Rosalind Picard: Affective Computing, Emotion, Privacy, and Health
心理与人性技术与编程AI 与机器学习音乐与艺术生物与进化
📋 章节目录
暂无章节信息
🔑 关键词
donhumanscienceemotioncomputerdatabetterlookingbrainputneedspersonhelpdeepfuturemovieintelligencerobotinteractionmachine
💬 精彩语录
暂无语录
🎙️ 完整对话（1327 条）
Lex Fridman (00:00.000)
The following is a conversation with Rosalind Picard.
以下是与罗莎琳德·皮卡德的对话。
Lex Fridman (00:02.880)
She's a professor at MIT,
她是麻省理工学院的教授
Lex Fridman (00:04.540)
director of the Effective Computing Research Group
有效计算研究组主任
Lex Fridman (00:06.880)
at the MIT Media Lab,
在麻省理工学院媒体实验室，
Lex Fridman (00:08.360)
and cofounder of two companies, Affectiva and Empatica.
以及 Affectiva 和 Empatica 两家公司的联合创始人。
Rosalind Picard (00:12.440)
Over two decades ago,
二十多年前，
Lex Fridman (00:13.560)
she launched a field of effective computing
她开创了有效计算领域
Rosalind Picard (00:15.420)
with her book of the same name.
和她的同名书。
Lex Fridman (00:17.560)
This book described the importance of emotion
这本书阐述了情感的重要性
Rosalind Picard (00:20.040)
in artificial and natural intelligence.
在人工智能和自然智能方面。
Lex Fridman (00:23.040)
The vital role of emotional communication
情感沟通的重要作用
Rosalind Picard (00:25.320)
has to the relationship between people in general
与一般人之间的关系有关
Lex Fridman (00:28.520)
and human robot interaction.
以及人机交互。
Rosalind Picard (00:30.880)
I really enjoy talking with Ros over so many topics,
我真的很喜欢和 Ros 谈论很多话题，
Lex Fridman (00:34.000)
including emotion, ethics, privacy, wearable computing,
包括情感、道德、隐私、可穿戴计算、
Lex Fridman (00:37.440)
and her recent research in epilepsy,
以及她最近对癫痫的研究，
Lex Fridman (00:39.680)
and even love and meaning.
甚至爱和意义。
Rosalind Picard (00:42.600)
This conversation is part
这段对话是一部分
Lex Fridman (00:43.960)
of the Artificial Intelligence Podcast.
人工智能播客。
Rosalind Picard (00:46.000)
If you enjoy it, subscribe on YouTube, iTunes,
如果您喜欢，请在 YouTube、iTunes、
Lex Fridman (00:48.720)
or simply connect with me on Twitter at Lex Friedman,
Rosalind Picard (00:51.920)
spelled F R I D.
Lex Fridman (00:53.960)
And now, here's my conversation with Rosalind Picard.
Rosalind Picard (00:59.480)
More than 20 years ago,
Lex Fridman (01:00.720)
you've coined the term effective computing
Lex Fridman (01:03.320)
and led a lot of research in this area since then.
Lex Fridman (01:06.680)
As I understand, the goal is to make the machine detect
Lex Fridman (01:09.220)
and interpret the emotional state of a human being
Lex Fridman (01:12.380)
and adapt the behavior of the machine
Rosalind Picard (01:14.200)
based on the emotional state.
Lex Fridman (01:16.120)
So how is your understanding of the problem space
Lex Fridman (01:19.920)
defined by effective computing changed in the past 24 years?
Lex Fridman (01:25.360)
So it's the scope, the applications, the challenges,
Lex Fridman (01:28.880)
what's involved, how has that evolved over the years?
Lex Fridman (01:32.120)
Yeah, actually, originally,
Rosalind Picard (01:33.400)
when I defined the term affective computing,
Lex Fridman (01:36.880)
it was a bit broader than just recognizing
Lex Fridman (01:40.120)
and responding intelligently to human emotion,
Lex Fridman (01:42.260)
although those are probably the two pieces
Rosalind Picard (01:44.520)
that we've worked on the hardest.
Lex Fridman (01:47.120)
The original concept also encompassed machines
Rosalind Picard (01:50.680)
that would have mechanisms
Lex Fridman (01:52.480)
that functioned like human emotion does inside them.
Rosalind Picard (01:55.680)
It would be any computing that relates to arises from
Lex Fridman (01:59.000)
or deliberately influences human emotion.
Lex Fridman (02:02.560)
So the human computer interaction part
Lex Fridman (02:05.160)
is the part that people tend to see,
Rosalind Picard (02:07.880)
like if I'm really ticked off at my computer
Lex Fridman (02:11.000)
and I'm scowling at it and I'm cursing at it
Lex Fridman (02:13.480)
and it just keeps acting smiling and happy
Lex Fridman (02:15.720)
like that little paperclip used to do,
Rosalind Picard (02:17.880)
dancing, winking, that kind of thing
Lex Fridman (02:22.200)
just makes you even more frustrated, right?
Lex Fridman (02:24.640)
And I thought that stupid thing needs to see my affect.
Lex Fridman (02:29.120)
And if it's gonna be intelligent,
Rosalind Picard (02:30.640)
which Microsoft researchers had worked really hard on,
Lex Fridman (02:33.000)
it actually had some of the most sophisticated AI
Rosalind Picard (02:34.920)
in it at the time,
Lex Fridman (02:36.240)
that thing's gonna actually be smart.
Rosalind Picard (02:38.000)
It needs to respond to me and you,
Lex Fridman (02:41.600)
and we can send it very different signals.
Lex Fridman (02:45.360)
So by the way, just a quick interruption,
Lex Fridman (02:47.160)
the Clippy, maybe it's in Word 95, 98,
Rosalind Picard (02:52.600)
I don't remember when it was born,
Lex Fridman (02:54.360)
but many people, do you find yourself with that reference
Rosalind Picard (02:58.320)
that people recognize what you're talking about
Lex Fridman (03:00.320)
still to this point?
Rosalind Picard (03:01.680)
I don't expect the newest students to these days,
Lex Fridman (03:05.200)
but I've mentioned it to a lot of audiences,
Lex Fridman (03:07.200)
like how many of you know this Clippy thing?
Lex Fridman (03:09.240)
And still the majority of people seem to know it.
Lex Fridman (03:11.720)
So Clippy kind of looks at maybe natural language processing
Lex Fridman (03:15.340)
where you were typing and tries to help you complete,
Rosalind Picard (03:18.200)
I think.
Lex Fridman (03:19.280)
I don't even remember what Clippy was, except annoying.
Rosalind Picard (03:22.520)
Yeah, some people actually liked it.
Lex Fridman (03:25.840)
I would hear those stories.
Lex Fridman (03:27.520)
You miss it?
Lex Fridman (03:28.480)
Well, I miss the annoyance.
Rosalind Picard (03:31.300)
They felt like there's an element.
Lex Fridman (03:34.080)
Someone was there.
Rosalind Picard (03:34.920)
Somebody was there and we were in it together
Lex Fridman (03:36.960)
and they were annoying.
Rosalind Picard (03:37.800)
It's like a puppy that just doesn't get it.
Lex Fridman (03:40.880)
They keep stripping up the couch kind of thing.
Lex Fridman (03:42.200)
And in fact, they could have done it smarter like a puppy.
Lex Fridman (03:44.960)
If they had done, like if when you yelled at it
Rosalind Picard (03:48.000)
or cursed at it,
Lex Fridman (03:49.040)
if it had put its little ears back in its tail down
Lex Fridman (03:51.800)
and shrugged off,
Lex Fridman (03:52.960)
probably people would have wanted it back, right?
Lex Fridman (03:55.900)
But instead, when you yelled at it, what did it do?
Lex Fridman (03:58.600)
It smiled, it winked, it danced, right?
Rosalind Picard (04:01.260)
If somebody comes to my office and I yell at them,
Lex Fridman (04:03.200)
they start smiling, winking and dancing.
Rosalind Picard (04:04.760)
I'm like, I never want to see you again.
Lex Fridman (04:06.760)
So Bill Gates got a standing ovation
Rosalind Picard (04:08.520)
when he said it was going away
Lex Fridman (04:10.160)
because people were so ticked.
Lex Fridman (04:12.360)
It was so emotionally unintelligent, right?
Lex Fridman (04:15.040)
It was intelligent about whether you were writing a letter,
Lex Fridman (04:18.160)
what kind of help you needed for that context.
Lex Fridman (04:20.880)
It was completely unintelligent about,
Rosalind Picard (04:23.440)
hey, if you're annoying your customer,
Lex Fridman (04:25.760)
don't smile in their face when you do it.
Lex Fridman (04:28.400)
So that kind of mismatch was something
Lex Fridman (04:32.360)
the developers just didn't think about.
Lex Fridman (04:35.080)
And intelligence at the time was really all about math
Lex Fridman (04:39.520)
and language and chess and games,
Rosalind Picard (04:44.960)
problems that could be pretty well defined.
Lex Fridman (04:47.920)
Social emotional interaction is much more complex
Rosalind Picard (04:50.880)
than chess or Go or any of the games
Lex Fridman (04:53.640)
that people are trying to solve.
Lex Fridman (04:56.060)
And in order to understand that required skills
Lex Fridman (04:58.720)
that most people in computer science
Rosalind Picard (05:00.320)
actually were lacking personally.
Lex Fridman (05:02.600)
Well, let's talk about computer science.
Rosalind Picard (05:03.800)
Have things gotten better since the work,
Lex Fridman (05:06.400)
since the message,
Rosalind Picard (05:07.920)
since you've really launched the field
Lex Fridman (05:09.520)
with a lot of research work in this space?
Rosalind Picard (05:11.320)
I still find as a person like yourself,
Lex Fridman (05:14.080)
who's deeply passionate about human beings
Lex Fridman (05:16.680)
and yet am in computer science,
Lex Fridman (05:18.860)
there still seems to be a lack of,
Rosalind Picard (05:22.440)
sorry to say empathy in as computer scientists.
Lex Fridman (05:26.800)
Yeah, well.
Rosalind Picard (05:27.800)
Or hasn't gotten better.
Lex Fridman (05:28.880)
Let's just say there's a lot more variety
Rosalind Picard (05:30.720)
among computer scientists these days.
Lex Fridman (05:32.400)
Computer scientists are a much more diverse group today
Rosalind Picard (05:35.000)
than they were 25 years ago.
Lex Fridman (05:37.600)
And that's good.
Rosalind Picard (05:39.000)
We need all kinds of people to become computer scientists
Lex Fridman (05:41.760)
so that computer science reflects more what society needs.
Lex Fridman (05:45.580)
And there's brilliance among every personality type.
Lex Fridman (05:49.080)
So it need not be limited to people
Rosalind Picard (05:52.000)
who prefer computers to other people.
Lex Fridman (05:54.080)
How hard do you think it is?
Rosalind Picard (05:55.800)
Your view of how difficult it is to recognize emotion
Lex Fridman (05:58.580)
or to create a deeply emotionally intelligent interaction.
Rosalind Picard (06:03.920)
Has it gotten easier or harder
Lex Fridman (06:06.000)
as you've explored it further?
Lex Fridman (06:07.440)
And how far away are we from cracking this?
Lex Fridman (06:12.400)
If you think of the Turing test solving the intelligence,
Rosalind Picard (06:16.040)
looking at the Turing test for emotional intelligence.
Lex Fridman (06:20.720)
I think it is as difficult as I thought it was gonna be.
Rosalind Picard (06:25.560)
I think my prediction of its difficulty is spot on.
Lex Fridman (06:29.240)
I think the time estimates are always hard
Rosalind Picard (06:33.120)
because they're always a function of society's love
Lex Fridman (06:37.280)
and hate of a particular topic.
