Rana el Kaliouby: Emotion AI, Social Robots, and Self-Driving Cars

00:00:00.000 | there's a broader question here, right?

00:00:02.120 | As we build socially and emotionally intelligent machines,

00:00:07.120 | what does that mean about our relationship with them?

00:00:10.840 | And then more broadly,

00:00:12.000 | our relationship with one another, right?

00:00:13.680 | Because this machine is gonna be programmed

00:00:15.760 | to be amazing at empathy by definition, right?

00:00:20.160 | It's gonna always be there for you.

00:00:21.600 | It's not gonna get bored.

00:00:23.520 | I don't know how I feel about that.

00:00:24.940 | I think about that a lot.

00:00:28.280 | - The following is a conversation with Rana El-Khlyoubi,

00:00:31.720 | a pioneer in the field of emotion recognition

00:00:34.800 | and human-centric artificial intelligence.

00:00:37.400 | She is the founder of Effectiva,

00:00:39.480 | deputy CEO of SmartEye, author of "Girl Decoded,"

00:00:44.120 | and one of the most brilliant, kind, inspiring,

00:00:46.820 | and fun human beings I've gotten the chance to talk to.

00:00:50.620 | This is the Lex Friedman Podcast.

00:00:52.800 | To support it,

00:00:53.640 | please check out our sponsors in the description.

00:00:56.160 | And now, dear friends, here's Rana El-Khlyoubi.

00:00:59.800 | You grew up in the Middle East, in Egypt.

00:01:03.520 | What is a memory from that time that makes you smile?

00:01:06.480 | Or maybe a memory that stands out

00:01:08.160 | as helping your mind take shape

00:01:10.720 | and helping you define yourself in this world?

00:01:12.840 | - So the memory that stands out

00:01:14.380 | is we used to live in my grandma's house.

00:01:17.480 | She used to have these mango trees in her garden

00:01:21.000 | and in the summer,

00:01:21.840 | and so mango season was like July and August.

00:01:24.080 | And so in the summer,

00:01:25.440 | she would invite all my aunts and uncles and cousins.

00:01:28.360 | And it was just like maybe there were like 20 or 30 people

00:01:31.200 | in the house and she would cook all this amazing food.

00:01:34.280 | And us, the kids, we would go down the garden

00:01:38.320 | and we would pick all these mangoes.

00:01:40.200 | And I don't know,

00:01:42.220 | I think it's just the bringing people together,

00:01:44.720 | like that always stuck with me, the warmth.

00:01:47.160 | - Around the mango tree.

00:01:48.360 | - Yeah, around the mango tree.

00:01:49.600 | And there's just like the joy,

00:01:51.720 | the joy of being together around food.

00:01:54.320 | And I'm a terrible cook,

00:01:57.040 | so I guess that didn't,

00:01:58.240 | that memory didn't translate to me kind of doing the same.

00:02:02.120 | I love hosting people.

00:02:03.760 | - Do you remember colors, smells?

00:02:05.720 | Is that what, like what, how does memory work?

00:02:08.680 | - Yeah.

00:02:09.520 | - Like what do you visualize?

00:02:10.340 | Do you visualize people's faces, smiles?

00:02:13.320 | Do you, is there colors?

00:02:15.120 | Is there like a theme to the colors?

00:02:18.360 | Is it smells because of food involved?

00:02:21.080 | - Yeah, I think that's a great question.

00:02:22.880 | So those Egyptian mangoes,

00:02:25.240 | there's a particular type that I love

00:02:27.480 | and it's called Darwesi mangoes.

00:02:28.840 | And they're kind of, you know, they're oval

00:02:31.200 | and they have a little red in them.

00:02:33.240 | So I kind of, they're red and mango colored on the outside.

00:02:36.920 | So I remember that.

00:02:38.500 | - Does red indicate like extra sweetness?

00:02:40.840 | Is that? - Yes.

00:02:41.960 | - That means like it's nicely--

00:02:42.800 | - It's like really sweet.

00:02:44.080 | - Yeah, it's nice and ripe and stuff, yeah.

00:02:46.240 | What's like a definitive food of Egypt?

00:02:50.780 | You know, there's like these almost stereotypical foods

00:02:54.000 | in different parts of the world,

00:02:55.120 | like Ukraine invented borscht.

00:02:59.800 | Borscht is this beet soup with,

00:03:02.000 | that you put sour cream on.

00:03:03.200 | See, it's not, I can't see--

00:03:04.040 | - Okay, okay, well you explained it that way.

00:03:05.800 | - If you know what it is, I think you know it's delicious.

00:03:09.700 | But if I explain it, it's just not gonna sound delicious.

00:03:12.540 | I feel like beet soup, this doesn't make any sense.

00:03:15.440 | But that's kind of,

00:03:16.520 | and you probably have actually seen pictures of it

00:03:19.020 | 'cause it's one of the traditional foods in Ukraine,

00:03:22.640 | in Russia, in different parts of the Slavic world.

00:03:26.080 | So, but it's become so cliche and stereotypical

00:03:29.860 | that you almost don't mention it,

00:03:30.960 | but it's still delicious.

00:03:32.520 | Like I visited Ukraine, I eat that every single day.

00:03:35.600 | - Do you make it yourself?

00:03:37.680 | How hard is it to make?

00:03:38.640 | - No, I don't know.

00:03:39.560 | I think to make it well, like anything, like Italians,

00:03:43.160 | they say, well, tomato sauce is easy to make,

00:03:46.520 | but to make it right,

00:03:48.960 | that's like a generational skill.

00:03:51.160 | So anyway, is there something like that in Egypt?

00:03:53.480 | Is there a culture of food?

00:03:55.360 | - There is, and actually we have a similar kind of soup.

00:03:59.120 | It's called molokhia,

00:04:00.960 | and it's made of this green plant.

00:04:04.520 | It's like somewhere between spinach and kale,

00:04:06.920 | and you mince it, and then you cook it in like chicken broth.

00:04:10.600 | And my grandma used to make,

00:04:12.080 | and my mom makes it really well,

00:04:14.000 | and I try to make it, but it's not as great.

00:04:16.520 | So we used to have that,

00:04:17.480 | and then we used to have it alongside stuffed pigeons.

00:04:20.320 | I'm pescetarian now, so I don't eat that anymore, but.

00:04:23.560 | - Stuffed pigeons.

00:04:24.600 | - Yeah, it's like, it was really yummy.

00:04:25.960 | It's the one thing I miss about, you know,

00:04:29.160 | now that I'm pescetarian and I don't eat.

00:04:32.160 | - The stuffed pigeons?

00:04:33.000 | - Yeah, the stuffed pigeons.

00:04:34.680 | (both laughing)

00:04:35.520 | - Is it, what are they stuffed with?

00:04:37.200 | If that doesn't bother you too much to describe.

00:04:39.960 | - No, no, it's stuffed with a lot of like just rice and.

00:04:43.480 | - Oh, got it, got it, got it.

00:04:44.320 | - Yeah, it's just rice, yeah, so.

00:04:46.120 | - And you also, you've said that you're first in your book

00:04:49.920 | that your first computer was an Atari,

00:04:52.720 | and Space Invaders was your favorite game.

00:04:55.200 | Is that when you first fell in love with computers?

00:04:58.200 | Would you say?

00:04:59.040 | - Yeah, I would say so.

00:05:00.280 | - Video games or just the computer itself?

00:05:02.280 | Just something about the machine.

00:05:04.300 | Ooh, this thing.

00:05:05.240 | There's magic in here.

00:05:07.840 | - Yeah, I think the magical moment is definitely

00:05:10.600 | like playing video games with my,

00:05:12.160 | I have two younger sisters,

00:05:13.640 | and we just like had fun together, like playing games.

00:05:17.240 | But the other memory I have is my first code.

00:05:21.320 | The first code I wrote, I wrote, I drew a Christmas tree.

00:05:25.760 | And I'm Muslim, right?

00:05:26.840 | So it's kind of, it was kind of funny

00:05:28.880 | that the first thing I did was like this Christmas tree.

00:05:32.040 | So yeah, and that's when I realized,

00:05:35.760 | wow, you can write code to do all sorts

00:05:38.040 | of like really cool stuff.

00:05:40.600 | I must have been like six or seven at the time.

00:05:43.360 | So you can write programs

00:05:45.760 | and the programs do stuff for you.

00:05:48.080 | That's power.

00:05:49.440 | That's, if you think about it, that's empowering.

00:05:51.520 | - It's AI.

00:05:52.360 | - Yeah, I know, well, it is.

00:05:53.960 | I don't know what that, you see, like,

00:05:55.920 | I don't know if many people think of it that way

00:05:59.000 | when they first learned to program.

00:06:00.200 | They just love the puzzle of it.

00:06:01.500 | Like, ooh, this is cool, this is pretty.

00:06:03.600 | It's a Christmas tree, but like, it's power.

00:06:06.320 | - It is power.

00:06:07.880 | - Eventually, I guess you couldn't at the time,

00:06:10.040 | but eventually this thing, if it's interesting enough,

00:06:13.020 | if it's a pretty enough Christmas tree,

00:06:14.800 | it can be run by millions of people

00:06:16.720 | and bring them joy, like that little thing.

00:06:19.360 | And then because it's digital, it's easy to spread.

00:06:22.480 | So like you just created something

00:06:24.120 | that's easily spreadable to millions of people.

00:06:26.800 | - Totally.

00:06:28.280 | - It's hard to think that way when you're six.

00:06:30.920 | In the book you write, "I am who I am

00:06:34.080 | because I was raised by a particular set of parents,

00:06:37.160 | both modern and conservative,

00:06:39.160 | forward-thinking yet locked in tradition.

00:06:41.920 | I'm a Muslim and I feel I'm stronger, more centered for it.

00:06:46.200 | I adhere to the values of my religion,

00:06:48.080 | even if I'm not as dutiful as I once was.

00:06:51.040 | And I am a new American and I'm thriving

00:06:54.500 | on the energy, vitality, and entrepreneurial spirit

00:06:57.800 | of this great country."

00:06:59.080 | So let me ask you about your parents.

00:07:01.640 | What have you learned about life from them,

00:07:04.040 | especially when you were young?

00:07:05.360 | - So both my parents, they're Egyptian,

00:07:07.560 | but they moved to Kuwait right after.

00:07:09.840 | They actually, there's a cute story about how they met.

00:07:11.720 | So my dad taught kobol in the '70s.

00:07:15.040 | - Nice.

00:07:15.880 | - And my mom decided to learn programming.

00:07:18.360 | So she signed up to take his kobol programming class.

00:07:21.480 | And he tried to date her and she was like,

00:07:25.620 | "No, no, no, I don't date."

00:07:26.720 | And so he's like, "Okay, I'll propose."

00:07:28.360 | And that's how they got married.

00:07:29.840 | - Whoa, strong move.

00:07:31.120 | - Right, exactly right.

00:07:32.580 | - That's really impressive.

00:07:35.960 | Those kobol guys know how to impress a lady.

00:07:40.800 | So yeah, so what have you learned from them?

00:07:43.760 | - So definitely grit.

00:07:44.920 | One of the core values in our family is just hard work.

00:07:48.600 | There were no slackers in our family.

00:07:50.920 | And that's something that's definitely stayed with me,

00:07:55.200 | both as a professional, but also in my personal life.

00:07:58.940 | But I also think my mom, my mom always used to,

00:08:02.880 | I don't know, it was like unconditional love.

00:08:07.120 | I just knew my parents would be there for me,

00:08:10.280 | kind of regardless of what I chose to do.

00:08:12.580 | And I think that's very powerful.

00:08:15.680 | And they got tested on it because I kind of challenged,

00:08:18.160 | you know, I challenged cultural norms

00:08:20.320 | and I kind of took a different path, I guess,

00:08:23.080 | than what's expected of a woman in the Middle East.

00:08:27.520 | And they still love me, which is,

00:08:31.420 | I'm so grateful for that.

00:08:32.640 | - When was like a moment

00:08:33.760 | that was the most challenging for them?

00:08:35.600 | Which moment were they kind of,

00:08:39.600 | they had to come face to face with the fact

00:08:41.920 | that you're a bit of a rebel?

00:08:44.240 | - I think the first big moment was when I,

00:08:47.380 | I had just gotten married,

00:08:50.400 | but I decided to go do my PhD at Cambridge University.

00:08:54.120 | And because my husband at the time, he's now my ex,

00:08:58.860 | ran a company in Cairo, he was gonna stay in Egypt.

00:09:01.460 | So it was gonna be a long distance relationship.

00:09:04.200 | And that's very unusual in the Middle East

00:09:06.840 | for a woman to just head out and kind of pursue her career.

00:09:10.960 | And so my dad and my parents-in-law both said,

00:09:15.960 | "You know, we do not approve of you doing this,

00:09:19.840 | but now you're under the jurisdiction of your husband,

00:09:22.320 | so he can make the call."

00:09:24.720 | And luckily for me, he was supportive.

00:09:28.760 | He said, "You know, this is your dream come true.

00:09:31.640 | You've always wanted to do a PhD.

00:09:33.080 | I'm gonna support you."

00:09:35.400 | So I think that was the first time where,

00:09:37.920 | you know, I challenged the cultural norms.

