is,
Ayanna Howard (58:49.000)
are the hard problems in human robot interactions
Lex Fridman (58:51.600)
fundamentally psychology, or is it still robotics,
Ayanna Howard (58:55.800)
the perception manipulation, planning,
Lex Fridman (58:57.720)
all that kind of stuff?
Ayanna Howard (58:59.460)
It's actually neither.
Lex Fridman (59:01.680)
The hardest part is the adaptation and the interaction.
Lex Fridman (59:06.120)
So it's the interface, it's the learning.
Lex Fridman (59:08.840)
And so if I think of,
Ayanna Howard (59:11.600)
like I've become much more of a roboticist slash AI person
Lex Fridman (59:17.180)
than when I, like originally, again,
Ayanna Howard (59:19.040)
I was about the bionics.
Lex Fridman (59:20.160)
I was electrical engineer, I was control theory, right?
Lex Fridman (59:24.040)
And then I started realizing that my algorithms
Lex Fridman (59:28.780)
needed like human data, right?
Lex Fridman (59:30.600)
And so then I was like, okay, what is this human thing?
Lex Fridman (59:32.760)
How do I incorporate human data?
Lex Fridman (59:34.360)
And then I realized that human perception had,
Lex Fridman (59:38.440)
like there was a lot in terms of how we perceive the world.
Lex Fridman (59:41.040)
And so trying to figure out
Lex Fridman (59:41.940)
how do I model human perception for my,
Lex Fridman (59:44.400)
and so I became a HRI person,
Lex Fridman (59:47.600)
human robot interaction person,
Ayanna Howard (59:49.320)
from being a control theory and realizing
Lex Fridman (59:51.760)
that humans actually offered quite a bit.
Lex Fridman (59:55.220)
And then when you do that,
Lex Fridman (59:56.060)
you become more of an artificial intelligence, AI.
Lex Fridman (59:59.280)
And so I see myself evolving more in this AI world
🔗 相关节目