Rosalind Picard (06:39.440)
If society gets excited and you get thousands of researchers
Lex Fridman (06:45.200)
working on it for a certain application,
Rosalind Picard (06:49.000)
that application gets solved really quickly.
Lex Fridman (06:52.000)
The general intelligence,
Rosalind Picard (06:54.320)
the computer's complete lack of ability
Lex Fridman (06:58.120)
to have awareness of what it's doing,
Rosalind Picard (07:03.480)
the fact that it's not conscious,
Lex Fridman (07:05.480)
the fact that there's no signs of it becoming conscious,
Rosalind Picard (07:08.580)
the fact that it doesn't read between the lines,
Lex Fridman (07:11.800)
those kinds of things that we have to teach it explicitly,
Lex Fridman (07:15.000)
what other people pick up implicitly.
Lex Fridman (07:17.440)
We don't see that changing yet.
Rosalind Picard (07:20.360)
There aren't breakthroughs yet that lead us to believe
Lex Fridman (07:23.540)
that that's gonna go any faster,
Rosalind Picard (07:25.280)
which means that it's still gonna be kind of stuck
Lex Fridman (07:28.640)
with a lot of limitations
Rosalind Picard (07:31.240)
where it's probably only gonna do the right thing
Lex Fridman (07:34.000)
in very limited, narrow, prespecified contexts
Rosalind Picard (07:37.120)
where we can prescribe pretty much
Lex Fridman (07:40.880)
what's gonna happen there.
Lex Fridman (07:42.800)
So I don't see the,
Lex Fridman (07:46.920)
it's hard to predict a date
Rosalind Picard (07:47.960)
because when people don't work on it, it's infinite.
Lex Fridman (07:51.720)
When everybody works on it, you get a nice piece of it
Rosalind Picard (07:56.000)
well solved in a short amount of time.
Lex Fridman (07:58.560)
I actually think there's a more important issue right now
Rosalind Picard (08:01.520)
than the difficulty of it.
Lex Fridman (08:04.480)
And that's causing some of us
Rosalind Picard (08:05.760)
to put the brakes on a little bit.
Lex Fridman (08:07.360)
Usually we're all just like step on the gas,
Rosalind Picard (08:09.320)
let's go faster.
Lex Fridman (08:11.120)
This is causing us to pull back and put the brakes on.
Lex Fridman (08:14.160)
And that's the way that some of this technology
Lex Fridman (08:18.640)
is being used in places like China right now.
Lex Fridman (08:21.160)
And that worries me so deeply
Lex Fridman (08:24.480)
that it's causing me to pull back myself
Rosalind Picard (08:27.760)
on a lot of the things that we could be doing.
Lex Fridman (08:30.040)
And try to get the community to think a little bit more
Rosalind Picard (08:33.640)
about, okay, if we're gonna go forward with that,
Lex Fridman (08:36.000)
how can we do it in a way that puts in place safeguards
Lex Fridman (08:39.240)
that protects people?
Lex Fridman (08:41.080)
So the technology we're referring to is
Rosalind Picard (08:43.480)
just when a computer senses the human being,
Lex Fridman (08:46.360)
like the human face, right?
Lex Fridman (08:48.560)
So there's a lot of exciting things there,
Lex Fridman (08:51.800)
like forming a deep connection with the human being.
Lex Fridman (08:53.880)
So what are your worries, how that could go wrong?
Lex Fridman (08:57.920)
Is it in terms of privacy?
Lex Fridman (08:59.400)
Is it in terms of other kinds of more subtle things?
Lex Fridman (09:02.880)
But let's dig into privacy.
Lex Fridman (09:04.200)
So here in the US, if I'm watching a video
Lex Fridman (09:07.680)
of say a political leader,
Lex Fridman (09:09.760)
and in the US we're quite free as we all know
Lex Fridman (09:13.520)
to even criticize the president of the United States, right?
Rosalind Picard (09:17.800)
Here that's not a shocking thing.
Lex Fridman (09:19.320)
It happens about every five seconds, right?
Lex Fridman (09:22.600)
But in China, what happens if you criticize
Lex Fridman (09:27.600)
the leader of the government, right?
Lex Fridman (09:30.800)
And so people are very careful not to do that.
Lex Fridman (09:34.080)
However, what happens if you're simply watching a video
Lex Fridman (09:37.600)
and you make a facial expression
Lex Fridman (09:40.760)
that shows a little bit of skepticism, right?
Rosalind Picard (09:45.000)
Well, and here we're completely free to do that.
Lex Fridman (09:47.920)
In fact, we're free to fly off the handle
Lex Fridman (09:50.440)
and say anything we want, usually.
Lex Fridman (09:54.440)
I mean, there are some restrictions
Rosalind Picard (09:56.280)
when the athlete does this
Lex Fridman (09:58.800)
as part of the national broadcast.
Rosalind Picard (10:00.800)
Maybe the teams get a little unhappy
Lex Fridman (10:03.800)
about picking that forum to do it, right?
Lex Fridman (10:05.840)
But that's more a question of judgment.
Lex Fridman (10:08.680)
We have these freedoms,
Lex Fridman (10:11.520)
and in places that don't have those freedoms,
Lex Fridman (10:14.120)
what if our technology can read
Lex Fridman (10:17.040)
your underlying affective state?
Lex Fridman (10:19.560)
What if our technology can read it even noncontact?
Lex Fridman (10:22.400)
What if our technology can read it
Lex Fridman (10:24.400)
without your prior consent?
Lex Fridman (10:28.800)
And here in the US,
Lex Fridman (10:30.360)
in my first company we started, Affectiva,
Rosalind Picard (10:32.920)
we have worked super hard to turn away money
Lex Fridman (10:35.560)
and opportunities that try to read people's affect
Rosalind Picard (10:38.400)
without their prior informed consent.
Lex Fridman (10:41.320)
And even the software that is licensable,
Rosalind Picard (10:45.120)
you have to sign things saying
Lex Fridman (10:46.680)
you will only use it in certain ways,
Lex Fridman (10:48.360)
which essentially is get people's buy in, right?
Lex Fridman (10:52.080)
Don't do this without people agreeing to it.
Rosalind Picard (10:56.760)
There are other countries where they're not interested
Lex Fridman (10:58.560)
in people's buy in.
Rosalind Picard (10:59.520)
They're just gonna use it.
Lex Fridman (11:01.400)
They're gonna inflict it on you.
Lex Fridman (11:03.000)
And if you don't like it,
Lex Fridman (11:04.400)
you better not scowl in the direction of any censors.
Lex Fridman (11:08.440)
So one, let me just comment on a small tangent.
Lex Fridman (11:11.400)
Do you know with the idea of adversarial examples
Lex Fridman (11:15.920)
and deep fakes and so on,
Lex Fridman (11:18.760)
what you bring up is actually,
Rosalind Picard (11:20.760)
in that one sense, deep fakes provide
Lex Fridman (11:23.680)
a comforting protection that you can no longer really trust
Rosalind Picard (11:30.640)
that the video of your face was legitimate.
Lex Fridman (11:34.560)
And therefore you always have an escape clause
Rosalind Picard (11:37.040)
if a government is trying,
Lex Fridman (11:38.440)
if a stable, balanced, ethical government
Rosalind Picard (11:44.800)
is trying to accuse you of something,
Lex Fridman (11:46.200)
at least you have protection.
Rosalind Picard (11:47.080)
You can say it was fake news, as is a popular term now.
Lex Fridman (11:50.600)
Yeah, that's the general thinking of it.
Rosalind Picard (11:52.360)
We know how to go into the video
Lex Fridman (11:54.360)
and see, for example, your heart rate and respiration
Lex Fridman (11:58.360)
and whether or not they've been tampered with.
Lex Fridman (12:02.200)
And we also can put like fake heart rate and respiration
Rosalind Picard (12:05.520)
in your video now too.
Lex Fridman (12:06.680)
We decided we needed to do that.
Rosalind Picard (12:10.440)
After we developed a way to extract it,
Lex Fridman (12:12.640)
we decided we also needed a way to jam it.
Lex Fridman (12:15.920)
And so the fact that we took time to do that other step too,
Lex Fridman (12:20.880)
that was time that I wasn't spending
Rosalind Picard (12:22.520)
making the machine more affectively intelligent.
Lex Fridman (12:25.240)
And there's a choice in how we spend our time,
Rosalind Picard (12:28.480)
which is now being swayed a little bit less by this goal
Lex Fridman (12:32.400)
and a little bit more like by concern
Rosalind Picard (12:34.320)
about what's happening in society
Lex Fridman (12:36.560)
and what kind of future do we wanna build.
Lex Fridman (12:38.840)
And as we step back and say,
Lex Fridman (12:41.640)
okay, we don't just build AI to build AI
Rosalind Picard (12:44.560)
to make Elon Musk more money
Lex Fridman (12:46.480)
or to make Amazon Jeff Bezos more money.
Rosalind Picard (12:48.760)
Good gosh, you know, that's the wrong ethic.
Lex Fridman (12:52.840)
Why are we building it?
Lex Fridman (12:54.080)
What is the point of building AI?
Lex Fridman (12:57.160)
It used to be, it was driven by researchers in academia
Rosalind Picard (13:01.520)
to get papers published and to make a career for themselves
Lex Fridman (13:04.120)
and to do something cool, right?
Rosalind Picard (13:05.760)
Like, cause maybe it could be done.
Lex Fridman (13:08.480)
Now we realize that this is enabling rich people
Rosalind Picard (13:12.440)
to get vastly richer, the poor are,
Lex Fridman (13:17.200)
the divide is even larger.
Lex Fridman (13:19.760)
And is that the kind of future that we want?
Lex Fridman (13:22.840)
Maybe we wanna think about, maybe we wanna rethink AI.
Rosalind Picard (13:25.880)
Maybe we wanna rethink the problems in society
Lex Fridman (13:29.080)
that are causing the greatest inequity
Lex Fridman (13:32.720)
and rethink how to build AI
Lex Fridman (13:35.000)
that's not about a general intelligence,
Lex Fridman (13:36.720)
but that's about extending the intelligence
Lex Fridman (13:39.280)
and capability of the have nots
Lex Fridman (13:41.200)
so that we close these gaps in society.
Lex Fridman (13:43.760)
Do you hope that kind of stepping on the brake
Lex Fridman (13:46.600)
happens organically?
Lex Fridman (13:47.920)
Because I think still majority of the force behind AI
Rosalind Picard (13:51.160)
is the desire to publish papers,
Lex Fridman (13:52.720)
is to make money without thinking about the why.
Lex Fridman (13:55.480)
Do you hope it happens organically?
Lex Fridman (13:57.200)
Is there room for regulation?
Rosalind Picard (14:01.040)
Yeah, yeah, yeah, great questions.
Lex Fridman (14:02.920)
I prefer the, you know,
Rosalind Picard (14:05.920)
they talk about the carrot versus the stick.
Lex Fridman (14:07.320)
I definitely prefer the carrot to the stick.
Rosalind Picard (14:09.120)
And, you know, in our free world,
Lex Fridman (14:12.360)
we, there's only so much stick, right?
Rosalind Picard (14:14.880)
You're gonna find a way around it.
Lex Fridman (14:17.240)
I generally think less regulation is better.
Rosalind Picard (14:21.160)
That said, even though my position is classically carrot,
Lex Fridman (14:24.400)
no stick, no regulation,
Rosalind Picard (14:26.240)
I think we do need some regulations in this space.
Lex Fridman (14:29.040)
I do think we need regulations
Rosalind Picard (14:30.680)
around protecting people with their data,
Lex Fridman (14:33.560)
that you own your data, not Amazon, not Google.