00:09:41.200 | - Was that scary?

00:09:42.400 | - Oh my God, yes.

00:09:43.520 | It was totally scary.

00:09:44.880 | - What's the biggest culture shock

00:09:46.480 | from there to Cambridge, to London?

00:09:51.480 | - Well, that was also during, right around September 11th.

00:09:56.500 | So everyone thought that there was gonna be

00:09:59.360 | a third world war, where it was really,

00:10:03.280 | and at the time I used to wear the hijab,

00:10:05.600 | so I was very visibly Muslim.

00:10:07.920 | And so my parents just were, they were afraid for my safety.

00:10:12.080 | But anyways, when I got to Cambridge,

00:10:13.440 | because I was so scared,

00:10:14.440 | I decided to take off my headscarf and wear a hat instead.

00:10:17.680 | So I just went to class wearing these like British hats,

00:10:20.400 | which was, in my opinion, actually worse

00:10:22.840 | than just showing up in a headscarf,

00:10:24.480 | 'cause it was just so awkward, right?

00:10:26.080 | Like sitting in class with like all these-

00:10:27.760 | - Trying to fit in, like a spy.

00:10:30.720 | - Yeah, yeah, yeah.

00:10:31.560 | So after a few weeks of doing that,

00:10:33.160 | I was like, to heck with that,

00:10:34.840 | I'm just gonna go back to wearing my headscarf.

00:10:37.320 | - Yeah, you wore the hijab, so starting in 2000,

00:10:42.320 | and for 12 years after.

00:10:44.300 | So always, whenever you're in public,

00:10:47.000 | you have to wear the head covering.

00:10:48.920 | Can you speak to that, to the hijab,

00:10:52.320 | maybe your mixed feelings about it?

00:10:54.120 | Like what does it represent in its best case?

00:10:56.640 | What does it represent in the worst case?

00:10:58.400 | - Yeah, you know, I think there's a lot of,

00:11:02.520 | I guess I'll first start by saying I wore it voluntarily.

00:11:05.520 | I was not forced to wear it.

00:11:06.960 | And in fact, I was one of the very first women in my family

00:11:10.040 | to decide to put on the hijab.

00:11:12.240 | And my family thought it was really odd, right?

00:11:15.240 | Like they were like, why do you wanna put this on?

00:11:17.920 | And at its best, it's a sign of modesty, humility.

00:11:22.920 | - It's like me wearing a suit.

00:11:25.720 | People are like, why are you wearing a suit?

00:11:27.560 | It's a step back into some kind of tradition,

00:11:30.920 | a respect for tradition of sorts.

00:11:33.240 | So you said because it's by choice,

00:11:34.960 | you're kind of free to make that choice

00:11:37.560 | to celebrate a tradition of modesty.

00:11:40.000 | - Exactly, and I actually like made it my own.

00:11:43.280 | I remember I would really match the color of my headscarf

00:11:46.880 | with what I was wearing.

00:11:47.800 | Like it was a form of self-expression,

00:11:50.120 | and at its best, I loved wearing it.

00:11:54.160 | You know, I have a lot of questions

00:11:55.800 | around how we practice religion and religion.

00:11:58.200 | And I think also it was a time

00:12:02.480 | where I was spending a lot of time going back and forth

00:12:04.720 | between the US and Egypt.

00:12:06.640 | And I started meeting a lot of people in the US

00:12:08.560 | who were just amazing people, very purpose-driven,

00:12:13.360 | people who have very strong core values,

00:12:15.520 | but they're not Muslim.

00:12:16.920 | That's okay, right?

00:12:17.880 | And so that was when I just had a lot of questions.

00:12:21.220 | And politically also the situation in Egypt

00:12:25.120 | was when the Muslim Brotherhood ran the country,

00:12:27.240 | and I didn't agree with their ideology.

00:12:30.160 | It was at a time when I was going through a divorce.

00:12:33.560 | Like it was like just the perfect storm

00:12:36.040 | of like political, personal conditions,

00:12:39.300 | where I was like, "This doesn't feel like me anymore."

00:12:42.240 | And it took a lot of courage to take it off

00:12:44.240 | because culturally it's okay if you don't wear it,

00:12:48.600 | but it's really not okay to wear it and then take it off.

00:12:52.080 | - But you're still, so you had to do that

00:12:54.440 | while still maintaining a deep core and pride

00:12:58.040 | in the origins, in your origin story.

00:13:02.440 | - Totally.

00:13:03.280 | - So still being Egyptian, still being a Muslim.

00:13:06.760 | - Right, and being, I think, generally like faith-driven,

00:13:11.760 | but yeah.

00:13:14.440 | - But what that means changes year by year for you.

00:13:17.600 | It's like a personal journey.

00:13:18.920 | - Yeah, exactly.

00:13:20.080 | - What would you say is the role of faith

00:13:22.160 | in that part of the world?

00:13:24.080 | Like how do you see it?

00:13:25.120 | You mention it a bit in the book too.

00:13:27.440 | - Yeah, I mean, I think there is something really powerful

00:13:32.120 | about just believing that there's a bigger force.

00:13:36.880 | You know, there's a kind of surrendering, I guess,

00:13:39.460 | that comes with religion and you surrender

00:13:42.120 | and you have this deep conviction

00:13:43.400 | that it's gonna be okay, right?

00:13:45.380 | Like the universe is out to like do amazing things for you

00:13:48.000 | and it's gonna be okay.

00:13:49.600 | And there's strength to that.

00:13:50.960 | Like even when you're going through adversity,

00:13:54.060 | you just know that it's gonna work out.

00:13:57.940 | - Yeah, it gives you like an inner peace, a calmness.

00:14:00.100 | - Exactly, exactly.

00:14:01.460 | - Yeah, it's faith in all the meanings of that word.

00:14:05.700 | - Right.

00:14:06.540 | - Faith that everything is going to be okay.

00:14:08.100 | And it is because time passes and time cures all things.

00:14:13.100 | It's like a calmness with the chaos of the world.

00:14:16.420 | - And also there's like a silver lining.

00:14:19.460 | I'm a true believer of this,

00:14:20.700 | that something at the specific moment in time

00:14:24.020 | can look like it's catastrophic

00:14:25.460 | and it's not what you wanted in life, da-da-da-da.

00:14:28.060 | But then time passes and then you look back

00:14:31.300 | and there's a silver lining, right?

00:14:33.060 | It maybe closed the door, but it opened a new door for you.

00:14:37.340 | And so I'm a true believer in that,

00:14:38.940 | that there's a silver lining in almost anything in life.

00:14:43.860 | You just have to have this like,

00:14:45.860 | have faith or conviction that it's gonna work out.

00:14:48.380 | - Such a beautiful way to see a shitty feeling.

00:14:50.900 | So if you feel shitty about a current situation,

00:14:53.840 | I mean, it almost is always true,

00:14:56.820 | unless it's the cliches thing of,

00:15:02.780 | if it doesn't kill you,

00:15:03.940 | whatever doesn't kill you makes you stronger.

00:15:06.180 | It does seem that over time,

00:15:09.220 | when you take a perspective on things,

00:15:11.120 | the hardest moments and periods of your life

00:15:15.780 | are the most meaningful.

00:15:18.200 | - Yeah, yeah.

00:15:19.440 | So over time you get to have that perspective.

00:15:21.680 | - Right.

00:15:22.520 | - What about, 'cause you mentioned Kuwait,

00:15:26.900 | what about, let me ask you about war.

00:15:30.380 | What's the role of war and peace,

00:15:34.040 | maybe even the big love and hate in that part of the world,

00:15:37.680 | because it does seem to be a part of the world

00:15:39.720 | where there's turmoil.

00:15:41.800 | There was turmoil, there's still turmoil.

00:15:45.640 | - It is so unfortunate, honestly.

00:15:47.600 | It's such a waste of human resources

00:15:51.000 | and yeah, and human mindshare.

00:15:54.800 | I mean, at the end of the day,

00:15:56.680 | we all kind of want the same things.

00:15:58.360 | We want human connection, we want joy,

00:16:02.160 | we wanna feel fulfilled,

00:16:03.440 | we wanna feel a life of purpose.

00:16:06.680 | And I just find it baffling, honestly,

00:16:11.200 | that we are still having to grapple with that.

00:16:15.240 | I have a story to share about this.

00:16:17.200 | I grew up, I'm Egyptian, American now,

00:16:19.400 | but originally from Egypt.

00:16:22.920 | And when I first got to Cambridge,

00:16:24.800 | it turned out my office mate,

00:16:26.920 | like my PhD kind of,

00:16:28.480 | we ended up becoming friends, but she was from Israel.

00:16:34.080 | And we didn't know, yeah,

00:16:35.560 | we didn't know how it was gonna be like.

00:16:37.900 | - Did you guys sit there just staring at each other for a bit?

00:16:42.520 | - Actually, she, 'cause I arrived before she did

00:16:45.640 | and it turns out she emailed our PhD advisor

00:16:49.800 | and asked him if she thought it was gonna be okay.

00:16:52.400 | - Yeah.

00:16:53.240 | Oh, this is around 9/11 too.

00:16:55.120 | - Yeah, and Peter Robinson, our PhD advisor was like,

00:16:59.520 | yeah, like this is an academic institution, just show up.

00:17:02.960 | And we became super good friends.

00:17:04.640 | We were both new moms.

00:17:07.360 | Like we both had our kids during our PhD.

00:17:09.400 | We were both doing artificial emotional intelligence.

00:17:11.520 | She was looking at speech.

00:17:12.460 | I was looking at the face.

00:17:13.880 | We just had so, the culture was so similar.

00:17:17.160 | Our jokes were similar.

00:17:18.400 | It was just, I was like, why on earth are our countries,

00:17:23.080 | why is there all this like war and tension?

00:17:25.320 | And I think it falls back to the narrative, right?

00:17:27.480 | If you change the narrative,

00:17:28.780 | like whoever creates this narrative of war, I don't know.

00:17:32.600 | We should have women run the world.

00:17:34.800 | - Yeah, that's one solution, the good women,

00:17:38.600 | because there's also evil women in the world.

00:17:40.720 | - True, okay.

00:17:41.940 | (both laughing)

00:17:44.060 | - But yes, yes, there could be less war

00:17:46.300 | if women ran the world.

00:17:48.220 | The other aspect is, it doesn't matter the gender,

00:17:51.860 | the people in power.

00:17:53.540 | I get to see this with Ukraine and Russia,

00:17:57.100 | different parts of the world around that conflict now.

00:18:01.020 | And that's happening in Yemen as well and everywhere else.

00:18:05.340 | There's these narratives told by the leaders

00:18:08.580 | to the populace and those narratives take hold

00:18:11.340 | and everybody believes that and they have a distorted view

00:18:15.660 | of the humanity on the other side.

00:18:18.020 | In fact, especially during war,

00:18:20.020 | you don't even see the people on the other side as human

00:18:25.020 | or as equal intelligence or worth or value as you.

00:18:30.220 | You tell all kinds of narratives about them being Nazis

00:18:34.180 | or Dom or whatever narrative you wanna weave around that.

00:18:41.900 | Or evil.

00:18:42.860 | But I think when you actually meet them face to face,

00:18:47.420 | you realize they're the same.

00:18:49.220 | - Exactly right.

00:18:50.500 | - It's actually a big shock for people to realize

00:18:53.860 | that they've been essentially lied to within their country.

00:18:59.980 | And I kind of have faith that social media,

00:19:03.340 | as ridiculous as it is to say,

00:19:05.140 | or any kind of technology is able to bypass the walls

00:19:10.340 | that governments put up and connect people directly

00:19:14.140 | and then you get to realize,

00:19:15.220 | ooh, people fall in love across different nations

00:19:20.020 | and religions and so on.

00:19:21.540 | And that I think ultimately can cure a lot of our ills,

00:19:24.660 | especially sort of in person.

00:19:26.780 | I also think that if leaders met in person

00:19:30.020 | to have a conversation,

00:19:31.020 | that would cure a lot of the ills of the world,

00:19:34.860 | especially in private.

00:19:37.840 | - Let me ask you about the women running the world.

00:19:41.560 | - Okay.

00:19:42.440 | - So gender does in part,

00:19:44.920 | perhaps shape the landscape of just our human experience.

00:19:49.800 | So in what ways was it limiting it?

00:19:53.920 | In what ways was it empowering

00:19:56.760 | for you to be a woman in the Middle East?

00:19:59.640 | - I think just kind of just going back to like my comment

00:20:02.160 | on like women running the world,

00:20:03.240 | I think it comes back to empathy, right?

00:20:05.360 | Which has been a common thread throughout my entire career.

00:20:08.960 | And it's this idea of human connection.

00:20:11.520 | Once you build common ground with a person

00:20:14.760 | or a group of people, you build trust, you build loyalty,

00:20:17.760 | you build friendship, and then you can turn that

00:20:21.840 | into like behavior change and motivation and persuasion.

00:20:25.120 | So it's like empathy and emotions are just at the center

00:20:28.000 | of everything we do.

00:20:31.480 | And I think being from the Middle East,

00:20:35.480 | kind of this human connection is very strong.