Rosalind Picard (14:38.160)
I would like to see people own their own data.
Lex Fridman (14:40.760)
I would also like to see the regulations
Rosalind Picard (14:42.440)
that we have right now around lie detection
Lex Fridman (14:44.480)
being extended to emotion recognition in general,
Rosalind Picard (14:48.120)
that right now you can't use a lie detector on an employee
Lex Fridman (14:50.960)
when you're, on a candidate
Rosalind Picard (14:52.680)
when you're interviewing them for a job.
Lex Fridman (14:54.640)
I think similarly, we need to put in place protection
Rosalind Picard (14:57.720)
around reading people's emotions without their consent
Lex Fridman (15:00.520)
and in certain cases,
Rosalind Picard (15:02.120)
like characterizing them for a job and other opportunities.
Lex Fridman (15:06.080)
So I'm also, I also think that when we're reading emotion
Rosalind Picard (15:09.120)
that's predictive around mental health,
Lex Fridman (15:11.640)
that that should, even though it's not medical data,
Rosalind Picard (15:14.120)
that that should get the kinds of protections
Lex Fridman (15:16.040)
that our medical data gets.
Lex Fridman (15:18.440)
What most people don't know yet
Lex Fridman (15:19.960)
is right now with your smartphone use,
Lex Fridman (15:22.560)
and if you're wearing a sensor
Lex Fridman (15:25.160)
and you wanna learn about your stress and your sleep
Lex Fridman (15:27.680)
and your physical activity
Lex Fridman (15:28.960)
and how much you're using your phone
Lex Fridman (15:30.760)
and your social interaction,
Lex Fridman (15:32.560)
all of that nonmedical data,
Rosalind Picard (15:34.880)
when we put it together with machine learning,
Lex Fridman (15:37.880)
now called AI, even though the founders of AI
Rosalind Picard (15:40.080)
wouldn't have called it that,
Lex Fridman (15:42.840)
that capability can not only tell that you're calm right now
Rosalind Picard (15:48.360)
or that you're getting a little stressed,
Lex Fridman (15:50.760)
but it can also predict how you're likely to be tomorrow.
Rosalind Picard (15:53.840)
If you're likely to be sick or healthy,
Lex Fridman (15:55.760)
happy or sad, stressed or calm.
Rosalind Picard (15:58.640)
Especially when you're tracking data over time.
Lex Fridman (16:00.560)
Especially when we're tracking a week of your data or more.
Lex Fridman (16:03.680)
Do you have an optimism towards,
Lex Fridman (16:05.600)
you know, a lot of people on our phones
Rosalind Picard (16:07.720)
are worried about this camera that's looking at us.
Lex Fridman (16:10.280)
For the most part, on balance,
Rosalind Picard (16:12.480)
are you optimistic about the benefits
Lex Fridman (16:16.000)
that can be brought from that camera
Lex Fridman (16:17.400)
that's looking at billions of us?
Lex Fridman (16:19.560)
Or should we be more worried?
Rosalind Picard (16:24.520)
I think we should be a little bit more worried
Lex Fridman (16:28.840)
about who's looking at us and listening to us.
Rosalind Picard (16:32.480)
The device sitting on your countertop in your kitchen,
Lex Fridman (16:36.680)
whether it's, you know, Alexa or Google Home or Apple, Siri,
Rosalind Picard (16:42.160)
these devices want to listen
Lex Fridman (16:47.520)
while they say ostensibly to help us.
Lex Fridman (16:49.680)
And I think there are great people in these companies
Lex Fridman (16:52.080)
who do want to help people.
Rosalind Picard (16:54.360)
Let me not brand them all bad.
Lex Fridman (16:56.160)
I'm a user of products from all of these companies
Rosalind Picard (16:59.320)
I'm naming all the A companies, Alphabet, Apple, Amazon.
Lex Fridman (17:04.360)
They are awfully big companies, right?
Rosalind Picard (17:09.120)
They have incredible power.
Lex Fridman (17:11.520)
And you know, what if China were to buy them, right?
Lex Fridman (17:17.200)
And suddenly all of that data
Lex Fridman (17:19.880)
were not part of free America,
Lex Fridman (17:22.440)
but all of that data were part of somebody
Lex Fridman (17:24.400)
who just wants to take over the world
Lex Fridman (17:26.640)
and you submit to them.
Lex Fridman (17:27.920)
And guess what happens if you so much as smirk the wrong way
Lex Fridman (17:32.120)
when they say something that you don't like?
Lex Fridman (17:34.560)
Well, they have reeducation camps, right?
Rosalind Picard (17:37.440)
That's a nice word for them.
Lex Fridman (17:39.000)
By the way, they have a surplus of organs
Rosalind Picard (17:41.440)
for people who have surgery these days.
Lex Fridman (17:43.320)
They don't have an organ donation problem
Rosalind Picard (17:45.040)
because they take your blood and they know you're a match.
Lex Fridman (17:48.040)
And the doctors are on record of taking organs
Rosalind Picard (17:51.800)
from people who are perfectly healthy and not prisoners.
Lex Fridman (17:55.360)
They're just simply not the favored ones of the government.
Lex Fridman (17:59.600)
And you know, that's a pretty freaky evil society.
Lex Fridman (18:04.480)
And we can use the word evil there.
Rosalind Picard (18:06.480)
I was born in the Soviet Union.
Lex Fridman (18:07.840)
I can certainly connect to the worry that you're expressing.
Rosalind Picard (18:13.080)
At the same time, probably both you and I
Lex Fridman (18:15.440)
and you very much so,
Rosalind Picard (18:19.120)
you know, there's an exciting possibility
Lex Fridman (18:23.160)
that you can have a deep connection with a machine.
Rosalind Picard (18:27.720)
Yeah, yeah.
Lex Fridman (18:28.640)
Right, so.
Rosalind Picard (18:30.920)
Those of us, I've admitted students who say that they,
Lex Fridman (18:35.440)
you know, when you list like,
Rosalind Picard (18:36.760)
who do you most wish you could have lunch with
Lex Fridman (18:39.400)
or dinner with, right?
Lex Fridman (18:41.400)
And they'll write like, I don't like people.
Lex Fridman (18:43.360)
I just like computers.
Lex Fridman (18:44.800)
And one of them said to me once
Lex Fridman (18:46.360)
when I had this party at my house,
Rosalind Picard (18:49.520)
I want you to know,
Lex Fridman (18:51.160)
this is my only social event of the year,
Rosalind Picard (18:53.160)
my one social event of the year.
Lex Fridman (18:55.560)
Like, okay, now this is a brilliant
Lex Fridman (18:57.680)
machine learning person, right?
Lex Fridman (18:59.280)
And we need that kind of brilliance in machine learning.
Lex Fridman (19:01.920)
And I love that computer science welcomes people
Lex Fridman (19:04.760)
who love people and people who are very awkward
Rosalind Picard (19:07.200)
around people.
Lex Fridman (19:08.040)
I love that this is a field that anybody could join.
Rosalind Picard (19:12.720)
We need all kinds of people
Lex Fridman (19:14.960)
and you don't need to be a social person.
Rosalind Picard (19:16.720)
I'm not trying to force people who don't like people
Lex Fridman (19:19.000)
to suddenly become social.
Rosalind Picard (19:21.720)
At the same time,
Lex Fridman (19:23.880)
if most of the people building the AIs of the future
Rosalind Picard (19:26.480)
are the kind of people who don't like people,
Lex Fridman (19:29.400)
we've got a little bit of a problem.
Rosalind Picard (19:31.040)
Well, hold on a second.
Lex Fridman (19:31.920)
So let me push back on that.
Lex Fridman (19:33.400)
So don't you think a large percentage of the world
Lex Fridman (19:38.640)
can, you know, there's loneliness.
Rosalind Picard (19:40.880)
There is a huge problem with loneliness that's growing.
Lex Fridman (19:44.400)
And so there's a longing for connection.
Lex Fridman (19:47.560)
Do you...
Lex Fridman (19:49.080)
If you're lonely, you're part of a big and growing group.
Rosalind Picard (19:51.400)
Yes.
Lex Fridman (19:52.240)
So we're in it together, I guess.
Rosalind Picard (19:54.320)
If you're lonely, join the group.
Lex Fridman (19:56.120)
You're not alone.
Rosalind Picard (19:56.960)
You're not alone.
Lex Fridman (19:57.960)
That's a good line.
Lex Fridman (20:00.160)
But do you think there's...
Lex Fridman (20:03.160)
You talked about some worry,
Lex Fridman (20:04.600)
but do you think there's an exciting possibility
Lex Fridman (20:07.600)
that something like Alexa and these kinds of tools
Rosalind Picard (20:11.560)
can alleviate that loneliness
Lex Fridman (20:14.240)
in a way that other humans can't?
Rosalind Picard (20:16.640)
Yeah, yeah, definitely.
Lex Fridman (20:18.920)
I mean, a great book can kind of alleviate loneliness
Rosalind Picard (20:22.120)
because you just get sucked into this amazing story
Lex Fridman (20:25.000)
and you can't wait to go spend time with that character.
Lex Fridman (20:27.760)
And they're not a human character.
Lex Fridman (20:30.360)
There is a human behind it.
Lex Fridman (20:33.200)
But yeah, it can be an incredibly delightful way
Lex Fridman (20:35.400)
to pass the hours and it can meet needs.
Rosalind Picard (20:39.480)
Even, you know, I don't read those trashy romance books,
Lex Fridman (20:43.440)
but somebody does, right?
Lex Fridman (20:44.760)
And what are they getting from this?
Lex Fridman (20:46.200)
Well, probably some of that feeling of being there, right?
Rosalind Picard (20:50.720)
Being there in that social moment,
Lex Fridman (20:52.920)
that romantic moment or connecting with somebody.
Rosalind Picard (20:56.240)
I've had a similar experience
Lex Fridman (20:57.560)
reading some science fiction books, right?
Lex Fridman (20:59.400)
And connecting with the character.
Lex Fridman (21:00.560)
Orson Scott Card, you know, just amazing writing
Lex Fridman (21:04.160)
and Ender's Game and Speaker for the Dead, terrible title.
Lex Fridman (21:07.560)
But those kind of books that pull you into a character
Lex Fridman (21:11.000)
and you feel like you're, you feel very social.
Lex Fridman (21:13.880)
It's very connected, even though it's not responding to you.
Lex Fridman (21:17.280)
And a computer, of course, can respond to you.
Lex Fridman (21:19.720)
So it can deepen it, right?
Rosalind Picard (21:21.440)
You can have a very deep connection,
Lex Fridman (21:25.480)
much more than the movie Her, you know, plays up, right?
Rosalind Picard (21:29.400)
Well, much more.
Lex Fridman (21:30.640)
I mean, movie Her is already a pretty deep connection, right?
Lex Fridman (21:34.760)
Well, but it's just a movie, right?
Lex Fridman (21:36.760)
It's scripted.
Rosalind Picard (21:37.600)
It's just, you know, but I mean,
Lex Fridman (21:39.560)
like there can be a real interaction
Rosalind Picard (21:42.680)
where the character can learn and you can learn.
Lex Fridman (21:46.600)
You could imagine it not just being you and one character.
Rosalind Picard (21:49.560)
You could imagine a group of characters.
Lex Fridman (21:51.600)
You can imagine a group of people and characters,
Rosalind Picard (21:53.600)
human and AI connecting,
Lex Fridman (21:56.440)
where maybe a few people can't sort of be friends
Rosalind Picard (22:00.800)
with everybody, but the few people
Lex Fridman (22:02.880)
and their AIs can befriend more people.
Rosalind Picard (22:07.000)
There can be an extended human intelligence in there
Lex Fridman (22:10.320)
where each human can connect with more people that way.
Lex Fridman (22:14.880)
But it's still very limited, but there are just,
Lex Fridman (22:19.480)
what I mean is there are many more possibilities
Rosalind Picard (22:21.560)
than what's in that movie.
Lex Fridman (22:22.760)
So there's a tension here.
Lex Fridman (22:24.680)
So one, you expressed a really serious concern
Lex Fridman (22:27.360)
about privacy, about how governments
Rosalind Picard (22:29.120)
can misuse the information,
Lex Fridman (22:31.120)
and there's the possibility of this connection.
Lex Fridman (22:34.080)
So let's look at Alexa.
Lex Fridman (22:36.200)
So personal assistance.
Rosalind Picard (22:37.760)
For the most part, as far as I'm aware,
Lex Fridman (22:40.840)
they ignore your emotion.
Rosalind Picard (22:42.840)
They ignore even the context or the existence of you,
Lex Fridman (22:47.400)
the intricate, beautiful, complex aspects of who you are,
Rosalind Picard (22:52.200)
except maybe aspects of your voice
Lex Fridman (22:54.160)
that help it recognize for speech recognition.
Lex Fridman (22:58.360)
Do you think they should move towards
Lex Fridman (23:00.600)
trying to understand your emotion?
Rosalind Picard (23:03.160)
All of these companies are very interested
Lex Fridman (23:04.960)
in understanding human emotion.
Rosalind Picard (23:07.440)
They want, more people are telling Siri every day
Lex Fridman (23:11.400)
they want to kill themselves.
Rosalind Picard (23:13.720)
Apple wants to know the difference between
Lex Fridman (23:15.640)
if a person is really suicidal versus if a person
Lex Fridman (23:18.480)
is just kind of fooling around with Siri, right?
Lex Fridman (23:21.400)
The words may be the same, the tone of voice
Lex Fridman (23:25.560)
and what surrounds those words is pivotal to understand
Lex Fridman (23:31.360)
if they should respond in a very serious way,
Rosalind Picard (23:34.200)
bring help to that person,
Lex Fridman (23:35.920)
or if they should kind of jokingly tease back,
Lex Fridman (23:40.640)
ah, you just want to sell me for something else, right?
Lex Fridman (23:44.960)
Like, how do you respond when somebody says that?
Rosalind Picard (23:47.920)
Well, you do want to err on the side of being careful
Lex Fridman (23:51.440)
and taking it seriously.
Rosalind Picard (23:53.640)
People want to know if the person is happy or stressed
Lex Fridman (23:59.120)
in part, well, so let me give you an altruistic reason
Lex Fridman (24:03.160)
and a business profit motivated reason.
Lex Fridman (24:08.320)
And there are people in companies that operate
Rosalind Picard (24:11.000)
on both principles.
Lex Fridman (24:12.720)
The altruistic people really care about their customers
Lex Fridman (24:16.920)
and really care about helping you feel a little better
Lex Fridman (24:19.320)
at the end of the day.
Lex Fridman (24:20.240)
And it would just make those people happy
Lex Fridman (24:22.680)
if they knew that they made your life better.
Rosalind Picard (24:24.320)
If you came home stressed and after talking
Lex Fridman (24:27.000)
with their product, you felt better.
Rosalind Picard (24:29.920)
There are other people who maybe have studied
Lex Fridman (24:32.960)
the way affect affects decision making
Lex Fridman (24:35.120)
and prices people pay.
Lex Fridman (24:36.440)
And they know, I don't know if I should tell you,
Rosalind Picard (24:38.760)
like the work of Jen Lerner on heartstrings and purse strings,
Lex Fridman (24:43.960)
you know, if we manipulate you into a slightly sadder mood,
Lex Fridman (24:47.960)
you'll pay more, right?
Lex Fridman (24:50.800)
You'll pay more to change your situation.
Rosalind Picard (24:53.800)
You'll pay more for something you don't even need
Lex Fridman (24:55.800)
to make yourself feel better.
Rosalind Picard (24:58.040)
So, you know, if they sound a little sad,
Lex Fridman (25:00.120)
maybe I don't want to cheer them up.
Rosalind Picard (25:01.240)
Maybe first I want to help them get something,
Lex Fridman (25:04.800)
a little shopping therapy, right?
Rosalind Picard (25:07.400)
That helps them.
Lex Fridman (25:08.480)
Which is really difficult for a company
Rosalind Picard (25:09.880)
that's primarily funded on advertisement.
Lex Fridman (25:12.120)
So they're encouraged to get you to offer you products
Rosalind Picard (25:16.160)
or Amazon that's primarily funded
Lex Fridman (25:17.840)
on you buying things from their store.
Lex Fridman (25:20.040)
So I think we should be, you know,
Lex Fridman (25:22.120)
maybe we need regulation in the future
Rosalind Picard (25:24.120)
to put a little bit of a wall between these agents
Lex Fridman (25:27.240)
that have access to our emotion
Lex Fridman (25:29.120)
and agents that want to sell us stuff.
Lex Fridman (25:32.280)
Maybe there needs to be a little bit more
Rosalind Picard (25:35.560)
of a firewall in between those.
Lex Fridman (25:38.400)
So maybe digging in a little bit
Rosalind Picard (25:40.480)
on the interaction with Alexa,
Lex Fridman (25:42.200)
you mentioned, of course, a really serious concern
Rosalind Picard (25:44.880)
about like recognizing emotion,
Lex Fridman (25:46.680)
if somebody is speaking of suicide or depression and so on,
Lex Fridman (25:49.680)
but what about the actual interaction itself?
Lex Fridman (25:55.000)
Do you think, so if I, you know,
Rosalind Picard (25:57.840)
you mentioned Clippy and being annoying,
Lex Fridman (26:01.480)
what is the objective function we're trying to optimize?
Lex Fridman (26:04.200)
Is it minimize annoyingness or minimize or maximize happiness?
Lex Fridman (26:09.480)
Or if we look at human to human relations,
Rosalind Picard (26:12.440)
I think that push and pull, the tension, the dance,
Lex Fridman (26:15.480)
you know, the annoying, the flaws, that's what makes it fun.
Lex Fridman (26:19.840)
So is there a room for, like what is the objective function?
Lex Fridman (26:24.160)
There are times when you want to have a little push and pull,
Lex Fridman (26:26.720)
I think of kids sparring, right?
Lex Fridman (26:29.120)
You know, I see my sons and they,
Rosalind Picard (26:31.200)
one of them wants to provoke the other to be upset
Lex Fridman (26:33.720)
and that's fun.
Lex Fridman (26:34.720)
And it's actually healthy to learn where your limits are,
Lex Fridman (26:38.520)
to learn how to self regulate.
Rosalind Picard (26:40.080)
You can imagine a game where it's trying to make you mad
Lex Fridman (26:43.000)
and you're trying to show self control.
Lex Fridman (26:45.080)
And so if we're doing a AI human interaction
Lex Fridman (26:48.640)
that's helping build resilience and self control,
Rosalind Picard (26:51.240)
whether it's to learn how to not be a bully
Lex Fridman (26:54.040)
or how to turn the other cheek
Rosalind Picard (26:55.560)
or how to deal with an abusive person in your life,
Lex Fridman (26:58.920)
then you might need an AI that pushes your buttons, right?
Lex Fridman (27:04.480)
But in general, do you want an AI that pushes your buttons?
Lex Fridman (27:10.440)
Probably depends on your personality.
Rosalind Picard (27:12.080)
I don't, I want one that's respectful,
Lex Fridman (27:15.320)
that is there to serve me
Lex Fridman (27:18.160)
and that is there to extend my ability to do things.
Lex Fridman (27:23.200)
I'm not looking for a rival,
Rosalind Picard (27:25.160)
I'm looking for a helper.
Lex Fridman (27:27.240)
And that's the kind of AI I'd put my money on.
Rosalind Picard (27:30.200)
Your sense is for the majority of people in the world,
Lex Fridman (27:33.680)
in order to have a rich experience,
Rosalind Picard (27:35.120)
that's what they're looking for as well.
Lex Fridman (27:37.120)
So they're not looking,
Rosalind Picard (27:37.960)
if you look at the movie Her, spoiler alert,
Lex Fridman (27:40.800)
I believe the program that the woman in the movie Her
Rosalind Picard (27:46.280)
leaves the person for somebody else,
Lex Fridman (27:51.320)
says they don't wanna be dating anymore, right?
Rosalind Picard (27:54.360)
Like, do you, your sense is if Alexa said,
Lex Fridman (27:58.280)
you know what, I'm actually had enough of you for a while,
Lex Fridman (28:02.840)
so I'm gonna shut myself off.
Lex Fridman (28:04.880)
You don't see that as...
Lex Fridman (28:07.120)
I'd say you're trash, cause I paid for you, right?
Lex Fridman (28:10.120)
You, we've got to remember,
Lex Fridman (28:14.000)
and this is where this blending human AI
Lex Fridman (28:18.160)
as if we're equals is really deceptive
Rosalind Picard (28:22.520)
because AI is something at the end of the day
Lex Fridman (28:26.400)
that my students and I are making in the lab.
Lex Fridman (28:28.840)
And we're choosing what it's allowed to say,
Lex Fridman (28:33.120)
when it's allowed to speak, what it's allowed to listen to,
Lex Fridman (28:36.600)
what it's allowed to act on given the inputs
Lex Fridman (28:40.760)
that we choose to expose it to,
Lex Fridman (28:43.400)
what outputs it's allowed to have.
Lex Fridman (28:45.880)
It's all something made by a human.
Lex Fridman (28:49.320)
And if we wanna make something
Lex Fridman (28:50.560)
that makes our lives miserable, fine.
Rosalind Picard (28:52.920)
I wouldn't invest in it as a business,
Lex Fridman (28:56.640)
unless it's just there for self regulation training.
Lex Fridman (28:59.520)
But I think we need to think about
Lex Fridman (29:01.960)
what kind of future we want.
Lex Fridman (29:02.800)
And actually your question, I really like the,
Lex Fridman (29:05.560)
what is the objective function?
Lex Fridman (29:06.760)
Is it to calm people down?
Lex Fridman (29:09.320)
Sometimes.
Lex Fridman (29:10.560)
Is it to always make people happy and calm them down?
Lex Fridman (29:14.400)
Well, there was a book about that, right?
Rosalind Picard (29:16.080)
The brave new world, make everybody happy,
Lex Fridman (29:18.840)
take your Soma if you're unhappy, take your happy pill.
Lex Fridman (29:22.520)
And if you refuse to take your happy pill,
Lex Fridman (29:24.360)
well, we'll threaten you by sending you to Iceland
Rosalind Picard (29:28.600)
to live there.
Lex Fridman (29:29.600)
I lived in Iceland three years.
Rosalind Picard (29:30.840)
It's a great place.
Lex Fridman (29:31.800)
Don't take your Soma, then go to Iceland.
Rosalind Picard (29:35.240)
A little TV commercial there.
Lex Fridman (29:37.520)
Now I was a child there for a few years.
Rosalind Picard (29:39.240)
It's a wonderful place.
Lex Fridman (29:40.600)
So that part of the book never scared me.
Lex Fridman (29:43.240)
But really like, do we want AI to manipulate us
Lex Fridman (29:46.720)
into submission, into making us happy?
Rosalind Picard (29:49.080)
Well, if you are a, you know,
Lex Fridman (29:52.640)
like a power obsessed sick dictator individual
Rosalind Picard (29:56.080)
who only wants to control other people
Lex Fridman (29:57.640)
to get your jollies in life, then yeah,
Rosalind Picard (29:59.640)
you wanna use AI to extend your power and your scale
Lex Fridman (30:03.720)
to force people into submission.