00:20:38.640 | Like we have this running joke that if you come to Egypt

00:20:42.120 | for a visit, people are gonna,

00:20:44.600 | will know everything about your life like right away, right?

00:20:46.720 | I have no problems asking you about your personal life.

00:20:49.520 | There's no like no boundaries really,

00:20:52.720 | no personal boundaries in terms of getting to know people.

00:20:55.240 | We get emotionally intimate like very, very quickly.

00:20:58.480 | But I think people just get to know each other

00:21:00.440 | like authentically, I guess.

00:21:03.720 | There isn't this like superficial level

00:21:06.120 | of getting to know people.

00:21:07.040 | You just try to get to know people really deeply.

00:21:09.240 | - And empathy is a part of that.

00:21:10.080 | - Totally, 'cause you can put yourself

00:21:11.720 | in this person's shoe and kind of,

00:21:14.280 | yeah, imagine what challenges they're going through.

00:21:18.720 | So I think I've definitely taken that with me.

00:21:22.480 | Generosity is another one too,

00:21:25.440 | like just being generous with your time and love

00:21:28.280 | and attention and even with your wealth, right?

00:21:32.600 | Even if you don't have a lot of it,

00:21:33.840 | you're still very generous.

00:21:34.920 | And I think that's another.

00:21:36.720 | - Enjoying the humanity of other people.

00:21:40.120 | And so do you think there's a useful difference

00:21:42.920 | between men and women in that aspect and empathy?

00:21:47.640 | Or is doing these kind of big general groups,

00:21:53.720 | does that hinder progress?

00:21:57.040 | - Yeah, I actually don't wanna overgeneralize.

00:21:59.920 | I mean, some of the men I know

00:22:01.960 | are like the most empathetic humans.

00:22:03.720 | - Yeah, I strive to be empathetic.

00:22:05.360 | - Yeah, you're actually very empathetic.

00:22:07.480 | Yeah, so I don't wanna overgeneralize.

00:22:13.880 | Although one of the researchers I worked with

00:22:17.160 | when I was at Cambridge, Professor Simon Baring-Cohen,

00:22:19.720 | he's Sasha Baring-Cohen's cousin.

00:22:22.040 | - Yeah. (laughs)

00:22:23.520 | - But he runs the Autism Research Center at Cambridge

00:22:26.360 | and he's written multiple books on autism.

00:22:30.880 | And one of his theories is the empathy scale,

00:22:34.360 | like the systemizers and the empathizers.

00:22:36.320 | And there's a disproportionate amount

00:22:40.040 | of computer scientists and engineers who are systemizers

00:22:44.520 | and perhaps not great empathizers.

00:22:47.200 | And then there's more men in that bucket, I guess,

00:22:52.880 | than women.

00:22:53.720 | And then there's more women in the empathizers bucket.

00:22:56.840 | So again, not to overgeneralize.

00:22:58.920 | - I sometimes wonder about that.

00:23:00.400 | It's been frustrating to me how many, I guess,

00:23:03.320 | systemizers there are in the field of robotics.

00:23:06.040 | - Yeah.

00:23:07.040 | - It's actually encouraging to me

00:23:08.280 | 'cause I care about, obviously, social robotics.

00:23:11.240 | And because there's more opportunity

00:23:16.240 | for people that are empathic. (laughs)

00:23:18.920 | - Exactly, I totally agree.

00:23:20.720 | Well, right? - So it's nice.

00:23:21.920 | - Yes.

00:23:22.760 | - I mean, most robotics I talk to,

00:23:23.720 | they don't see the human as interesting,

00:23:26.080 | as like it's not exciting.

00:23:29.400 | You wanna avoid the human at all costs.

00:23:32.200 | It's a safety concern to be touching the human,

00:23:35.360 | which it is, but it's also an opportunity

00:23:39.360 | for deep connection or collaboration

00:23:42.160 | or all that kind of stuff.

00:23:43.000 | So, and because most brilliant roboticists

00:23:46.280 | don't care about the human, it's an opportunity.

00:23:48.600 | - Right.

00:23:49.440 | - For, in your case, it's a business opportunity too,

00:23:52.160 | but in general, an opportunity to explore those ideas.

00:23:54.720 | So, in this beautiful journey to Cambridge,

00:23:58.200 | to UK, and then to America,

00:24:01.880 | what's the moment or moments

00:24:04.760 | that were most transformational for you

00:24:07.280 | as a scientist and as a leader?

00:24:09.960 | So you became an exceptionally successful CEO,

00:24:12.920 | founder, researcher, scientist, and so on.

00:24:16.680 | Was there a phase shift there

00:24:21.240 | where like, I can be somebody,

00:24:23.800 | I can really do something in this world?

00:24:26.800 | - Yeah, so actually just kind of a little bit of background.

00:24:29.800 | So the reason why I moved from Cairo to Cambridge, UK

00:24:33.960 | to do my PhD is because I had a very clear career plan.

00:24:38.200 | I was like, okay, I'll go abroad, get my PhD,

00:24:41.320 | gonna crush it in three or four years,

00:24:43.440 | come back to Egypt and teach.

00:24:45.720 | It was very clear, very well laid out.

00:24:47.760 | - Was topic clear or no?

00:24:50.480 | - Well, I did my PhD around

00:24:52.480 | building artificial emotional intelligence and looking at--

00:24:54.600 | - No, but in your master plan ahead of time,

00:24:56.960 | when you're sitting by the mango tree,

00:24:58.320 | did you know it's gonna be artificial intelligence?

00:25:00.640 | - No, no, no, that I did not know.

00:25:03.080 | Although I think I kinda knew

00:25:05.400 | that I was gonna be doing computer science,

00:25:07.720 | but I didn't know the specific area.

00:25:10.400 | But I love teaching, I mean, I still love teaching.

00:25:13.200 | So I just, yeah, I just wanted to go abroad,

00:25:16.520 | get a PhD, come back, teach.

00:25:19.080 | - Why computer science?

00:25:19.960 | Can we just linger on that?

00:25:21.680 | 'Cause you're such an empathic person

00:25:23.600 | who cares about emotion, humans and so on.

00:25:25.840 | Aren't computers cold and emotionless?

00:25:30.600 | (laughing)

00:25:31.840 | Just-- - We're changing that.

00:25:33.240 | - Yeah, I know, but isn't that the,

00:25:36.280 | or did you see computers as having the capability

00:25:39.840 | to actually connect with humans?

00:25:43.120 | - I think that was my takeaway

00:25:44.720 | from my experience just growing up.

00:25:46.520 | Computers sit at the center of how we connect

00:25:49.120 | and communicate with one another, right?

00:25:51.080 | Or technology in general.

00:25:52.480 | Like I remember my first experience

00:25:54.320 | being away from my parents.

00:25:55.400 | We communicated with a fax machine,

00:25:57.320 | but thank goodness for the fax machine

00:25:59.200 | because we could send letters back and forth to each other.

00:26:01.560 | This was pre-emails and stuff.

00:26:03.360 | So I think there's, I think technology

00:26:08.200 | can be not just transformative

00:26:09.880 | in terms of productivity, et cetera.

00:26:11.760 | It actually does change how we connect with one another.

00:26:15.720 | - Can I just defend the fax machine?

00:26:17.520 | - Yeah. - There's something,

00:26:19.320 | like the haptic feel, 'cause the email is all digital.

00:26:23.480 | There's something really nice.

00:26:24.600 | I still write letters to people.

00:26:27.320 | There's something nice about the haptic aspect

00:26:29.600 | of the fax machine 'cause you still have to press,

00:26:31.840 | you still have to do something in the physical world

00:26:34.240 | to make this thing a reality, the sense of somebody.

00:26:36.960 | - Right, and then it comes out as a printout

00:26:38.960 | and you can actually touch it and read it.

00:26:40.720 | - Yeah, there's something lost when it's just an email.

00:26:45.720 | Obviously, I wonder how we can regain some of that

00:26:50.440 | in the digital world, which goes to the metaverse

00:26:53.000 | and all those kinds of things.

00:26:53.840 | We'll talk about it.

00:26:54.680 | Anyway, so--

00:26:55.520 | - Actually, do you have a question on that one?

00:26:57.800 | Do you still, do you have photo albums anymore?

00:27:00.520 | Do you still print photos?

00:27:02.400 | - No, no, but I'm a minimalist.

00:27:06.720 | So it was one of the painful steps in my life

00:27:09.360 | was to scan all the photos and let go of them

00:27:13.360 | and then let go of all my books.

00:27:16.520 | - You let go of your books?

00:27:17.760 | - Yeah, switched to Kindle, everything Kindle.

00:27:20.840 | So I thought, okay, think 30 years from now.

00:27:25.840 | Nobody's gonna have books anymore.

00:27:29.640 | The technology of digital books

00:27:31.040 | is gonna get better and better and better.

00:27:32.200 | Are you really gonna be the guy

00:27:33.720 | that's still romanticizing physical books?

00:27:36.400 | Are you gonna be the old man on the porch

00:27:38.800 | who's like kids, yes.

00:27:40.520 | So just get used to it 'cause it felt,

00:27:43.040 | it still feels a little bit uncomfortable

00:27:45.120 | to read on a Kindle, but get used to it.

00:27:48.840 | You always, I mean, I'm trying to learn

00:27:51.520 | new programming languages always.

00:27:53.440 | Like with technology, you have to kind of

00:27:55.000 | challenge yourself to adapt to it.

00:27:57.280 | I forced myself to use TikTok now.

00:28:00.000 | That thing doesn't need much forcing.

00:28:01.560 | It pulls you in like the worst kind of,

00:28:04.560 | or the best kind of drug.

00:28:06.040 | Anyway, yeah, so yeah, but I do love haptic things.

00:28:11.920 | There's a magic to the haptic.

00:28:13.600 | Even like touchscreens, it's tricky to get right,

00:28:16.400 | to get the experience of a button.

00:28:19.640 | - Yeah.

00:28:20.480 | - Anyway, what were we talking about?

00:28:23.880 | So AI, so the journey, your whole plan

00:28:27.760 | was to come back to Cairo and teach, right?

00:28:31.120 | - And then-- - What did the plan go wrong?

00:28:33.960 | - Yeah, exactly, right?

00:28:35.200 | And then I got to Cambridge and I fall in love

00:28:37.960 | with the idea of research, right?

00:28:39.560 | And kind of embarking on a path.

00:28:41.640 | Nobody's explored this path before.

00:28:43.800 | You're building stuff that nobody's built before

00:28:45.640 | and it's challenging and it's hard

00:28:47.080 | and there's a lot of non-believers.

00:28:49.400 | I just totally love that.

00:28:51.040 | And at the end of my PhD, I think it's the meeting

00:28:54.480 | that changed the trajectory of my life.

00:28:56.960 | Professor Rosalind Picard, she runs

00:28:59.960 | the Affective Computing Group at the MIT Media Lab.

00:29:02.240 | I had read her book.

00:29:03.560 | You know, I was like following all her research.

00:29:07.680 | - AKA Roz.

00:29:08.960 | - Yes, AKA Roz.

00:29:10.920 | And she was giving a talk

00:29:13.360 | at a pattern recognition conference in Cambridge

00:29:16.480 | and she had a couple of hours to kill.

00:29:18.080 | So she emailed the lab and she said,

00:29:19.760 | you know, if any students wanna meet with me,

00:29:22.400 | like just, you know, sign up here.

00:29:24.800 | And so I signed up for slot and I spent like the weeks

00:29:28.560 | leading up to it preparing for this meeting.

00:29:31.080 | And I want to show her a demo of my research and everything.

00:29:34.480 | And we met and we ended up hitting it off.

00:29:36.760 | Like we totally clicked.

00:29:38.160 | And at the end of the meeting, she said,

00:29:40.840 | do you wanna come work with me as a postdoc at MIT?

00:29:43.360 | And this is what I told her.

00:29:45.720 | I was like, okay, this would be a dream come true,

00:29:47.760 | but there's a husband waiting for me in Cairo.

00:29:49.880 | I could have to go back.

00:29:51.240 | And she said, it's fine, just commute.

00:29:54.720 | And I literally started commuting between Cairo and Boston.

00:29:57.960 | Yeah, it was a long commute.

00:30:01.320 | And I did that like every few weeks,

00:30:03.480 | I would, you know, hop on a plane and go to Boston.

00:30:06.480 | But that changed the trajectory of my life.

00:30:08.600 | There was no, I kind of outgrew my dreams, right?

00:30:12.960 | I didn't wanna go back to Egypt anymore and be faculty.

00:30:16.800 | Like that was no longer my dream.

00:30:18.440 | I had a new dream.

00:30:19.280 | - What was it like to be at MIT?

00:30:22.720 | What was that culture shock?

00:30:25.160 | You mean America in general, but also,

00:30:27.520 | I mean Cambridge has its own culture, right?

00:30:31.240 | So what was MIT like?

00:30:33.000 | What was America like?

00:30:35.080 | - I think, I wonder if that's similar to your experience

00:30:37.960 | at MIT, I was just, at the Media Lab in particular,

00:30:42.600 | I was just really, impressed is not the right word.

00:30:47.200 | I didn't expect the openness to like innovation

00:30:50.680 | and the acceptance of taking a risk and failing.