Rosalind Picard (30:07.080)
If you believe that the human race is better off
Lex Fridman (30:10.080)
being given freedom and the opportunity
Rosalind Picard (30:12.120)
to do things that might surprise you,
Lex Fridman (30:15.360)
then you wanna use AI to extend people's ability to build,
Rosalind Picard (30:20.200)
you wanna build AI that extends human intelligence,
Lex Fridman (30:22.960)
that empowers the weak and helps balance the power
Rosalind Picard (30:27.320)
between the weak and the strong,
Lex Fridman (30:28.840)
not that gives more power to the strong.
Lex Fridman (30:32.440)
So in this process of empowering people and sensing people,
Lex Fridman (30:39.280)
what is your sense on emotion
Lex Fridman (30:41.280)
in terms of recognizing emotion?
Lex Fridman (30:42.680)
The difference between emotion that is shown
Lex Fridman (30:44.680)
and emotion that is felt.
Lex Fridman (30:46.640)
So yeah, emotion that is expressed on the surface
Rosalind Picard (30:52.640)
through your face, your body, and various other things,
Lex Fridman (30:56.560)
and what's actually going on deep inside
Rosalind Picard (30:58.840)
on the biological level, on the neuroscience level,
Lex Fridman (31:01.760)
or some kind of cognitive level.
Rosalind Picard (31:03.680)
Yeah, yeah.
Lex Fridman (31:05.720)
Whoa, no easy questions here.
Rosalind Picard (31:07.840)
Well, yeah, I'm sure there's no definitive answer,
Lex Fridman (31:11.280)
but what's your sense?
Lex Fridman (31:12.360)
How far can we get by just looking at the face?
Lex Fridman (31:16.160)
We're very limited when we just look at the face,
Lex Fridman (31:18.480)
but we can get further than most people think we can get.
Lex Fridman (31:21.920)
People think, hey, I have a great poker face,
Rosalind Picard (31:25.920)
therefore all you're ever gonna get from me is neutral.
Lex Fridman (31:28.280)
Well, that's naive.
Rosalind Picard (31:30.160)
We can read with the ordinary camera
Lex Fridman (31:32.680)
on your laptop or on your phone.
Rosalind Picard (31:34.920)
We can read from a neutral face if your heart is racing.
Lex Fridman (31:39.320)
We can read from a neutral face
Rosalind Picard (31:41.280)
if your breathing is becoming irregular
Lex Fridman (31:44.760)
and showing signs of stress.
Rosalind Picard (31:46.920)
We can read under some conditions
Lex Fridman (31:50.720)
that maybe I won't give you details on,
Lex Fridman (31:53.200)
how your heart rate variability power is changing.
Lex Fridman (31:57.080)
That could be a sign of stress,
Rosalind Picard (31:58.640)
even when your heart rate is not necessarily accelerating.
Lex Fridman (32:02.920)
So...
Lex Fridman (32:03.760)
Sorry, from physio sensors or from the face?
Lex Fridman (32:06.080)
From the color changes that you cannot even see,
Lex Fridman (32:09.160)
but the camera can see.
Lex Fridman (32:11.680)
That's amazing.
Lex Fridman (32:12.520)
So you can get a lot of signal, but...
Lex Fridman (32:15.320)
So we get things people can't see using a regular camera.
Lex Fridman (32:18.680)
And from that, we can tell things about your stress.
Lex Fridman (32:21.920)
So if you were just sitting there with a blank face
Rosalind Picard (32:25.600)
thinking nobody can read my emotion, well, you're wrong.
Lex Fridman (32:30.120)
Right, so that's really interesting,
Lex Fridman (32:31.840)
but that's from sort of visual information from the face.
Lex Fridman (32:34.520)
That's almost like cheating your way
Rosalind Picard (32:37.120)
to the physiological state of the body,
Lex Fridman (32:39.120)
by being very clever with what you can do with vision.
Rosalind Picard (32:42.600)
With signal processing.
Lex Fridman (32:43.440)
With signal processing.
Lex Fridman (32:44.360)
So that's really impressive.
Lex Fridman (32:45.320)
But if you just look at the stuff we humans can see,
Rosalind Picard (32:49.320)
the poker, the smile, the smirks,
Lex Fridman (32:52.240)
the subtle, all the facial actions.
Lex Fridman (32:54.320)
So then you can hide that on your face
Lex Fridman (32:55.960)
for a limited amount of time.
Rosalind Picard (32:57.240)
Now, if you're just going in for a brief interview
Lex Fridman (33:00.600)
and you're hiding it, that's pretty easy for most people.
Rosalind Picard (33:03.880)
If you are, however, surveilled constantly everywhere you go,
Lex Fridman (33:08.800)
then it's gonna say, gee, you know, Lex used to smile a lot
Lex Fridman (33:13.280)
and now I'm not seeing so many smiles.
Lex Fridman (33:15.920)
And Roz used to laugh a lot
Lex Fridman (33:20.240)
and smile a lot very spontaneously.
Lex Fridman (33:22.280)
And now I'm only seeing
Rosalind Picard (33:23.480)
these not so spontaneous looking smiles.
Lex Fridman (33:26.440)
And only when she's asked these questions.
Rosalind Picard (33:28.920)
You know, that's something's changed here.
Lex Fridman (33:31.720)
Probably not getting enough sleep.
Rosalind Picard (33:33.720)
We could look at that too.
Lex Fridman (33:35.200)
So now I have to be a little careful too.
Rosalind Picard (33:37.000)
When I say we, you think we can't read your emotion
Lex Fridman (33:40.920)
and we can, it's not that binary.
Lex Fridman (33:42.760)
What we're reading is more some physiological changes
Lex Fridman (33:45.960)
that relate to your activation.
Rosalind Picard (33:48.760)
Now, that doesn't mean that we know everything
Lex Fridman (33:51.800)
about how you feel.
Rosalind Picard (33:52.640)
In fact, we still know very little about how you feel.
Lex Fridman (33:54.880)
Your thoughts are still private.
Rosalind Picard (33:56.880)
Your nuanced feelings are still completely private.
Lex Fridman (34:01.120)
We can't read any of that.
Lex Fridman (34:02.920)
So there's some relief that we can't read that.
Lex Fridman (34:07.000)
Even brain imaging can't read that.
Rosalind Picard (34:09.800)
Wearables can't read that.
Lex Fridman (34:12.280)
However, as we read your body state changes
Lex Fridman (34:16.000)
and we know what's going on in your environment
Lex Fridman (34:18.520)
and we look at patterns of those over time,
Rosalind Picard (34:21.480)
we can start to make some inferences
Lex Fridman (34:24.960)
about what you might be feeling.
Lex Fridman (34:26.960)
And that is where it's not just the momentary feeling
Lex Fridman (34:31.400)
but it's more your stance toward things.
Lex Fridman (34:34.120)
And that could actually be a little bit more scary
Lex Fridman (34:37.040)
with certain kinds of governmental control freak people
Rosalind Picard (34:42.840)
who want to know more about are you on their team
Lex Fridman (34:46.800)
or are you not?
Lex Fridman (34:48.320)
And getting that information through over time.
Lex Fridman (34:50.320)
So you're saying there's a lot of signal
Rosalind Picard (34:51.640)
by looking at the change over time.
Lex Fridman (34:53.680)
Yeah.
Lex Fridman (34:54.520)
So you've done a lot of exciting work
Lex Fridman (34:56.600)
both in computer vision
Lex Fridman (34:57.800)
and physiological sense like wearables.
Lex Fridman (35:00.560)
What do you think is the best modality for,
Lex Fridman (35:03.480)
what's the best window into the emotional soul?
Lex Fridman (35:08.360)
Is it the face?
Lex Fridman (35:09.200)
Is it the voice?
Lex Fridman (35:10.160)
Depends what you want to know.
Rosalind Picard (35:11.920)
It depends what you want to know.
Lex Fridman (35:13.120)
It depends what you want to know.
Rosalind Picard (35:13.960)
Everything is informative.
Lex Fridman (35:15.600)
Everything we do is informative.
Lex Fridman (35:17.440)
So for health and wellbeing and things like that,
Lex Fridman (35:20.160)
do you find the wearable physiotechnical,
Rosalind Picard (35:22.680)
measuring physiological signals
Lex Fridman (35:24.840)
is the best for health based stuff?
Lex Fridman (35:29.320)
So here I'm going to answer empirically
Lex Fridman (35:31.880)
with data and studies we've been doing.
Rosalind Picard (35:34.680)
We've been doing studies.
Lex Fridman (35:36.040)
Now these are currently running
Rosalind Picard (35:38.280)
with lots of different kinds of people
Lex Fridman (35:39.600)
but where we've published data
Lex Fridman (35:41.880)
and I can speak publicly to it,
Lex Fridman (35:44.080)
the data are limited right now
Rosalind Picard (35:45.520)
to New England college students.
Lex Fridman (35:47.680)
So that's a small group.
Rosalind Picard (35:50.320)
Among New England college students,
Lex Fridman (35:52.440)
when they are wearing a wearable
Rosalind Picard (35:55.880)
like the empathic embrace here
Lex Fridman (35:57.640)
that's measuring skin conductance, movement, temperature.
Lex Fridman (36:01.760)
And when they are using a smartphone
Lex Fridman (36:05.800)
that is collecting their time of day
Rosalind Picard (36:09.400)
of when they're texting, who they're texting,
Lex Fridman (36:12.120)
their movement around it, their GPS,
Rosalind Picard (36:14.200)
the weather information based upon their location.
Lex Fridman (36:18.040)
And when it's using machine learning
Lex Fridman (36:19.360)
and putting all of that together
Lex Fridman (36:20.920)
and looking not just at right now
Lex Fridman (36:22.920)
but looking at your rhythm of behaviors
Lex Fridman (36:26.960)
over about a week.
Rosalind Picard (36:28.560)
When we look at that,
Lex Fridman (36:29.920)
we are very accurate at forecasting tomorrow's stress,
Rosalind Picard (36:33.560)
mood and happy, sad mood and health.
Lex Fridman (36:38.560)
And when we look at which pieces of that are most useful,
Rosalind Picard (36:43.680)
first of all, if you have all the pieces,
Lex Fridman (36:45.560)
you get the best results.
Rosalind Picard (36:48.320)
If you have only the wearable,
Lex Fridman (36:50.600)
you get the next best results.
Lex Fridman (36:52.680)
And that's still better than 80% accurate
Lex Fridman (36:56.520)
at forecasting tomorrow's levels.
Rosalind Picard (37:00.080)
Isn't that exciting because the wearable stuff
Lex Fridman (37:02.800)
with physiological information,
Rosalind Picard (37:05.320)
it feels like it violates privacy less
Lex Fridman (37:08.000)
than the noncontact face based methods.
Rosalind Picard (37:12.720)
Yeah, it's interesting.
Lex Fridman (37:14.040)
I think what people sometimes don't,
Rosalind Picard (37:16.560)
it's funny in the early days people would say,
Lex Fridman (37:18.880)
oh, wearing something or giving blood is invasive, right?
Rosalind Picard (37:22.560)
Whereas a camera is less invasive
Lex Fridman (37:24.400)
because it's not touching you.
Rosalind Picard (37:26.920)
I think on the contrary,
Lex Fridman (37:28.320)
the things that are not touching you are maybe the scariest
Rosalind Picard (37:31.200)
because you don't know when they're on or off.
Lex Fridman (37:33.880)
And you don't know who's behind it, right?
Rosalind Picard (37:39.920)
A wearable, depending upon what's happening
Lex Fridman (37:43.480)
to the data on it, if it's just stored locally
Rosalind Picard (37:46.760)
or if it's streaming and what it is being attached to,
Lex Fridman (37:52.640)
in a sense, you have the most control over it
Lex Fridman (37:54.960)
because it's also very easy to just take it off, right?
Lex Fridman (37:59.400)
Now it's not sensing me.
Lex Fridman (38:01.440)
So if I'm uncomfortable with what it's sensing,
Lex Fridman (38:05.320)
now I'm free, right?
Rosalind Picard (38:07.240)
If I'm comfortable with what it's sensing,
Lex Fridman (38:09.960)
then, and I happen to know everything about this one
Lex Fridman (38:12.800)
and what it's doing with it,
Lex Fridman (38:13.720)
so I'm quite comfortable with it,
Rosalind Picard (38:15.600)
then I have control, I'm comfortable.
Lex Fridman (38:20.240)
Control is one of the biggest factors for an individual
Rosalind Picard (38:24.720)
in reducing their stress.
Lex Fridman (38:26.600)
If I have control over it,
Rosalind Picard (38:28.160)
if I know all there is to know about it,
Lex Fridman (38:30.200)
then my stress is a lot lower
Lex Fridman (38:32.480)
and I'm making an informed choice
Lex Fridman (38:34.960)
about whether to wear it or not,
Rosalind Picard (38:36.640)
or when to wear it or not.
Lex Fridman (38:38.040)
I wanna wear it sometimes, maybe not others.
Rosalind Picard (38:40.320)
Right, so that control, yeah, I'm with you.
Lex Fridman (38:42.760)
That control, even if, yeah, the ability to turn it off,
Rosalind Picard (38:47.520)
that is a really important thing.
Lex Fridman (38:49.000)
It's huge.
Lex Fridman (38:49.840)
And we need to, maybe, if there's regulations,
Lex Fridman (38:53.440)
maybe that's number one to protect
Rosalind Picard (38:55.080)
is people's ability to, it's easy to opt out as to opt in.
Lex Fridman (38:59.960)
Right, so you've studied a bit of neuroscience as well.
Lex Fridman (39:04.480)
How have looking at our own minds,
Lex Fridman (39:08.240)
sort of the biological stuff or the neurobiological,
Rosalind Picard (39:12.840)
the neuroscience to get the signals in our brain,
Lex Fridman (39:17.400)
helped you understand the problem
Lex Fridman (39:18.760)
and the approach of effective computing, so?
Lex Fridman (39:21.880)
Originally, I was a computer architect
Lex Fridman (39:23.720)
and I was building hardware and computer designs
Lex Fridman (39:26.320)
and I wanted to build ones that worked like the brain.
Lex Fridman (39:28.240)
So I've been studying the brain
Lex Fridman (39:29.880)
as long as I've been studying how to build computers.
Lex Fridman (39:33.920)
Have you figured out anything yet?
Lex Fridman (39:36.160)
Very little.
Rosalind Picard (39:37.000)
It's so amazing.
Lex Fridman (39:39.400)
You know, they used to think like,
Rosalind Picard (39:40.720)
oh, if you remove this chunk of the brain
Lex Fridman (39:42.560)
and you find this function goes away,
Rosalind Picard (39:44.040)
well, that's the part of the brain that did it.
Lex Fridman (39:45.760)
And then later they realized
Rosalind Picard (39:46.880)
if you remove this other chunk of the brain,
Lex Fridman (39:48.320)
that function comes back and,
Rosalind Picard (39:50.120)
oh no, we really don't understand it.
Lex Fridman (39:52.800)
Brains are so interesting and changing all the time
Lex Fridman (39:56.200)
and able to change in ways
Lex Fridman (39:58.080)
that will probably continue to surprise us.
Rosalind Picard (40:02.120)
When we were measuring stress,
Lex Fridman (40:04.520)
you may know the story where we found
Rosalind Picard (40:07.160)
an unusual big skin conductance pattern on one wrist
Lex Fridman (40:10.680)
in one of our kids with autism.
Lex Fridman (40:14.120)
And in trying to figure out how on earth
Lex Fridman (40:15.480)
you could be stressed on one wrist and not the other,
Lex Fridman (40:17.760)
like how can you get sweaty on one wrist, right?
Lex Fridman (40:20.160)
When you get stressed
Rosalind Picard (40:21.520)
with that sympathetic fight or flight response,
Lex Fridman (40:23.280)
like you kind of should like sweat more
Rosalind Picard (40:25.120)
in some places than others,
Lex Fridman (40:26.240)
but not more on one wrist than the other.
Rosalind Picard (40:27.920)
That didn't make any sense.
Lex Fridman (40:30.840)
We learned that what had actually happened
Rosalind Picard (40:33.120)
was a part of his brain had unusual electrical activity
Lex Fridman (40:37.080)
and that caused an unusually large sweat response
Rosalind Picard (40:41.240)
on one wrist and not the other.
Lex Fridman (40:44.040)
And since then we've learned
Rosalind Picard (40:45.480)
that seizures cause this unusual electrical activity.
Lex Fridman (40:49.360)
And depending where the seizure is,
Rosalind Picard (40:51.240)
if it's in one place and it's staying there,
Lex Fridman (40:53.360)
you can have a big electrical response
Rosalind Picard (40:55.520)
we can pick up with a wearable at one part of the body.
Lex Fridman (40:58.480)
You can also have a seizure
Rosalind Picard (40:59.400)
that spreads over the whole brain,
Lex Fridman (41:00.480)
generalized grand mal seizure.
Lex Fridman (41:02.400)
And that response spreads
Lex Fridman (41:04.040)
and we can pick it up pretty much anywhere.
Rosalind Picard (41:07.160)
As we learned this and then later built Embrace
Lex Fridman (41:10.200)
that's now FDA cleared for seizure detection,
Rosalind Picard (41:13.120)
we have also built relationships
Lex Fridman (41:15.760)
with some of the most amazing doctors in the world
Rosalind Picard (41:18.480)
who not only help people
Lex Fridman (41:20.400)
with unusual brain activity or epilepsy,
Lex Fridman (41:23.040)
but some of them are also surgeons
Lex Fridman (41:24.440)
and they're going in and they're implanting electrodes,
Rosalind Picard (41:27.160)
not just to momentarily read the strange patterns
Lex Fridman (41:31.200)
of brain activity that we'd like to see return to normal,
Lex Fridman (41:35.080)
but also to read out continuously what's happening
Lex Fridman (41:37.320)
in some of these deep regions of the brain
Rosalind Picard (41:39.120)
during most of life when these patients are not seizing.
Lex Fridman (41:41.640)
Most of the time they're not seizing,
Rosalind Picard (41:42.960)
most of the time they're fine.
Lex Fridman (41:44.960)
And so we are now working on mapping
Rosalind Picard (41:47.920)
those deep brain regions
Lex Fridman (41:49.960)
that you can't even usually get with EEG scalp electrodes
Rosalind Picard (41:53.680)
because the changes deep inside don't reach the surface.
Lex Fridman (41:58.640)
But interesting when some of those regions
Rosalind Picard (42:00.480)
are activated, we see a big skin conductance response.
Lex Fridman (42:04.280)
Who would have thunk it, right?
Rosalind Picard (42:05.800)
Like nothing here, but something here.
Lex Fridman (42:07.840)
In fact, right after seizures
Rosalind Picard (42:10.400)
that we think are the most dangerous ones
Lex Fridman (42:12.600)
that precede what's called SUDEP,
Rosalind Picard (42:14.120)
Sudden Unexpected Death and Epilepsy,
Lex Fridman (42:16.960)
there's a period where the brainwaves go flat
Lex Fridman (42:19.520)
and it looks like the person's brain has stopped,
Lex Fridman (42:21.640)
but it hasn't.
Rosalind Picard (42:23.120)
The activity has gone deep into a region
Lex Fridman (42:26.400)
that can make the cortical activity look flat,
Rosalind Picard (42:29.120)
like a quick shutdown signal here.
Lex Fridman (42:32.600)
It can unfortunately cause breathing to stop
Rosalind Picard (42:35.560)
if it progresses long enough.
Lex Fridman (42:38.360)
Before that happens, we see a big skin conductance response
Rosalind Picard (42:42.080)
in the data that we have.
Lex Fridman (42:43.760)
The longer this flattening, the bigger our response here.
Lex Fridman (42:46.880)
So we have been trying to learn, you know, initially,
Lex Fridman (42:49.480)
like why are we getting a big response here
Lex Fridman (42:51.720)
when there's nothing here?
Lex Fridman (42:52.560)
Well, it turns out there's something much deeper.
Lex Fridman (42:55.600)
So we can now go inside the brains
Lex Fridman (42:57.760)
of some of these individuals, fabulous people
Rosalind Picard (43:01.280)
who usually aren't seizing,
Lex Fridman (43:03.480)
and get this data and start to map it.
Lex Fridman (43:05.600)
So that's the active research that we're doing right now
Lex Fridman (43:07.600)
with top medical partners.
Lex Fridman (43:09.360)
So this wearable sensor that's looking at skin conductance
Lex Fridman (43:12.960)
can capture sort of the ripples of the complexity
Rosalind Picard (43:17.160)
of what's going on in our brain.
Lex Fridman (43:18.880)
So this little device, you have a hope
Rosalind Picard (43:22.240)
that you can start to get the signal
Lex Fridman (43:24.920)
from the interesting things happening in the brain.
Rosalind Picard (43:27.880)
Yeah, we've already published the strong correlations
Lex Fridman (43:30.600)
between the size of this response
Lex Fridman (43:32.200)
and the flattening that happens afterwards.
Lex Fridman (43:35.360)
And unfortunately, also in a real SUDEP case
Rosalind Picard (43:38.320)
where the patient died because the, well, we don't know why.
Lex Fridman (43:42.360)
We don't know if somebody was there,
Rosalind Picard (43:43.600)
it would have definitely prevented it.
Lex Fridman (43:45.280)
But we know that most SUDEPs happen
Rosalind Picard (43:47.040)
when the person's alone.
Lex Fridman (43:48.600)
And in this case, a SUDEP is an acronym, S U D E P.
Lex Fridman (43:53.600)
And it stands for the number two cause
Lex Fridman (43:56.680)
of years of life lost actually
Rosalind Picard (43:58.960)
among all neurological disorders.
Lex Fridman (44:01.040)
Stroke is number one, SUDEP is number two,
Lex Fridman (44:03.480)
but most people haven't heard of it.
Lex Fridman (44:05.640)
Actually, I'll plug my TED talk,
Rosalind Picard (44:07.240)
it's on the front page of TED right now
Lex Fridman (44:09.280)
that talks about this.
Lex Fridman (44:11.160)
And we hope to change that.
Lex Fridman (44:13.560)
I hope everybody who's heard of SIDS and stroke
Rosalind Picard (44:17.160)
will now hear of SUDEP
Lex Fridman (44:18.760)
because we think in most cases it's preventable
Rosalind Picard (44:21.280)
if people take their meds and aren't alone
Lex Fridman (44:24.720)
when they have a seizure.
Rosalind Picard (44:26.200)
Not guaranteed to be preventable.
Lex Fridman (44:27.800)
There are some exceptions,
Lex Fridman (44:29.680)
but we think most cases probably are.
Lex Fridman (44:31.560)
So you had this embrace now in the version two wristband,
Rosalind Picard (44:35.000)
right, for epilepsy management.
Lex Fridman (44:39.280)
That's the one that's FDA approved?
Rosalind Picard (44:41.440)
Yes.
Lex Fridman (44:42.640)
Which is kind of a clear.
Rosalind Picard (44:43.480)
FDA cleared, they say.