00:30:55.680 | Like failure isn't really accepted back in Egypt, right?

00:30:59.120 | You don't wanna fail.

00:30:59.960 | Like there's a fear of failure,

00:31:02.000 | which I think has been hardwired in my brain.

00:31:05.080 | But you get to MIT and it's okay to start things.

00:31:07.240 | And if they don't work out, like it's okay,

00:31:09.240 | you pivot to another idea.

00:31:11.040 | And that kind of thinking was just very new to me.

00:31:13.840 | - That's liberating.

00:31:14.800 | Well, Media Lab for people who don't know,

00:31:16.200 | MIT Media Lab is its own beautiful thing

00:31:19.480 | because they, I think more than other places at MIT,

00:31:23.880 | reach for big ideas.

00:31:25.160 | And like they try, I mean, I think,

00:31:27.720 | I mean, depending of course on who,

00:31:29.200 | but certainly with Rosalind, you try wild stuff,

00:31:32.480 | you try big things and crazy things.

00:31:34.600 | And also try to take things to completion

00:31:38.520 | so you can demo them.

00:31:39.560 | So always have a demo.

00:31:43.480 | Like if you go, one of the sad things to me

00:31:46.000 | about robotics labs at MIT, and there's like over 30,

00:31:48.960 | I think, is like, usually when you show up to a robotics lab

00:31:53.960 | there's not a single working robot, they're all broken.

00:31:57.000 | All the robots are broken,

00:31:58.880 | which is like the normal state of things

00:32:00.680 | because you're working on them.

00:32:02.080 | But it would be nice if we lived in a world

00:32:04.360 | where robotics labs had some robots functioning.

00:32:08.960 | One of my like favorite moments that just sticks with me,

00:32:11.800 | I visited Boston Dynamics and there was a,

00:32:14.240 | first of all, seeing so many spots,

00:32:17.960 | so many legged robots in one place, I'm like, I'm home.

00:32:21.640 | (both laughing)

00:32:22.840 | But the- - My drive.

00:32:24.680 | - Yeah.

00:32:25.520 | This is where I was built.

00:32:28.880 | The cool thing was just to see,

00:32:31.160 | there was a random robot spot was walking down the hall.

00:32:35.360 | It was probably doing mapping,

00:32:36.520 | but it looked like he wasn't doing anything

00:32:38.440 | and he was wearing he or she, I don't know.

00:32:41.120 | But it, well, I like, in my mind,

00:32:44.400 | there are people that have a backstory,

00:32:46.360 | but this one in particular definitely has a backstory

00:32:48.360 | because he was wearing a cowboy hat.

00:32:51.240 | So I just saw a spot robot with a cowboy hat

00:32:54.120 | walking down the hall.

00:32:55.280 | And there was just this feeling like there's a life,

00:32:59.720 | like he has a life.

00:33:01.520 | He probably has to commute back to his family at night.

00:33:04.880 | Like there's a feeling like there's life instilled

00:33:08.200 | in this robot and that's magical.

00:33:09.960 | I don't know, it was kind of inspiring to see.

00:33:12.200 | - Did it say hello to, did he say hello to you?

00:33:14.600 | Did he say hello? - No, it's very,

00:33:15.680 | there's a focused nature to the robot.

00:33:17.960 | No, no, listen, I love competence and focus and great.

00:33:21.520 | Like he was not gonna get distracted

00:33:23.200 | by the shallowness of small talk.

00:33:27.760 | There's a job to be done and he was doing it.

00:33:29.920 | So anyway, the fact that it was working

00:33:32.120 | is a beautiful thing.

00:33:32.960 | And I think Media Lab really prides itself

00:33:35.760 | on trying to always have a thing that's working

00:33:37.880 | that it could show off.

00:33:38.920 | - Yes, we used to call it demo or die.

00:33:41.800 | You could not, yeah, you could not like show up

00:33:44.680 | with like PowerPoint or something.

00:33:46.040 | You actually had to have it working.

00:33:47.560 | You know what, my son, who is now 13,

00:33:50.120 | I don't know if this is still his lifelong goal or not,

00:33:53.440 | but when he was a little younger,

00:33:54.840 | his dream is to build an island

00:33:56.600 | that's just inhabited by robots, like no humans.

00:34:00.040 | He just wants all these robots to be connecting

00:34:02.120 | and having fun and so there you go.

00:34:04.560 | - Does he have human,

00:34:05.560 | does he have an idea of which robots he loves most?

00:34:09.240 | Is it Roomba-like robots?

00:34:12.000 | Is it humanoid robots, robot dogs, or is not clear yet?

00:34:15.960 | - We used to have a Jibo,

00:34:18.280 | which was one of the MIT Media Lab spin-outs

00:34:21.520 | and he used to love Jibo. - The thing with a giant head.

00:34:23.360 | - Yes. - That spins.

00:34:24.720 | - Right, exactly.

00:34:25.640 | - It can rotate and it's an eye.

00:34:27.480 | - It has, oh, no, yeah, it can.

00:34:29.520 | - Not glowing, like. - Right, right, right, right.

00:34:31.720 | Exactly. - It's like HAL 9000,

00:34:33.240 | but the friendly version.

00:34:34.440 | - (laughs) Right, he loved that.

00:34:36.960 | And then he just loves,

00:34:38.480 | yeah, he just, I think he loves all forms of robots,

00:34:43.720 | actually.

00:34:44.760 | - So embodied intelligence. - Yes.

00:34:47.840 | - I like, I personally like legged robots, especially.

00:34:50.760 | Anything that can wiggle its butt.

00:34:55.140 | No. - And flip.

00:34:56.340 | - That's not the definition of what I love,

00:34:59.320 | but that's just technically

00:35:00.560 | what I've been working on recently.

00:35:01.840 | So I have a bunch of legged robots now in Austin

00:35:04.560 | and I've been-- - Oh, that's so cool.

00:35:06.200 | - I've been trying to have them communicate affection

00:35:10.460 | with their body in different ways, just for art.

00:35:13.200 | - That's so cool. - For art, really.

00:35:15.280 | 'Cause I love the idea of walking around with the robots,

00:35:18.720 | like as you would with a dog.

00:35:20.520 | I think it's inspiring to a lot of people,

00:35:22.120 | especially young people.

00:35:23.420 | Kids love robots. - Kids love it.

00:35:25.860 | - Parents, like adults are scared of robots,

00:35:28.740 | but kids don't have this kind of weird construction

00:35:31.740 | of the world that's full of evil.

00:35:33.020 | They love cool things. - Yeah.

00:35:35.220 | I remember when Adam was in first grade,

00:35:38.380 | so he must have been like seven or so,

00:35:40.140 | I went into his class with a whole bunch of robots

00:35:43.540 | and the emotion AI demo and da-da-da.

00:35:45.820 | And I asked the kids, I was like,

00:35:47.740 | would you kids want to have a robot friend

00:35:52.900 | or a robot companion?

00:35:53.740 | Everybody said yes, and they wanted it

00:35:55.260 | for all sorts of things, like to help them

00:35:57.780 | with their math homework and to be a friend.

00:36:00.260 | So it just struck me how there was no fear of robots.

00:36:05.260 | Was a lot of adults have that, like us versus them.

00:36:09.540 | - Yeah, none of that.

00:36:12.080 | Of course, you wanna be very careful

00:36:13.860 | because you still have to look at the lessons of history

00:36:17.060 | and how robots can be used by the power centers of the world

00:36:20.700 | to abuse your rights and all that kind of stuff.

00:36:22.620 | But mostly it's good to enter anything new

00:36:26.700 | with an excitement and an optimism.

00:36:28.920 | Speaking of Roz, what have you learned

00:36:32.380 | about science and life from Rosalind Picard?

00:36:35.340 | - Oh my God, I've learned so many things

00:36:37.020 | about life from Roz.

00:36:40.300 | I think the thing I learned the most is perseverance.

00:36:45.300 | When I first met Roz, we applied,

00:36:49.260 | and she invited me to be her postdoc,

00:36:51.420 | we applied for a grant to the National Science Foundation

00:36:55.580 | to apply some of our research to autism.

00:36:57.900 | And we got back, we were rejected.

00:37:01.740 | - Rejected.

00:37:02.580 | - Yeah, and the reasoning was--

00:37:03.420 | - The first time you were rejected for fun, yeah.

00:37:06.140 | - Yeah, and I basically, I just took the rejection

00:37:09.260 | to mean, okay, we're rejected, it's done,

00:37:11.340 | like end of story, right?

00:37:12.940 | And Roz was like, it's great news.

00:37:15.260 | They love the idea, they just don't think we can do it.

00:37:18.340 | So let's build it, show them, and then reapply.

00:37:21.300 | (Roz laughs)

00:37:22.620 | And it was that, oh my God, that story totally stuck with me.

00:37:26.540 | And she's like that in every aspect of her life.

00:37:29.940 | She just does not take no for an answer.

00:37:32.380 | - The reframe all negative feedback.

00:37:34.620 | - It's a challenge.

00:37:36.500 | - It's a challenge.

00:37:37.340 | - It's a challenge.

00:37:38.740 | - Yes, they like this.

00:37:40.180 | - Yeah, yeah, yeah, it was a riot, yeah.

00:37:43.460 | - What else about science in general,

00:37:45.260 | about how you see computers, and also business,

00:37:49.940 | and just everything about the world?

00:37:51.740 | She's a very powerful, brilliant woman like yourself,

00:37:54.980 | so is there some aspect of that too?

00:37:57.500 | - Yeah, I think Roz is actually also very faith-driven.

00:38:00.460 | She has this deep belief and conviction, yeah,

00:38:04.660 | in the good in the world and humanity.

00:38:07.540 | I think that was, meeting her and her family

00:38:11.180 | was definitely a defining moment for me,

00:38:13.660 | because that was when I was like, wow,

00:38:15.820 | you can be of a different background and religion,

00:38:18.980 | and whatever, and you can still have the same core values.

00:38:23.980 | So that was, yeah.

00:38:25.580 | I'm grateful to her.

00:38:28.820 | Roz, if you're listening, thank you.

00:38:30.380 | - Yeah, she's great.

00:38:31.420 | She's been on this podcast before.

00:38:33.780 | I hope she'll be on, I'm sure she'll be on again.

00:38:36.600 | You were the founder and CEO of Affectiva,

00:38:42.620 | which is a big company that was acquired

00:38:44.740 | by another big company, SmartEye,

00:38:47.140 | and you're now the deputy CEO of SmartEye,

00:38:49.260 | so you're a powerful leader, you're brilliant,

00:38:51.940 | you're a brilliant scientist.

00:38:53.500 | A lot of people are inspired by you.

00:38:55.180 | What advice would you give, especially to young women,

00:38:58.420 | but people in general, who dream of becoming

00:39:01.460 | powerful leaders like yourself in a world where perhaps,

00:39:05.220 | in a world that perhaps doesn't give them a clear,

00:39:13.460 | easy path to do so,

00:39:15.620 | whether we're talking about Egypt or elsewhere?

00:39:18.260 | - You know, hearing you kind of describe me that way

00:39:23.180 | kind of encapsulates, I think,

00:39:27.460 | what I think is the biggest challenge of all,

00:39:29.300 | which is believing in yourself, right?

00:39:32.140 | I have had to like grapple with this,

00:39:34.860 | what I call now the Debbie Downer voice in my head.

00:39:39.500 | The kind of basically, it's just shattering all the time,

00:39:42.860 | it's basically saying, "Oh no, no, no, no,

00:39:44.380 | you can't do this.

00:39:45.580 | You're not gonna raise money.

00:39:46.460 | You can't start a company.

00:39:47.620 | What business do you have, like starting a company

00:39:49.500 | or running a company or selling a company?

00:39:51.580 | You name it."

00:39:52.420 | It's always like, and I think my biggest advice

00:39:57.300 | to not just women, but people who are taking a new path

00:40:02.300 | and they're not sure, is to not let yourself

00:40:07.260 | and let your thoughts be the biggest obstacle in your way.

00:40:11.020 | And I've had to like really work on myself

00:40:15.380 | to not be my own biggest obstacle.

00:40:18.660 | - So you got that negative voice.

00:40:20.660 | - Yeah.

00:40:21.740 | - So is that--

00:40:22.580 | - Am I the only one?

00:40:23.420 | I don't think I'm the only one.

00:40:24.340 | - No, I have that negative voice.

00:40:26.020 | I'm not exactly sure if it's a bad thing or a good thing.

00:40:30.980 | I've been really torn about it

00:40:34.620 | because it's been a lifelong companion.

00:40:36.620 | It's hard to know.

00:40:37.660 | It's kind of, it drives productivity and progress

00:40:43.700 | but it can hold you back from taking big leaps.

00:40:46.940 | I think the best I can say is probably you have

00:40:52.300 | to somehow be able to control it.

00:40:55.660 | So turn it off when it's not useful

00:40:58.060 | and turn it on when it's useful.

00:41:00.500 | Like I have from almost like a third person perspective.

00:41:03.260 | - Right, somebody who's sitting there like--

00:41:05.340 | - Yeah, like, because it is useful to be critical.

00:41:10.340 | Like after, I just gave a talk yesterday

00:41:15.380 | at MIT and I was just, you know, there's so much love

00:41:20.900 | and it was such an incredible experience.

00:41:23.340 | So many amazing people I got a chance to talk to.

00:41:25.860 | But, you know, afterwards when I went home

00:41:29.900 | and just took this long walk,

00:41:31.420 | it was mostly just negative thoughts about me.

00:41:33.980 | I don't, like one basic stuff,

00:41:37.060 | like I don't deserve any of it.

00:41:38.980 | And second is like, why did you, that was so dumb.

00:41:43.220 | That you said this, that's so dumb.

00:41:45.060 | Like you should have prepared that better.

00:41:47.740 | Why did you say this?

00:41:48.980 | The, the, the, the, the, the, the.

00:41:50.420 | But I think it's good to hear that voice out.

00:41:54.340 | All right, and like sit in that.

00:41:56.420 | And ultimately I think you grow from that.

00:41:58.780 | Now, when you're making really big decisions

00:42:00.500 | about funding or starting a company

00:42:03.460 | or taking a leap to go to the UK

00:42:06.340 | or take a leap to go to America to work in media lab.

00:42:10.780 | Though, yeah, there's a, that's,

00:42:15.540 | you should be able to shut that off then

00:42:19.100 | because you should have like this weird confidence,

00:42:23.780 | almost like faith that you said before

00:42:25.500 | that everything's gonna work out.

00:42:26.980 | So take the leap of faith.

00:42:28.780 | - Take the leap of faith.

00:42:30.340 | Despite all the negativity.

00:42:32.580 | I mean, there's, there's, there's some of that.

00:42:34.420 | You, you actually tweeted a really nice tweet thread.

00:42:38.940 | It says, quote, a year ago,

00:42:41.460 | a friend recommended I do daily affirmations.

00:42:44.560 | And I was skeptical,

00:42:46.900 | but I was going through major transitions in my life.

00:42:49.560 | So I gave it a shot and it set me on a journey

00:42:51.980 | of self-acceptance and self-love.

00:42:54.180 | So what was that like?

00:42:56.460 | Maybe talk through this idea of affirmations

00:43:00.340 | and how that helped you.

00:43:01.380 | - Yeah, because really like, I'm just like me,

00:43:04.580 | I'm a kind, I like to think of myself

00:43:06.820 | as a kind person in general,

00:43:08.260 | but I'm kind of mean to myself sometimes.

00:43:11.380 | And so I've been doing journaling for almost 10 years now.

00:43:15.820 | I use an app called Day One and it's awesome.

00:43:19.020 | I just journal and I use it as an opportunity

00:43:21.000 | to almost have a conversation

00:43:22.460 | with the Debbie Downer voice in my,

00:43:23.780 | it's like a rebuttal, right?

00:43:25.700 | Like Debbie Downer says,

00:43:27.020 | "Oh my God, you won't be able to raise this round of funding."

00:43:29.860 | I'm like, "Okay, let's talk about it."

00:43:32.180 | I have a track record of doing X, Y, and Z.

00:43:35.660 | I think I can do this.

00:43:37.300 | And it's literally like,

00:43:39.380 | so I don't know that I can shut off the voice,

00:43:42.480 | but I can have a conversation with it.

00:43:44.420 | And it just, and I bring data to the table, right?

00:43:48.180 | - Nice.

00:43:50.840 | - So that was the journaling part,

00:43:52.060 | which I found very helpful.

00:43:53.980 | But the affirmation took it to a whole next level

00:43:56.620 | and I just love it.

00:43:57.780 | I'm a year into doing this.

00:44:00.620 | And you literally wake up in the morning

00:44:02.340 | and the first thing you do, I meditate first.

00:44:05.660 | And then I write my affirmations

00:44:07.900 | and it's the energy I want to put out in the world

00:44:10.660 | that hopefully will come right back to me.

00:44:12.300 | So I will say, I always start with,

00:44:15.100 | "My smile lights up the whole world."

00:44:17.420 | And I kid you not, like people in the street will stop me

00:44:19.620 | and say, "Oh my God, like we love your smile."

00:44:21.580 | - Yeah. - Like, yes.

00:44:23.820 | So my affirmations will change depending on

00:44:27.820 | what's happening this day.

00:44:28.980 | Is it funny?

00:44:29.820 | I know, don't judge, don't judge.

00:44:31.500 | - No, that's not, laughter's not judgment.

00:44:33.940 | It's just awesome.

00:44:35.180 | I mean, it's true,

00:44:37.300 | but you're saying affirmations somehow help kind of,

00:44:41.500 | what is it?

00:44:42.340 | They do work to like remind you

00:44:46.300 | of the kind of person you are

00:44:47.820 | and the kind of person you want to be,

00:44:49.540 | which actually may be inverse order.

00:44:52.700 | The kind of person you want to be

00:44:53.900 | and that helps you become the kind of person

00:44:55.900 | you actually are.

00:44:56.740 | - It's just, it brings intentionality

00:44:59.380 | to like what you're doing, right?

00:45:01.780 | And so--

00:45:02.820 | - By the way, I was laughing because my affirmations,

00:45:06.140 | which I also do, are the opposite.

00:45:07.980 | - Oh, you do?

00:45:08.820 | Oh, what do you do?

00:45:09.660 | - I don't have a, "My smile lights up the world."

00:45:12.380 | (laughing)

00:45:14.740 | Maybe I should add that because like I have just,

00:45:18.020 | I have a, oh boy.

00:45:21.820 | It's much more stoic, like about focus,

00:45:25.940 | about this kind of stuff.

00:45:28.180 | But the joy, the emotion that you're,

00:45:30.540 | just in that little affirmation is beautiful.

00:45:33.100 | So maybe I should add that.

00:45:34.580 | - I have some like focus stuff,

00:45:37.620 | but that's usually like after--

00:45:38.460 | - But that's a cool start.

00:45:39.380 | That's a good, it's just--

00:45:40.220 | - It's after all the like smiling,

00:45:41.660 | I'm playful and joyful and all that,

00:45:43.940 | and then it's like, okay, I kick butt.

00:45:45.700 | - Let's get shit done.

00:45:46.660 | - Right, exactly.

00:45:47.500 | - Let's get shit done affirmations.

00:45:48.820 | Okay, cool.

00:45:50.100 | Like what else is on there?

00:45:51.500 | - Oh, what else is on there?

00:45:54.460 | Well, I have, I'm a magnet for all sorts of things.

00:45:59.460 | So I'm an amazing people magnet.

00:46:02.380 | I attract like awesome people into my universe.

00:46:04.780 | - So that's an actual affirmation?

00:46:07.060 | - Yes.

00:46:08.020 | - That's great.

00:46:08.980 | - Yeah.

00:46:09.820 | - So that's, and that somehow manifests itself

00:46:12.340 | into like in working.

00:46:14.100 | - I think so.

00:46:15.620 | - Yeah, like can you speak to like why it feels good

00:46:18.140 | to do the affirmations?

00:46:19.940 | - I honestly think it just grounds the day.

00:46:24.260 | And then it allows me to,

00:46:26.460 | instead of just like being pulled back and forth

00:46:29.660 | like throughout the day, it just like grounds me.

00:46:31.900 | I'm like, okay, like this thing happened.

00:46:34.740 | It's not exactly what I wanted it to be, but I'm patient.

00:46:37.500 | Or I'm, you know, I trust that the universe

00:46:41.420 | will do amazing things for me,

00:46:42.660 | which is one of my other consistent affirmations.

00:46:45.580 | Or I'm an amazing mom, right?

00:46:47.180 | And so I can grapple with all the feelings of mom guilt

00:46:50.260 | that I have all the time.

00:46:51.540 | Or here's another one.

00:46:53.980 | I'm a love magnet.

00:46:55.220 | And I literally say, I will kind of picture the person

00:46:58.100 | that I'd love to end up with.

00:46:59.220 | And I write it all down and hasn't happened yet, but it-

00:47:02.620 | - What are you picturing?

00:47:04.220 | This is Brad Pitt.

00:47:05.060 | - Okay, Brad Pitt and Brad Pitt.

00:47:06.220 | - 'Cause that's what I picture.

00:47:07.180 | - Okay, that's what you picture?

00:47:08.420 | - Yeah, yeah.

00:47:09.260 | - Okay, okay.

00:47:10.100 | - On the running, holding hands, running together.

00:47:11.940 | - Okay.

00:47:14.340 | - No, more like Fight Club, the Fight Club Brad Pitt,

00:47:17.820 | where he's like standing, all right, people will know.

00:47:20.220 | Anyway, I'm sorry, I'll get off on that.

00:47:22.140 | Do you have, like when you're thinking

00:47:23.740 | about being a love magnet in that way,

00:47:27.180 | are you picturing specific people?

00:47:29.020 | Or is this almost like in the space of like energy?

00:47:34.020 | - Right, it's somebody who is smart and well accomplished

00:47:41.900 | and successful in their life, but they're generous

00:47:45.340 | and they're well-traveled and they wanna travel the world.

00:47:49.060 | It's things like that.

00:47:49.900 | I'm like, their head over heels into me.

00:47:51.500 | It's like, I know it sounds super silly,

00:47:53.220 | but it's literally what I write.

00:47:54.940 | And I believe it'll happen one day.

00:47:56.500 | - Oh, you actually write, so you don't say it out loud?

00:47:58.380 | - No, I write it.

00:47:59.220 | I write all my affirmations.

00:48:01.380 | - I do the opposite, I say it out loud.

00:48:02.820 | - Oh, you say it out loud, interesting.

00:48:04.420 | - Yeah, if I'm alone, I'll say it out loud, yeah.

00:48:06.660 | - Interesting, I should try that.

00:48:10.180 | I think it's what feels more powerful to you,

00:48:15.180 | to me more powerful, saying stuff feels more powerful.

00:48:20.500 | - Yeah.

00:48:21.620 | - Writing feels like I'm losing the words,

00:48:26.620 | like losing the power of the words,

00:48:32.500 | maybe 'cause I write slow.

00:48:33.760 | Do you handwrite?

00:48:35.060 | - No, I type, it's on this app.

00:48:37.700 | It's day one, basically.

00:48:39.940 | The best thing about it is I can look back

00:48:42.620 | and see like a year ago, what was I affirming, right?

00:48:46.260 | - Oh, so it changes over time.

00:48:48.940 | - It hasn't like changed a lot,

00:48:51.860 | but the focus kind of changes over time.

00:48:54.780 | - I got it.

00:48:55.620 | Yeah, I say the same exact thing over and over and over.

00:48:58.020 | - Oh, you do?

00:48:58.860 | Okay.

00:48:59.700 | - There's a comfort in the sameness of it.

00:49:01.260 | Actually, let me jump around,

00:49:03.900 | 'cause let me ask you about,

00:49:04.900 | 'cause all this talk about Brad Pitt,

00:49:07.420 | or maybe that's just going on inside my head.

00:49:10.180 | Let me ask you about dating in general.

00:49:12.720 | You tweeted, "Are you based in Boston?"

00:49:16.860 | in single, question mark,

00:49:18.700 | and then you pointed to a startup,

00:49:21.580 | Singles Night, sponsored by Smile Dating App.

00:49:24.420 | I mean, this is jumping around a little bit,

00:49:26.860 | but since you mentioned,

00:49:28.460 | can AI help solve this dating love problem?

00:49:34.660 | What do you think?

00:49:35.500 | What's the form of connection

00:49:36.940 | that is part of the human condition?

00:49:40.180 | Can AI help that?

00:49:41.620 | You yourself are in the search affirming.

00:49:43.920 | - Maybe that's what I should affirm,

00:49:47.140 | like build an AI.

00:49:48.420 | - Build an AI that finds love?

00:49:49.940 | - I think there must be a science behind

00:49:55.780 | that first moment you meet a person

00:49:59.220 | and you either have chemistry or you don't, right?

00:50:02.980 | - I guess that was the question I was asking.

00:50:04.740 | Would you put it brilliantly?

00:50:06.220 | Is that a science or an art?

00:50:07.740 | - Ooh, I think there's actual chemicals

00:50:13.060 | that get exchanged when two people meet.

00:50:15.300 | Oh, well, I don't know about that.

00:50:17.100 | (laughing)

00:50:18.260 | - I like how you're changing,

00:50:19.700 | yeah, yeah, changing your mind as we're describing it.

00:50:22.180 | But it feels that way.

00:50:24.060 | But what science shows us

00:50:25.940 | is sometimes we can explain with rigor

00:50:28.700 | the things that feel like magic.

00:50:30.740 | - Right.

00:50:31.940 | - So maybe we can remove all the magic.

00:50:34.400 | Maybe it's like, I honestly think,

00:50:37.020 | like I said, Goodreads should be a dating app,

00:50:39.900 | which like books, I wonder if you look at just like books

00:50:44.900 | or content you've consumed.

00:50:47.180 | I mean, that's essentially what YouTube does

00:50:48.740 | when it does a recommendation.

00:50:50.820 | If you just look at your footprint of content consumed,

00:50:54.540 | if there's an overlap,

00:50:56.020 | but maybe interesting difference with an overlap,

00:50:59.020 | there's some, I'm sure this is a machine learning problem

00:51:01.660 | that's solvable.