Lex Fridman (44:45.200)
Sorry.
Rosalind Picard (44:46.040)
No, it's okay.
Lex Fridman (44:46.880)
It essentially means it's approved for marketing.
Rosalind Picard (44:49.400)
Got it.
Lex Fridman (44:50.240)
Just a side note, how difficult is that to do?
Rosalind Picard (44:52.960)
It's essentially getting FDA approval
Lex Fridman (44:54.880)
for computer science technology.
Rosalind Picard (44:57.000)
It's so agonizing.
Lex Fridman (44:58.000)
It's much harder than publishing multiple papers
Rosalind Picard (45:01.920)
in top medical journals.
Lex Fridman (45:04.120)
Yeah, we've published peer reviewed
Rosalind Picard (45:05.920)
top medical journal neurology, best results,
Lex Fridman (45:08.840)
and that's not good enough for the FDA.
Rosalind Picard (45:10.800)
Is that system,
Lex Fridman (45:12.520)
so if we look at the peer review of medical journals,
Rosalind Picard (45:14.880)
there's flaws, there's strengths,
Lex Fridman (45:16.800)
is the FDA approval process,
Lex Fridman (45:19.560)
how does it compare to the peer review process?
Lex Fridman (45:21.560)
Does it have the strength?
Rosalind Picard (45:23.160)
I'll take peer review over FDA any day.
Lex Fridman (45:25.720)
But is that a good thing?
Lex Fridman (45:26.600)
Is that a good thing for FDA?
Lex Fridman (45:28.040)
You're saying, does it stop some amazing technology
Lex Fridman (45:31.160)
from getting through?
Lex Fridman (45:32.320)
Yeah, it does.
Rosalind Picard (45:33.400)
The FDA performs a very important good role
Lex Fridman (45:36.240)
in keeping people safe.
Rosalind Picard (45:37.600)
They keep things,
Lex Fridman (45:39.480)
they put you through tons of safety testing
Lex Fridman (45:41.800)
and that's wonderful and that's great.
Lex Fridman (45:44.240)
I'm all in favor of the safety testing.
Lex Fridman (45:46.240)
But sometimes they put you through additional testing
Lex Fridman (45:51.000)
that they don't have to explain why they put you through it
Lex Fridman (45:54.080)
and you don't understand why you're going through it
Lex Fridman (45:56.680)
and it doesn't make sense.
Lex Fridman (45:58.440)
And that's very frustrating.
Lex Fridman (46:00.680)
And maybe they have really good reasons
Lex Fridman (46:04.400)
and they just would,
Lex Fridman (46:05.480)
it would do people a service to articulate those reasons.
Rosalind Picard (46:09.720)
Be more transparent.
Lex Fridman (46:10.640)
Be more transparent.
Lex Fridman (46:12.120)
So as part of Empatica, you have sensors.
Lex Fridman (46:15.760)
So what kind of problems can we crack?
Lex Fridman (46:17.800)
What kind of things from seizures to autism
Lex Fridman (46:24.480)
to I think I've heard you mentioned depression.
Lex Fridman (46:28.000)
What kind of things can we alleviate?
Lex Fridman (46:29.760)
Can we detect?
Rosalind Picard (46:30.720)
What's your hope of what,
Lex Fridman (46:32.240)
how we can make the world a better place
Lex Fridman (46:33.960)
with this wearable tech?
Lex Fridman (46:35.760)
I would really like to see my fellow brilliant researchers
Rosalind Picard (46:40.760)
step back and say, what are the really hard problems
Lex Fridman (46:46.200)
that we don't know how to solve
Rosalind Picard (46:47.760)
that come from people maybe we don't even see
Lex Fridman (46:50.440)
in our normal life because they're living
Rosalind Picard (46:52.800)
in the poor places.
Lex Fridman (46:54.240)
They're stuck on the bus.
Rosalind Picard (46:56.160)
They can't even afford the Uber or the Lyft
Lex Fridman (46:58.440)
or the data plan or all these other wonderful things
Rosalind Picard (47:02.200)
we have that we keep improving on.
Lex Fridman (47:04.360)
Meanwhile, there's all these folks left behind in the world
Lex Fridman (47:07.080)
and they're struggling with horrible diseases
Lex Fridman (47:09.520)
with depression, with epilepsy, with diabetes,
Rosalind Picard (47:12.960)
with just awful stuff that maybe a little more time
Lex Fridman (47:19.200)
and attention hanging out with them
Lex Fridman (47:20.440)
and learning what are their challenges in life?
Lex Fridman (47:22.800)
What are their needs?
Lex Fridman (47:24.000)
How do we help them have job skills?
Lex Fridman (47:25.640)
How do we help them have a hope and a future
Lex Fridman (47:28.520)
and a chance to have the great life
Lex Fridman (47:31.240)
that so many of us building technology have?
Lex Fridman (47:34.960)
And then how would that reshape the kinds of AI
Lex Fridman (47:37.920)
that we build? How would that reshape the new apps
Rosalind Picard (47:41.400)
that we build or the maybe we need to focus
Lex Fridman (47:44.040)
on how to make things more low cost and green
Lex Fridman (47:46.880)
instead of thousand dollar phones?
Lex Fridman (47:49.320)
I mean, come on, why can't we be thinking more
Lex Fridman (47:52.840)
about things that do more with less for these folks?
Lex Fridman (47:56.840)
Quality of life is not related to the cost of your phone.
Rosalind Picard (48:00.200)
It's not something that, it's been shown that what about
Lex Fridman (48:03.960)
$75,000 of income and happiness is the same, okay?
Rosalind Picard (48:08.440)
However, I can tell you, you get a lot of happiness
Lex Fridman (48:10.720)
from helping other people.
Rosalind Picard (48:12.320)
You get a lot more than $75,000 buys.
Lex Fridman (48:15.200)
So how do we connect up the people who have real needs
Rosalind Picard (48:19.320)
with the people who have the ability to build the future
Lex Fridman (48:21.920)
and build the kind of future that truly improves the lives
Lex Fridman (48:25.840)
of all the people that are currently being left behind?
Lex Fridman (48:28.720)
So let me return just briefly on a point,
Rosalind Picard (48:32.720)
maybe in the movie, Her.
Lex Fridman (48:35.280)
So do you think if we look farther into the future,
Rosalind Picard (48:37.720)
you said so much of the benefit from making our technology
Lex Fridman (48:41.240)
more empathetic to us human beings would make them
Rosalind Picard (48:46.240)
better tools, empower us, make our lives better.
Lex Fridman (48:50.320)
Well, if we look farther into the future,
Lex Fridman (48:51.960)
do you think we'll ever create an AI system
Lex Fridman (48:54.560)
that we can fall in love with?
Rosalind Picard (48:56.920)
That we can fall in love with and loves us back
Lex Fridman (49:00.280)
on a level that is similar to human to human interaction,
Lex Fridman (49:04.920)
like in the movie Her or beyond?
Lex Fridman (49:07.680)
I think we can simulate it in ways that could,
Rosalind Picard (49:13.920)
you know, sustain engagement for a while.
Lex Fridman (49:17.280)
Would it be as good as another person?
Rosalind Picard (49:20.160)
I don't think so, if you're used to like good people.
Lex Fridman (49:24.560)
Now, if you've just grown up with nothing but abuse
Lex Fridman (49:27.120)
and you can't stand human beings,
Lex Fridman (49:29.080)
can we do something that helps you there
Lex Fridman (49:32.160)
that gives you something through a machine?
Lex Fridman (49:34.000)
Yeah, but that's pretty low bar, right?
Rosalind Picard (49:36.160)
If you've only encountered pretty awful people.
Lex Fridman (49:39.120)
If you've encountered wonderful, amazing people,
Rosalind Picard (49:41.680)
we're nowhere near building anything like that.
Lex Fridman (49:44.800)
And I would not bet on building it.
Rosalind Picard (49:49.480)
I would bet instead on building the kinds of AI
Lex Fridman (49:53.160)
that helps kind of raise all boats,
Rosalind Picard (49:56.880)
that helps all people be better people,
Lex Fridman (49:59.480)
helps all people figure out if they're getting sick tomorrow
Lex Fridman (50:02.280)
and helps give them what they need to stay well tomorrow.
Lex Fridman (50:05.400)
That's the kind of AI I wanna build
Rosalind Picard (50:07.000)
that improves human lives,
Lex Fridman (50:09.040)
not the kind of AI that just walks on The Tonight Show
Lex Fridman (50:11.640)
and people go, wow, look how smart that is.
Lex Fridman (50:14.600)
Really?
Lex Fridman (50:15.800)
And then it goes back in a box, you know?
Lex Fridman (50:18.640)
So on that point,
Rosalind Picard (50:19.960)
if we continue looking a little bit into the future,
Lex Fridman (50:23.440)
do you think an AI that's empathetic
Lex Fridman (50:25.200)
and does improve our lives
Lex Fridman (50:28.960)
need to have a physical presence, a body?
Lex Fridman (50:31.840)
And even let me cautiously say the C word consciousness
Lex Fridman (50:38.720)
and even fear of mortality.
Lex Fridman (50:40.760)
So some of those human characteristics,
Lex Fridman (50:42.760)
do you think it needs to have those aspects
Rosalind Picard (50:45.920)
or can it remain simply a machine learning tool
Lex Fridman (50:50.880)
that learns from data of behavior
Rosalind Picard (50:53.400)
that learns to make us,
Lex Fridman (50:56.640)
based on previous patterns, feel better?
Lex Fridman (51:00.080)
Or does it need those elements of consciousness?
Lex Fridman (51:02.560)
It depends on your goals.
Rosalind Picard (51:03.760)
If you're making a movie, it needs a body.
Lex Fridman (51:06.720)
It needs a gorgeous body.
Rosalind Picard (51:08.000)
It needs to act like it has consciousness.
Lex Fridman (51:10.040)
It needs to act like it has emotion, right?
Rosalind Picard (51:11.680)
Because that's what sells.
Lex Fridman (51:13.360)
That's what's gonna get me to show up and enjoy the movie.
Rosalind Picard (51:16.280)
Okay.
Lex Fridman (51:17.800)
In real life, does it need all that?
Rosalind Picard (51:19.800)
Well, if you've read Orson Scott Card,
Lex Fridman (51:21.840)
Ender's Game, Speaker of the Dead,
Lex Fridman (51:23.520)
it could just be like a little voice in your earring, right?
Lex Fridman (51:26.720)
And you could have an intimate relationship
Lex Fridman (51:28.360)
and it could get to know you.
Lex Fridman (51:29.520)
And it doesn't need to be a robot.
Lex Fridman (51:34.160)
But that doesn't make this compelling of a movie, right?
Lex Fridman (51:37.160)
I mean, we already think it's kind of weird
Rosalind Picard (51:38.440)
when a guy looks like he's talking to himself on the train,
Lex Fridman (51:41.600)
even though it's earbuds.
Lex Fridman (51:43.760)
So we have these, embodied is more powerful.
Lex Fridman (51:49.680)
Embodied, when you compare interactions
Rosalind Picard (51:51.760)
with an embodied robot versus a video of a robot
Lex Fridman (51:55.280)
versus no robot, the robot is more engaging.
Rosalind Picard (52:00.160)
The robot gets our attention more.
Lex Fridman (52:01.720)
The robot, when you walk in your house,
Rosalind Picard (52:03.080)
is more likely to get you to remember to do the things
Lex Fridman (52:05.440)
that you asked it to do,
Rosalind Picard (52:06.480)
because it's kind of got a physical presence.
Lex Fridman (52:09.000)
You can avoid it if you don't like it.
Rosalind Picard (52:10.840)
It could see you're avoiding it.
Lex Fridman (52:12.440)
There's a lot of power to being embodied.
Rosalind Picard (52:14.760)
There will be embodied AIs.
Lex Fridman (52:17.160)
They have great power and opportunity and potential.
Rosalind Picard (52:22.040)
There will also be AIs that aren't embodied,
Lex Fridman (52:24.600)
that just are little software assistants
Rosalind Picard (52:28.520)
that help us with different things
Lex Fridman (52:30.240)
that may get to know things about us.
Lex Fridman (52:33.480)
Will they be conscious?
Lex Fridman (52:34.880)
There will be attempts to program them
Rosalind Picard (52:36.640)
to make them appear to be conscious.
Lex Fridman (52:39.280)
We can already write programs that make it look like,
Lex Fridman (52:41.760)
oh, what do you mean?
Lex Fridman (52:42.600)
Of course I'm aware that you're there, right?
Rosalind Picard (52:43.800)
I mean, it's trivial to say stuff like that.
Lex Fridman (52:45.720)
It's easy to fool people,
Lex Fridman (52:48.720)
but does it actually have conscious experience like we do?
Lex Fridman (52:53.400)
Nobody has a clue how to do that yet.
Rosalind Picard (52:55.920)
That seems to be something that is beyond
Lex Fridman (52:58.720)
what any of us knows how to build now.
Lex Fridman (53:01.480)
Will it have to have that?
Lex Fridman (53:03.720)
I think you can get pretty far
Rosalind Picard (53:05.520)
with a lot of stuff without it.
Lex Fridman (53:07.280)
But will we accord it rights?
Rosalind Picard (53:10.960)
Well, that's more a political game
Lex Fridman (53:13.320)
than it is a question of real consciousness.
Rosalind Picard (53:16.480)
Yeah, can you go to jail for turning off Alexa
Lex Fridman (53:18.720)
is the question for an election maybe a few decades from now.
Rosalind Picard (53:24.960)
Well, Sophia Robot's already been given rights
Lex Fridman (53:27.640)
as a citizen in Saudi Arabia, right?
Rosalind Picard (53:30.040)
Even before women have full rights.
Lex Fridman (53:33.680)
Then the robot was still put back in the box
Rosalind Picard (53:36.960)
to be shipped to the next place
Lex Fridman (53:39.600)
where it would get a paid appearance, right?
Rosalind Picard (53:42.760)
Yeah, it's dark and almost comedic, if not absurd.
Lex Fridman (53:50.160)
So I've heard you speak about your journey in finding faith.
Rosalind Picard (53:54.640)
Sure.
Lex Fridman (53:55.960)
And how you discovered some wisdoms about life
Lex Fridman (54:00.040)
and beyond from reading the Bible.
Lex Fridman (54:03.240)
And I've also heard you say that,
Rosalind Picard (54:05.440)
you said scientists too often assume
Lex Fridman (54:07.040)
that nothing exists beyond what can be currently measured.
Rosalind Picard (54:11.360)
Yeah, materialism.
Lex Fridman (54:12.640)
Materialism.
Lex Fridman (54:13.480)
And scientism, yeah.
Lex Fridman (54:14.880)
So in some sense, this assumption enables
Rosalind Picard (54:17.360)
the near term scientific method,
Lex Fridman (54:20.280)
assuming that we can uncover the mysteries of this world
Rosalind Picard (54:25.360)
by the mechanisms of measurement that we currently have.
Lex Fridman (54:28.280)
But we easily forget that we've made this assumption.
Lex Fridman (54:33.960)
So what do you think we miss out on
Lex Fridman (54:35.640)
by making that assumption?
Rosalind Picard (54:38.760)
It's fine to limit the scientific method
Lex Fridman (54:42.640)
to things we can measure and reason about and reproduce.
Rosalind Picard (54:47.720)
That's fine.
Lex Fridman (54:49.640)
I think we have to recognize
Rosalind Picard (54:51.160)
that sometimes we scientists also believe
Lex Fridman (54:53.160)
in things that happen historically.
Rosalind Picard (54:55.560)
Like I believe the Holocaust happened.
Lex Fridman (54:57.920)
I can't prove events from past history scientifically.
Lex Fridman (55:03.880)
You prove them with historical evidence, right?
Lex Fridman (55:06.480)
With the impact they had on people,
Rosalind Picard (55:08.200)
with eyewitness testimony and things like that.
Lex Fridman (55:11.640)
So a good thinker recognizes that science
Rosalind Picard (55:15.680)
is one of many ways to get knowledge.
Lex Fridman (55:19.440)
It's not the only way.
Lex Fridman (55:21.640)
And there's been some really bad philosophy
Lex Fridman (55:24.280)
and bad thinking recently, you can call it scientism,
Rosalind Picard (55:27.840)
where people say science is the only way to get to truth.
Lex Fridman (55:31.200)
And it's not, it just isn't.
Rosalind Picard (55:33.360)
There are other ways that work also.
Lex Fridman (55:35.960)
Like knowledge of love with someone.
Lex Fridman (55:38.520)
You don't prove your love through science, right?
Lex Fridman (55:43.600)
So history, philosophy, love,
Rosalind Picard (55:48.160)
a lot of other things in life show us
Lex Fridman (55:50.840)
that there's more ways to gain knowledge and truth
Rosalind Picard (55:55.840)
if you're willing to believe there is such a thing,
Lex Fridman (55:57.800)
and I believe there is, than science.
Rosalind Picard (56:01.040)
I do, I am a scientist, however.
Lex Fridman (56:03.080)
And in my science, I do limit my science
Rosalind Picard (56:05.880)
to the things that the scientific method can do.
Lex Fridman (56:09.600)
But I recognize that it's myopic
Rosalind Picard (56:11.800)
to say that that's all there is.
Lex Fridman (56:13.720)
Right, there's, just like you listed,
Rosalind Picard (56:15.680)
there's all the why questions.
Lex Fridman (56:17.120)
And really we know, if we're being honest with ourselves,
Rosalind Picard (56:20.520)
the percent of what we really know is basically zero
Lex Fridman (56:25.520)
relative to the full mystery of the...
Rosalind Picard (56:28.120)
Measure theory, a set of measure zero,
Lex Fridman (56:30.360)
if I have a finite amount of knowledge, which I do.
Lex Fridman (56:34.520)
So you said that you believe in truth.
Lex Fridman (56:37.960)
So let me ask that old question.
Lex Fridman (56:40.600)
What do you think this thing is all about?
Lex Fridman (56:42.800)
What's the life on earth?
Lex Fridman (56:44.800)
Life, the universe, and everything?
Lex Fridman (56:46.200)
And everything, what's the meaning?
Rosalind Picard (56:47.040)
I can't quote Douglas Adams 42.
Lex Fridman (56:49.680)
It's my favorite number.
Rosalind Picard (56:51.240)
By the way, that's my street address.
Lex Fridman (56:52.640)
My husband and I guessed the exact same number
Rosalind Picard (56:54.880)
for our house, we got to pick it.
Lex Fridman (56:57.760)
And there's a reason we picked 42, yeah.
Lex Fridman (57:00.120)
So is it just 42 or is there,
Lex Fridman (57:02.320)
do you have other words that you can put around it?
Rosalind Picard (57:05.360)
Well, I think there's a grand adventure
Lex Fridman (57:07.760)
and I think this life is a part of it.
Rosalind Picard (57:09.840)
I think there's a lot more to it than meets the eye
Lex Fridman (57:12.200)
and the heart and the mind and the soul here.
Rosalind Picard (57:14.440)
I think we see but through a glass dimly in this life.
Lex Fridman (57:18.000)
We see only a part of all there is to know.
Rosalind Picard (57:22.720)
If people haven't read the Bible, they should,
Lex Fridman (57:25.800)
if they consider themselves educated
Lex Fridman (57:27.920)
and you could read Proverbs
Lex Fridman (57:30.400)
and find tremendous wisdom in there
Rosalind Picard (57:33.320)
that cannot be scientifically proven.
Lex Fridman (57:35.680)
But when you read it, there's something in you,
Rosalind Picard (57:38.200)
like a musician knows when the instruments played right
Lex Fridman (57:41.160)
and it's beautiful.
Rosalind Picard (57:42.680)
There's something in you that comes alive
Lex Fridman (57:45.200)
and knows that there's a truth there
Rosalind Picard (57:47.240)
that it's like your strings are being plucked by the master
Lex Fridman (57:50.160)
instead of by me, right, playing when I pluck it.
Lex Fridman (57:54.240)
But probably when you play, it sounds spectacular, right?
Lex Fridman (57:57.400)
And when you encounter those truths,
Rosalind Picard (58:01.520)
there's something in you that sings
Lex Fridman (58:03.240)
and knows that there is more
Rosalind Picard (58:06.320)
than what I can prove mathematically
Lex Fridman (58:09.200)
or program a computer to do.
Rosalind Picard (58:11.280)
Don't get me wrong, the math is gorgeous.
Lex Fridman (58:13.360)
The computer programming can be brilliant.
Lex Fridman (58:16.040)
It's inspiring, right?
Lex Fridman (58:17.200)
We wanna do more.
Rosalind Picard (58:19.200)
None of this squashes my desire to do science
Lex Fridman (58:21.840)
or to get knowledge through science.
Rosalind Picard (58:23.640)
I'm not dissing the science at all.
Lex Fridman (58:26.040)
I grow even more in awe of what the science can do
Rosalind Picard (58:29.560)
because I'm more in awe of all there is we don't know.
Lex Fridman (58:33.000)
And really at the heart of science,
Rosalind Picard (58:36.040)
you have to have a belief that there's truth,
Lex Fridman (58:38.200)
that there's something greater to be discovered.
Lex Fridman (58:41.080)
And some scientists may not wanna use the faith word,
Lex Fridman (58:44.480)
but it's faith that drives us to do science.
Rosalind Picard (58:47.800)
It's faith that there is truth,
Lex Fridman (58:49.920)
that there's something to know that we don't know,
Rosalind Picard (58:52.520)
that it's worth knowing, that it's worth working hard,
Lex Fridman (58:56.000)
and that there is meaning,
Rosalind Picard (58:58.320)
that there is such a thing as meaning,
Lex Fridman (58:59.880)
which by the way, science can't prove either.
Rosalind Picard (59:02.720)
We have to kind of start with some assumptions
Lex Fridman (59:04.760)
that there's things like truth and meaning.
Lex Fridman (59:06.880)
And these are really questions philosophers own, right?
Lex Fridman (59:10.680)
This is their space,
Rosalind Picard (59:11.760)
of philosophers and theologians at some level.
Lex Fridman (59:14.560)
So these are things science,
Rosalind Picard (59:19.160)
when people claim that science will tell you all truth,
Lex Fridman (59:23.000)
there's a name for that.
Rosalind Picard (59:23.920)
It's its own kind of faith.
Lex Fridman (59:25.600)
It's scientism and it's very myopic.
Rosalind Picard (59:29.000)
Yeah, there's a much bigger world out there to be explored
Lex Fridman (59:32.360)
in ways that science may not,
Rosalind Picard (59:34.800)
at least for now, allow us to explore.
Lex Fridman (59:37.120)
Yeah, and there's meaning and purpose and hope
Lex Fridman (59:40.080)
and joy and love and all these awesome things
Lex Fridman (59:43.240)
that make it all worthwhile too.
Rosalind Picard (59:45.480)
I don't think there's a better way to end it, Roz.
Lex Fridman (59:47.680)
Thank you so much for talking today.
Rosalind Picard (59:49.040)
Thanks Lex, what a pleasure.
Lex Fridman (59:50.360)
Great questions.
🔗 相关节目