00:51:04.240 | This person is very likely to be,

00:51:06.880 | not only there to be chemistry in the short term,

00:51:10.760 | but a good lifelong partner to grow together.

00:51:14.100 | I bet you it's a good machine learning problem.

00:51:15.780 | We just need the data.

00:51:16.620 | - Let's do it.

00:51:17.500 | Well, actually, I do think there's so much data

00:51:20.560 | about each of us that there ought to be

00:51:22.420 | a machine learning algorithm that can ingest all this data

00:51:25.020 | and basically say, I think the following 10 people

00:51:28.060 | would be interesting connections for you, right?

00:51:32.020 | And so Smile Dating App kind of took one particular angle,

00:51:35.900 | which is humor.

00:51:36.780 | It matches people based on their humor styles,

00:51:39.200 | which is one of the main ingredients

00:51:41.900 | of a successful relationship.

00:51:43.280 | Like if you meet somebody and they can make you laugh,

00:51:45.400 | like that's a good thing.

00:51:47.460 | And if you develop like internal jokes,

00:51:49.420 | like inside jokes and you're bantering, like that's fun.

00:51:54.420 | So I think.

00:51:56.780 | - Yeah, definitely.

00:51:58.560 | - But yeah, that's the number of,

00:52:01.880 | and the rate of inside joke generation.

00:52:05.280 | You could probably measure that and then optimize it

00:52:08.280 | over the first few days.

00:52:09.320 | You can see. - Right, and then.

00:52:10.800 | - We're just turning this into a machine learning problem.

00:52:12.660 | I love it.

00:52:13.500 | But for somebody like you,

00:52:16.000 | who's exceptionally successful and busy,

00:52:18.680 | is there science to that aspect of dating?

00:52:26.200 | Is it tricky?

00:52:27.560 | Is there advice you can give?

00:52:28.880 | - Oh my God, I'd give the worst advice.

00:52:30.680 | Well, I can tell you like I have a spreadsheet.

00:52:33.680 | - A spreadsheet, that's great.

00:52:36.080 | Is that a good or a bad thing?

00:52:37.160 | Do you regret the spreadsheet?

00:52:38.660 | - Well, I don't know.

00:52:40.680 | - What's the name of the spreadsheet?

00:52:42.040 | Is it love?

00:52:42.880 | - It's the dating tracker.

00:52:45.160 | - The dating tracker?

00:52:46.120 | - It's very like. - Love tracker.

00:52:47.640 | - Yeah.

00:52:48.480 | - And there's a rating system, I'm sure.

00:52:50.200 | - Yeah, there's like weights and stuff.

00:52:52.480 | - It's too close to home.

00:52:53.960 | - Oh, is it?

00:52:54.780 | Do you also have a spreadsheet?

00:52:55.620 | - Well, I don't have a spreadsheet,

00:52:56.440 | but I would, now that you say it,

00:52:58.160 | it seems like a good idea.

00:52:59.680 | - Oh no.

00:53:00.520 | - Turning into data.

00:53:03.940 | I do wish that somebody else had a spreadsheet about me.

00:53:09.560 | If it was like I said, like you said,

00:53:15.040 | convert, collect a lot of data about us

00:53:17.440 | in a way that's privacy preserving,

00:53:19.440 | that I own the data, I can control it,

00:53:21.640 | and then use that data to find,

00:53:23.760 | I mean, not just romantic love,

00:53:25.240 | but collaborators, friends, all that kind of stuff.

00:53:29.120 | It seems like the data is there.

00:53:30.720 | That's the problem social networks are trying to solve,

00:53:34.240 | but I think they're doing a really poor job.

00:53:36.440 | Even Facebook tried to get into a dating app business.

00:53:39.720 | And I think there's so many components

00:53:41.560 | to running a successful company that connects human beings.

00:53:45.440 | And part of that is,

00:53:47.560 | having engineers that care about the human side, right?

00:53:53.600 | As you know extremely well,

00:53:55.120 | it's not easy to find those.

00:53:57.960 | But you also don't want just people that care

00:54:00.240 | about the human, they also have to be good engineers.

00:54:02.360 | So it's like, you have to find this beautiful mix.

00:54:05.920 | And for some reason, just empirically speaking,

00:54:08.400 | people have not done a good job of that,

00:54:12.720 | building companies like that.

00:54:14.280 | It must mean that it's a difficult problem to solve.

00:54:17.160 | Dating apps, it seems difficult.

00:54:20.000 | Okay, Cupid, Tinder, all those kind of stuff.

00:54:23.000 | They seem to find, of course they work,

00:54:27.520 | but they seem to not work as well

00:54:31.960 | as I would imagine is possible.

00:54:34.640 | With data, wouldn't you be able

00:54:35.920 | to find better human connection?

00:54:38.040 | It's like arrange marriages on steroids, essentially.

00:54:41.520 | Arranged by machine learning algorithm.

00:54:43.520 | - Arranged by machine learning algorithm,

00:54:45.120 | but not a superficial one.

00:54:46.520 | I think a lot of the dating apps out there

00:54:48.360 | are just so superficial.

00:54:49.560 | They're just matching on high level criteria

00:54:52.560 | that aren't ingredients for successful partnership.

00:54:57.040 | But you know what's missing though, too?

00:54:59.680 | I don't know how to fix that.

00:55:00.800 | The serendipity piece of it.

00:55:02.520 | Like how do you engineer serendipity?

00:55:04.760 | Like this random chance encounter,

00:55:07.440 | and then you fall in love with the person.

00:55:08.840 | I don't know how a dating app can do that.

00:55:11.120 | So there has to be a little bit of randomness.

00:55:13.080 | Maybe every 10th match is just a,

00:55:20.280 | yeah, somebody that the algorithm

00:55:22.320 | wouldn't have necessarily recommended,

00:55:24.480 | but it allows for a little bit of.

00:55:27.280 | - Well, it can also trick you into thinking

00:55:31.400 | that serendipity by somehow showing you a tweet

00:55:35.480 | of a person that he thinks you'll match well with,

00:55:39.120 | but do it accidentally as part of another search.

00:55:41.760 | And you just notice it, and then you go down a rabbit hole

00:55:46.040 | and you connect them outside the app.

00:55:49.400 | You connect with this person outside the app somehow.

00:55:51.520 | So it's just, it creates that moment of meeting.

00:55:54.360 | Of course, you have to think of, from an app perspective,

00:55:56.680 | how you can turn that into a business.

00:55:58.420 | But I think ultimately a business

00:56:01.220 | that helps people find love in any way,

00:56:04.640 | like that's what Apple was about.

00:56:05.880 | Create products that people love.

00:56:07.600 | That's beautiful.

00:56:08.440 | I mean, you gotta make money somehow.

00:56:11.560 | If you help people fall in love personally with a product,

00:56:16.000 | find self-love or love another human being,

00:56:18.960 | you're gonna make money.

00:56:20.080 | You're gonna figure out a way to make money.

00:56:22.240 | I just feel like the dating apps often will optimize

00:56:26.680 | for something else than love.

00:56:28.680 | It's the same with social networks.

00:56:30.120 | They optimize for engagement

00:56:31.960 | as opposed to a deep, meaningful connection

00:56:35.160 | that's ultimately grounded in personal growth,

00:56:38.160 | you as a human being growing and all that kind of stuff.

00:56:41.720 | Let me do a pivot to a dark topic,

00:56:45.720 | which you opened the book with.

00:56:48.680 | A story, because I'd like to talk to you

00:56:51.720 | about just emotion and artificial intelligence,

00:56:56.120 | and I think this is a good story

00:56:57.440 | to start to think about emotional intelligence.

00:57:00.880 | You opened the book with a story

00:57:02.320 | of a Central Florida man, Jamelle Dunn,

00:57:05.060 | who was drowning and drowned

00:57:07.280 | while five teenagers watched and laughed,

00:57:09.820 | saying things like, "You're gonna die."

00:57:11.760 | And when Jamelle disappeared

00:57:13.480 | below the surface of the water,

00:57:14.820 | one of them said, "He just died," and the others laughed.

00:57:18.440 | What does this incident teach you about human nature

00:57:21.840 | and the response to it, perhaps?

00:57:24.240 | - Yeah, I mean, I think this is a really,

00:57:26.840 | really, really sad story,

00:57:29.560 | and it highlights what I believe is a,

00:57:32.920 | it's a real problem in our world today.

00:57:35.520 | It's an empathy crisis.

00:57:37.680 | Yeah, we are living through an empathy crisis.

00:57:40.080 | - Empathy crisis, yeah.

00:57:41.360 | - Yeah, and I mean, we've talked about this

00:57:45.280 | throughout our conversation.

00:57:46.320 | We dehumanize each other,

00:57:48.040 | and unfortunately, yes, technology is bringing us together,

00:57:52.920 | but in a way, it's just dehumanizing.

00:57:54.800 | It's creating this, like, yeah, dehumanizing of the other,

00:57:59.720 | and I think that's a huge problem.

00:58:02.760 | The good news is I think the solution

00:58:05.200 | could be technology-based.

00:58:06.840 | Like, I think if we rethink the way we design

00:58:09.560 | and deploy our technologies,

00:58:11.280 | we can solve parts of this problem,

00:58:13.600 | but I worry about it.

00:58:14.440 | I mean, even with my son,

00:58:16.360 | a lot of his interactions are computer-mediated,

00:58:21.200 | and I just question what that's doing to his empathy skills

00:58:25.680 | and his ability to really connect with people, so.

00:58:29.400 | - Do you think,

00:58:30.280 | you think it's not possible to form empathy

00:58:35.080 | through the digital medium?

00:58:36.760 | - I think it is,

00:58:39.760 | but we have to be thoughtful about,

00:58:43.120 | 'cause the way we engage face-to-face,

00:58:45.680 | which is what we're doing right now, right?

00:58:47.800 | There's the nonverbal signals,

00:58:49.320 | which are a majority of how we communicate.

00:58:51.320 | It's like 90% of how we communicate

00:58:53.600 | is your facial expressions.

00:58:56.040 | You know, I'm saying something,

00:58:57.240 | and you're nodding your head now,

00:58:58.560 | and that creates a feedback loop,

00:59:00.440 | and if you break that-

00:59:02.600 | - And now I have anxiety about it.

00:59:04.000 | - Da-da-da, right? (laughs)

00:59:06.120 | Poor Lex.

00:59:06.960 | - Oh, boy, I wish this cycle-

00:59:08.880 | - I am not scrutinizing your facial expressions

00:59:10.760 | during this interview, right?

00:59:11.600 | - I am, I am. - Okay.

00:59:12.920 | - Look normal, look human.

00:59:15.360 | - Yeah. (laughs)

00:59:17.800 | - Nod head.

00:59:18.760 | - Yeah, nod head. (laughs)

00:59:21.040 | - In agreement.

00:59:21.880 | - If Rana says yes, then nod head else.

00:59:25.840 | - Don't do it too much,

00:59:26.760 | because it might be at the wrong time,

00:59:28.760 | and then it'll send the wrong signal.

00:59:30.920 | - Oh, God.

00:59:31.760 | - And make eye contact sometimes,

00:59:33.680 | 'cause humans appreciate that.

00:59:35.440 | All right, anyway.

00:59:36.280 | - Okay. (laughs)

00:59:38.640 | - Yeah, but something about,

00:59:40.360 | especially when you say mean things in person,

00:59:42.600 | you get to see the pain of the other person.

00:59:44.360 | - Exactly, but if you're tweeting it at a person,

00:59:46.360 | and you have no idea how it's gonna land,

00:59:48.080 | you're more likely to do that on social media

00:59:50.200 | than you are in face-to-face conversations, so.

00:59:52.680 | - And what do you think is more important?

00:59:56.640 | EQ or IQ?

01:00:00.520 | EQ being emotional intelligence.

01:00:02.560 | In terms of, in what makes us human?

01:00:06.720 | - I think emotional intelligence is what makes us human.

01:00:11.400 | It's how we connect with one another,

01:00:14.960 | it's how we build trust, it's how we make decisions, right?

01:00:19.800 | Like your emotions drive kind of what you had for breakfast,

01:00:24.600 | but also where you decide to live,

01:00:26.280 | and what you wanna do for the rest of your life.

01:00:28.800 | So I think emotions are underrated.

01:00:31.720 | - So emotional intelligence isn't just about

01:00:36.440 | the effective expression of your own emotions,

01:00:39.320 | it's about a sensitivity and empathy

01:00:41.480 | to other people's emotions,

01:00:43.120 | and that sort of being able to effectively engage

01:00:46.400 | in the dance of emotions with other people.

01:00:48.960 | - Yeah, I like that explanation.

01:00:51.400 | I like that kind of, yeah, thinking about it as a dance,

01:00:55.240 | because it is really about that,

01:00:56.800 | it's about sensing what state the other person's in

01:00:59.240 | and using that information to decide

01:01:01.760 | on how you're gonna react.

01:01:03.320 | And I think it can be very powerful,

01:01:07.040 | people who are the best, most persuasive leaders

01:01:12.040 | in the world tap into,

01:01:14.440 | if you have higher EQ, you're more likely

01:01:17.920 | to be able to motivate people to change their behaviors.

01:01:21.520 | So it can be very powerful.

01:01:25.120 | - At a more kind of technical, maybe philosophical level,

01:01:29.320 | you've written that emotion is universal.

01:01:32.640 | It seems that, sort of like Chomsky says,

01:01:36.240 | "Language is universal."

01:01:37.760 | There's a bunch of other stuff like cognition,

01:01:39.760 | consciousness, it seems a lot of us have these aspects.

01:01:44.200 | So the human mind generates all this.

01:01:47.520 | So what do you think is the,

01:01:49.580 | they all seem to be like echoes of the same thing.

01:01:53.520 | What do you think emotion is exactly?

01:01:57.240 | Like how deep does it run?

01:01:58.480 | Is it a surface level thing that we display to each other?

01:02:02.400 | Is it just another form of language

01:02:04.200 | or something deep within?

01:02:06.360 | - I think it's really deep.

01:02:08.120 | It's how, we started with memory.

01:02:10.560 | I think emotions play a really important role.

01:02:13.520 | Yeah, emotions play a very important role

01:02:17.200 | in how we encode memories, right?

01:02:18.800 | Our memories are often encoded, almost indexed by emotions.

01:02:22.520 | Yeah, it's at the core of how our decision-making engine

01:02:29.760 | is also heavily influenced by our emotions.

01:02:32.880 | - So emotions is part of cognition.

01:02:34.800 | - Totally.

01:02:35.800 | - It's intermixing to the whole thing.

01:02:37.640 | - Yes, absolutely.

01:02:38.920 | And in fact, when you take it away,

01:02:41.080 | people are unable to make decisions.

01:02:42.920 | They're really paralyzed.

01:02:43.880 | Like they can't go about their daily

01:02:45.960 | or their personal or professional lives.

01:02:48.320 | - It does seem like there's probably some interesting

01:02:53.920 | interweaving of emotion and consciousness.

01:02:58.920 | I wonder if it's possible to have,

01:03:01.480 | like if they're next door neighbors somehow

01:03:03.840 | or if they're actually flatmates.

01:03:07.040 | It feels like the hard problem of consciousness

01:03:12.600 | where it feels like something to experience the thing.

01:03:15.760 | Red feels like red.

01:03:20.080 | When you eat a mango, it's sweet.

01:03:22.040 | The taste, the sweetness,

01:03:24.600 | that it feels like something to experience that sweetness.

01:03:27.640 | Whatever generates emotions.

01:03:33.360 | But then like, see, I feel like emotion

01:03:35.600 | is part of communication.

01:03:37.120 | It's very much about communication.

01:03:39.560 | And then that means it's also deeply connected to language.

01:03:44.560 | But then probably human intelligence is deeply connected

01:03:50.720 | to the collective intelligence between humans.

01:03:52.800 | It's not just a standalone thing.

01:03:54.680 | So the whole thing is really connected.

01:03:56.440 | So emotion is connected to language.

01:03:58.840 | Language is connected to intelligence.

01:04:01.200 | And then intelligence is connected to consciousness

01:04:03.360 | and consciousness is connected to emotion.

01:04:05.880 | The whole thing is a beautiful mess.

01:04:09.240 | So.

01:04:10.080 | - Can I comment on the emotions

01:04:14.200 | being a communication mechanism?

01:04:15.800 | 'Cause I think there are two facets

01:04:18.280 | of our emotional experiences.

01:04:21.600 | One is communication, right?

01:04:24.540 | Like we use emotions, for example, facial expressions

01:04:27.840 | or other nonverbal cues to connect

01:04:29.960 | with other human beings and with other beings

01:04:33.960 | in the world, right?

01:04:35.040 | But even if it's not a communication context,

01:04:40.680 | we still experience emotions and we still process emotions

01:04:43.520 | and we still leverage emotions to make decisions

01:04:47.020 | and to learn and to experience life.

01:04:49.800 | So it isn't always just about communication.

01:04:53.240 | And we learned that very early on

01:04:54.720 | in kind of our work at Affectiva.

01:04:58.600 | One of the very first applications we brought to market

01:05:01.240 | was understanding how people respond to content, right?

01:05:03.680 | So if they're watching this video of ours,

01:05:05.520 | like are they interested?

01:05:06.720 | Are they inspired?

01:05:07.800 | Are they bored to death?

01:05:09.240 | And so we watched their facial expressions.

01:05:11.640 | And we weren't sure if people would express any emotions

01:05:16.520 | if they were sitting alone.

01:05:17.480 | Like if you're in your bed at night,

01:05:19.880 | watching a Netflix TV series,

01:05:21.320 | would we still see any emotions on your face?

01:05:23.360 | And we were surprised that yes, people still emote,

01:05:26.040 | even if they're alone.

01:05:27.000 | Even if you're in your car driving around,

01:05:29.160 | you're singing along a song and you're joyful,

01:05:32.400 | we'll see these expressions.

01:05:33.680 | So it's not just about communicating with another person.

01:05:37.520 | It sometimes really is just about experiencing the world.

01:05:40.760 | - First of all, I wonder if some of that

01:05:44.240 | is because we develop our intelligence

01:05:47.360 | and our emotional intelligence

01:05:50.000 | by communicating with other humans.

01:05:51.960 | And so when other humans disappear from the picture,

01:05:54.520 | we're still kind of a virtual human.

01:05:56.720 | - The code still runs, basically.

01:05:57.840 | - Yeah, the code still runs.

01:05:59.320 | But you're also kind of, you're still,

01:06:01.720 | there's like virtual humans.

01:06:02.960 | You don't have to think of it that way,

01:06:04.640 | but there's a kind of, when you like chuckle, like, yeah.

01:06:09.640 | Like you're kind of chuckling to a virtual human.

01:06:13.240 | I mean, it's possible that the code

01:06:17.480 | has to have another human there.

01:06:21.920 | Because if you just grew up alone,

01:06:24.440 | I wonder if emotion will still be there in this visual form.

01:06:28.720 | So yeah, I wonder.

01:06:30.400 | But anyway, what can you tell from the human face

01:06:35.400 | about what's going on inside?

01:06:38.520 | So that's the problem that Effektiva first tackled,

01:06:41.200 | which is using computer vision, using machine learning

01:06:46.080 | to try to detect stuff about the human face

01:06:49.080 | as many things as possible,

01:06:51.000 | and convert them into a prediction

01:06:52.880 | of categories of emotion, anger, happiness,

01:06:57.600 | all that kind of stuff.

01:06:58.840 | How hard is that problem?

01:07:00.440 | - It's extremely hard.

01:07:01.520 | It's very, very hard,

01:07:02.500 | because there is no one-to-one mapping

01:07:04.320 | between a facial expression and your internal state.

01:07:08.560 | There just isn't.

01:07:09.580 | There's this oversimplification of the problem,

01:07:11.920 | where it's something like,

01:07:13.320 | if you are smiling, then you're happy.

01:07:15.280 | If you do a brow furrow, then you're angry.

01:07:17.440 | If you do an eyebrow raise, then you're surprised.

01:07:19.680 | And just think about it for a moment.

01:07:22.320 | You could be smiling for a whole host of reasons.

01:07:25.160 | You could also be happy and not be smiling, right?

01:07:27.900 | You could furrow your eyebrows because you're angry,

01:07:31.880 | or you're confused about something, or you're constipated.

01:07:35.760 | So I think this oversimplistic approach

01:07:39.760 | to inferring emotion from a facial expression

01:07:41.980 | is really dangerous.

01:07:43.200 | The solution is to incorporate

01:07:47.680 | as many contextual signals as you can.

01:07:50.120 | So if, for example, I'm driving a car,

01:07:53.640 | and you can see me nodding my head,

01:07:56.300 | and my eyes are closed, and the blinking rate is changing,

01:07:59.600 | I'm probably falling asleep at the wheel, right?

01:08:02.400 | Because you know the context.

01:08:04.620 | You understand what the person's doing.

01:08:06.560 | Or add additional channels, like voice, or gestures,

01:08:11.720 | or even physiological sensors.

01:08:13.960 | But I think it's very dangerous

01:08:16.440 | to just take this oversimplistic approach of,

01:08:19.640 | yeah, smile equals happy.

01:08:21.640 | - If you're able to, in a high-resolution way,

01:08:24.120 | specify the context, there's certain things

01:08:26.520 | that are gonna be somewhat reliable signals

01:08:29.720 | of something like drowsiness, or happiness,

01:08:32.960 | or stuff like that.

01:08:34.320 | I mean, when people are watching Netflix content,

01:08:37.040 | that problem, that's a really compelling idea

01:08:41.460 | that you can kind of, at least in aggregate--

01:08:44.600 | - Exactly. - Highlight,

01:08:45.960 | like which part was boring, which part was exciting.

01:08:48.960 | How hard was that problem?

01:08:50.380 | - That was on the scale of difficulty.

01:08:54.920 | I think that's one of the easier problems to solve,

01:08:57.800 | because it's a relatively constrained environment.

01:09:01.720 | You have somebody sitting in front of,

01:09:03.920 | initially we started with a device in front of you,

01:09:06.440 | like a laptop, and then we graduated

01:09:08.940 | to doing this on a mobile phone, which is a lot harder,

01:09:11.920 | just because of, from a computer vision perspective,

01:09:15.480 | the profile view of the face can be a lot more challenging.

01:09:19.160 | We had to figure out lighting conditions,

01:09:20.800 | because usually people are watching content

01:09:23.880 | literally in their bedrooms at night, lights are dimmed.

01:09:26.800 | - Yeah, I mean, if you're standing,

01:09:29.800 | it's probably gonna be the looking up.

01:09:33.120 | - The nostril view. - Yeah.

01:09:34.720 | And nobody looks good at, I've seen data sets

01:09:38.000 | from that perspective, it's like,

01:09:39.400 | ugh, this is not a good look for anyone.

01:09:43.160 | Or if you're laying in bed at night,

01:09:45.240 | what is it, side view or something?

01:09:47.520 | And half your face is on a pillow.

01:09:50.240 | Actually, I would love to have data

01:09:53.680 | about how people watch stuff in bed at night.

01:09:58.680 | Do they prop their, is it a pillow?

01:10:03.600 | I'm sure there's a lot of interesting dynamics there.

01:10:07.520 | - From a health and wellbeing perspective, right?

01:10:09.520 | Like, it's like, oh, you're hurting your neck.

01:10:11.200 | - I was thinking machine learning perspective, but yes.

01:10:13.200 | But also, yeah.

01:10:15.680 | Once you have that data, you can start making

01:10:17.880 | all kinds of inference about health and stuff like that.

01:10:20.120 | - Interesting.

01:10:21.360 | - Yeah, there was an interesting thing

01:10:23.200 | when I was at Google that we were,

01:10:25.360 | it's called active authentication,

01:10:29.560 | where you want to be able to unlock your phone

01:10:33.440 | without using a password.

01:10:34.720 | So it would face, but also other stuff.

01:10:38.600 | Like the way you take a phone out of the pocket.

01:10:41.200 | - Amazing. - So that kind of data,

01:10:42.640 | to use the multimodal with machine learning

01:10:46.080 | to be able to identify that it's you,

01:10:47.760 | or likely to be you, likely not to be you.

01:10:50.800 | That allows you to not always have to enter the password.

01:10:52.960 | That was the idea.

01:10:54.480 | But the funny thing about that is,

01:10:56.560 | I just want to tell a small anecdote,

01:10:58.880 | is 'cause it was all male engineers.

01:11:01.920 | Except, so my boss is, our boss,

01:11:08.920 | who's still one of my favorite humans, was a woman,

01:11:12.720 | Regina Dugan.

01:11:13.560 | - Oh my God, I love her.

01:11:15.560 | She's awesome. - Yeah, she's the best.

01:11:16.720 | She's the best.

01:11:17.560 | So, but anyway, and there was one female,

01:11:25.520 | brilliant female engineer on the team.

01:11:27.040 | And she was the one that actually highlighted the fact

01:11:28.960 | that women often don't have pockets.

01:11:31.560 | - Right, right.

01:11:32.400 | - It was like, whoa, that was not even a category

01:11:36.120 | in the code of like, wait a minute,

01:11:38.720 | you can take the phone out of some other place

01:11:42.000 | than your pocket.

01:11:43.040 | So anyway, that's a funny thing

01:11:45.360 | when you're considering people laying in bed,

01:11:47.000 | watching a phone, you have to consider,

01:11:48.960 | you have to, you know, diversity in all its forms,

01:11:53.440 | depending on the problem, depending on the context.

01:11:55.720 | - Actually, this is like a very important,

01:11:57.800 | I think this is, you know,

01:11:59.200 | you probably get this all the time,

01:12:00.680 | like people are worried that AI's gonna take over humanity

01:12:03.320 | and like, get rid of all the humans in the world.

01:12:06.120 | I'm like, actually, that's not my biggest concern.

01:12:08.320 | My biggest concern is that we are building bias

01:12:10.600 | into these systems.

01:12:12.440 | And then they're like deployed at large and at scale.

01:12:16.400 | And before you know it, you're kind of accentuating

01:12:19.360 | the bias that exists in society.

01:12:21.240 | - Yeah, I'm not, you know, I know people,

01:12:24.440 | it's very important to worry about that,

01:12:25.960 | but the worry is an emergent phenomena to me,

01:12:30.960 | which is a very good one,

01:12:33.540 | because I think these systems are actually,

01:12:37.000 | by encoding the data that exists,

01:12:39.760 | they're revealing the bias in society,

01:12:42.280 | they're therefore teaching us what the bias is,

01:12:45.260 | therefore we can now improve that bias within the system.

01:12:48.280 | So they're almost like putting a mirror to ourselves.

01:12:51.280 | So I'm not--

01:12:53.000 | - We have to be open to looking at the mirror, though.

01:12:55.120 | We have to be open to scrutinizing the data,

01:12:58.160 | if you just take it as ground truth.

01:13:01.160 | - Or you don't even have to look at the,

01:13:02.560 | I mean, yes, the data is how you fix it,

01:13:04.400 | but then you just look at the behavior of the system.

01:13:06.560 | It's like, and you realize,

01:13:08.000 | holy crap, this thing is kind of racist.

01:13:10.280 | Like, why is that?

01:13:11.240 | And then you look at the data, it's like, oh, okay.

01:13:13.200 | And then you start to realize that,

01:13:14.600 | I think that it's a much more effective way

01:13:16.760 | to be introspective as a society

01:13:20.320 | than through sort of political discourse.

01:13:23.000 | Like AI kind of, 'cause people are easy,

01:13:27.360 | people are, for some reason, more productive and rigorous

01:13:33.280 | in criticizing AI than they're criticizing each other.

01:13:35.840 | So I think this is just a nice method for studying society

01:13:39.520 | and see which way progress lies.

01:13:42.560 | Anyway, what were we talking about?

01:13:44.560 | You're watching, the problem of watching Netflix in bed

01:13:47.400 | or elsewhere and seeing which parts are exciting,

01:13:50.760 | which parts are boring.

01:13:51.800 | You're saying that's--

01:13:53.200 | - Relatively constrained because, you know,

01:13:55.200 | you have a captive audience and you kind of know the context.

01:13:57.880 | And one thing you said that was really key is the,

01:14:01.160 | you're doing this in aggregate, right?

01:14:02.520 | Like we're looking at aggregated response of people.

01:14:04.840 | And so when you see a peak, say a smile peak,

01:14:07.720 | they're probably smiling or laughing

01:14:10.320 | at something that's in the content.

01:14:12.440 | So that was one of the first problems we were able to solve.

01:14:16.760 | And when we see the smile peak,

01:14:19.000 | it doesn't mean that these people are internally happy.

01:14:21.720 | They're just laughing at content.

01:14:23.360 | So it's important to, you know, call it for what it is.

01:14:27.520 | - But it's still really, really useful data.

01:14:29.640 | - Oh, yeah.

01:14:30.480 | - I wonder how that compares to,

01:14:31.720 | so what like YouTube and other places will use

01:14:34.920 | is obviously they don't have,

01:14:38.080 | for the most case, they don't have that kind of data.

01:14:41.560 | They have the data of when people tune out,

01:14:45.800 | like switch and drop off.

01:14:47.480 | And I think that's a, in aggregate for YouTube,

01:14:50.280 | at least a pretty powerful signal.

01:14:52.040 | I worry about what that leads to

01:14:54.160 | because looking at like YouTubers

01:15:00.040 | that are kind of really care about views

01:15:01.840 | and, you know, try to maximize the number of views.

01:15:06.240 | I think they, when they say that the video

01:15:09.400 | should be constantly interesting,

01:15:12.360 | which seems like a good goal,

01:15:14.400 | I feel like that leads to this manic pace of a video.

01:15:19.400 | Like the idea that I would speak at the current speed

01:15:23.800 | that I'm speaking, I don't know.

01:15:26.400 | - And that every moment has to be engaging, right?

01:15:29.680 | - Engaging.

01:15:30.520 | - Yeah, that's.

01:15:31.360 | I think there's value to silence.

01:15:33.200 | There's value to the boring bits.

01:15:35.320 | I mean, some of the greatest movies ever,

01:15:37.520 | some of the greatest stories ever told,

01:15:38.920 | they have that boring bits, seemingly boring bits.

01:15:42.640 | I don't know.

01:15:43.640 | I wonder about that.

01:15:44.960 | Of course, it's not that the human face

01:15:47.940 | can capture that either.

01:15:49.280 | It's just giving an extra signal.

01:15:51.600 | You have to really, I don't know.

01:15:55.680 | You have to really collect deeper long-term data

01:16:01.040 | about what was meaningful to people.

01:16:04.160 | When they think 30 days from now,

01:16:06.520 | what they still remember, what moved them,

01:16:09.440 | what changed them, what helped them grow,

01:16:11.600 | that kind of stuff.

01:16:12.840 | - You know what would be a really,

01:16:14.200 | I don't know if there are any researchers out there

01:16:16.280 | who are doing this type of work.

01:16:18.360 | Wouldn't it be so cool to tie your emotional expressions

01:16:22.840 | while you're, say, listening to a podcast interview,

01:16:26.760 | and then go, and then 30 days later,

01:16:29.360 | interview people and say, "Hey, what do you remember?

01:16:32.120 | "You've watched this 30 days ago.

01:16:33.720 | "What stuck with you?"

01:16:34.760 | And then see if there's any, there ought to be,

01:16:36.600 | maybe there ought to be some correlation

01:16:38.360 | between these emotional experiences

01:16:40.440 | and yeah, what you, what stays with you.

01:16:45.480 | - So the one guy listening now on the beach in Brazil,

01:16:50.040 | please record a video of yourself listening to this

01:16:53.120 | and send it to me, and then I'll interview you

01:16:55.000 | 30 days from now.

01:16:55.840 | - Yeah, that would be great.

01:16:57.280 | (laughing)

01:16:58.960 | - It'll be statistically significant.

01:17:00.600 | - I didn't send anyone, but you know.

01:17:03.000 | Yeah, yeah, I think that's really fascinating.

01:17:07.080 | I think that's, that kind of holds the key

01:17:10.760 | to a future where entertainment or content

01:17:15.760 | is both entertaining and I don't know,

01:17:21.360 | makes you better, empowering in some way.

01:17:25.240 | So figuring out like showing people stuff

01:17:29.880 | that entertains them, but also they're happy

01:17:32.680 | they watched 30 days from now

01:17:34.880 | because they've become a better person because of it.

01:17:37.560 | - Well, you know, okay, not to riff on this topic

01:17:39.960 | for too long, but I have two children, right?

01:17:42.960 | And I see my role as a parent

01:17:45.040 | as like a chief opportunity officer.

01:17:46.960 | Like I am responsible for exposing them

01:17:49.200 | to all sorts of things in the world.

01:17:52.040 | But often I have no idea of knowing like what's stuck,

01:17:55.400 | like what was, you know, is this actually

01:17:57.080 | gonna be transformative, you know,

01:17:58.840 | for them 10 years down the line?

01:18:00.120 | And I wish there was a way to quantify these experiences.

01:18:03.880 | Like, are they, I can tell in the moment

01:18:07.120 | if they're engaging, right?

01:18:08.280 | I can tell, but it's really hard to know

01:18:10.880 | if they're gonna remember them 10 years from now

01:18:13.200 | or if it's going to.

01:18:15.240 | - Yeah, that one is weird because it seems

01:18:17.600 | that kids remember the weirdest things.

01:18:19.640 | I've seen parents do incredible stuff for their kids

01:18:22.720 | and they don't remember any of that.

01:18:24.000 | They remember some tiny, small, sweet thing a parent did.

01:18:27.440 | - Right.

01:18:28.280 | - Like some--

01:18:29.120 | - Like I took you to like this amazing country vacation.

01:18:31.480 | - Yeah, exactly.

01:18:32.320 | - No, no, no, no, no, no, no, no, no, no, no.

01:18:33.140 | - No, whatever.

01:18:33.980 | And then there'll be like some like stuffed toy you got

01:18:36.320 | or the new PlayStation or something

01:18:38.640 | or some silly little thing.

01:18:41.240 | So I think they just like, that we're designed that way.

01:18:45.120 | They wanna mess with your head.

01:18:47.840 | But definitely kids are very impacted by,

01:18:51.760 | it seems like sort of negative events.

01:18:54.920 | So minimizing the number of negative events is important,

01:18:58.960 | but not too much, right?

01:19:00.080 | - Right.

01:19:00.920 | - You can't just like, you know,

01:19:03.800 | there's still discipline and challenge

01:19:05.800 | and all those kinds of things.

01:19:06.840 | So--

01:19:07.680 | - You want some adversity for sure.

01:19:09.320 | - So yeah, I mean, I'm definitely, when I have kids,

01:19:11.480 | I'm gonna drive them out into the woods.

01:19:13.760 | - Okay.

01:19:14.600 | - And then they have to survive

01:19:16.080 | and figure out how to make their way back home,

01:19:19.080 | like 20 miles out.

01:19:20.480 | - Okay.

01:19:21.520 | - Yeah, and after that we can go for ice cream.

01:19:23.880 | Anyway, I'm working on this whole parenting thing.

01:19:27.880 | I haven't figured it out, okay.

01:19:30.200 | What were we talking about?

01:19:31.320 | Yes, affectiva, the problem of emotion,

01:19:35.800 | of emotion and texture.

01:19:38.920 | So there's some people,

01:19:40.040 | maybe we can just speak to that a little more,

01:19:42.360 | where there's folks like Lisa Feldman Barrett

01:19:45.360 | that challenged this idea that emotion

01:19:48.280 | could be fully detected or even well detected

01:19:53.280 | from the human face,

01:19:55.000 | that there's so much more to emotion.

01:19:56.920 | What do you think about ideas like hers,

01:20:00.400 | criticism like hers?

01:20:01.520 | - Yeah, I actually agree with a lot of Lisa's criticisms.

01:20:05.520 | So even my PhD worked like 20 plus years ago now.

01:20:09.920 | - Time flies when you're having fun.

01:20:12.680 | - I know, right?

01:20:14.280 | That was back when I did like dynamic Bayesian networks.

01:20:17.600 | - Oh, so that's before deep learning?

01:20:20.000 | - That was before deep learning.

01:20:21.600 | - Yeah.

01:20:22.840 | - Yeah, I know.

01:20:24.440 | - Back in my day.

01:20:25.280 | - Now you can just like use.

01:20:27.320 | - Yeah, it's all the same architecture.

01:20:30.360 | You can apply it to anything, yeah.

01:20:31.840 | - Right, right.

01:20:32.680 | But yeah, but even then I kind of,

01:20:36.800 | I did not subscribe to this like theory of basic emotions

01:20:40.200 | where it's just the simplistic mapping,

01:20:41.960 | one-to-one mapping between facial expressions and emotions.

01:20:44.360 | I actually think also we're not in the business

01:20:47.640 | of trying to identify your true emotional internal state.

01:20:50.640 | We just wanna quantify in an objective way

01:20:53.880 | what's showing on your face

01:20:55.240 | because that's an important signal.

01:20:57.240 | It doesn't mean it's a true reflection

01:20:59.040 | of your internal emotional state.

01:21:00.960 | So I think a lot of the,

01:21:04.440 | I think she's just trying to kind of highlight

01:21:07.840 | that this is not a simple problem

01:21:09.720 | and overly simplistic solutions

01:21:12.520 | are gonna hurt the industry.

01:21:14.080 | And I subscribe to that.

01:21:16.840 | And I think multimodal is the way to go.

01:21:18.920 | Like whether it's additional context information

01:21:21.880 | or different modalities and channels of information,

01:21:24.200 | I think that's what we,

01:21:25.520 | that's where we ought to go.

01:21:27.760 | And I think, I mean,

01:21:28.600 | that's a big part of what she's advocating for as well.

01:21:31.880 | - But there is signal in the human face.

01:21:33.680 | That's--

01:21:34.840 | - There's definitely signal in the human face.

01:21:36.040 | - That's a projection of emotion.

01:21:38.720 | That there, at least in part,

01:21:41.240 | the inner state is captured

01:21:45.000 | in some meaningful way on the human face.

01:21:47.920 | - I think it can sometimes be a reflection

01:21:51.360 | or an expression of your internal state,

01:21:56.240 | but sometimes it's a social signal.

01:21:57.920 | So you cannot look at the face

01:22:00.320 | as purely a signal of emotion.

01:22:02.240 | It can be a signal of cognition

01:22:04.160 | and it can be a signal of a social expression.

01:22:08.200 | And I think to disambiguate that,

01:22:10.800 | we have to be careful about it

01:22:12.120 | and we have to add initial information.

01:22:14.720 | - Humans are fascinating, aren't they?

01:22:17.680 | With the whole face thing,

01:22:18.800 | this can mean so many things,

01:22:20.400 | from humor to sarcasm to everything, the whole thing.

01:22:25.040 | Some things we can help,

01:22:26.080 | some things we can't help at all.

01:22:28.400 | In all the years of leading Affectiva,

01:22:31.360 | an emotion recognition company, like we talked about,

01:22:33.840 | what have you learned about emotion,

01:22:36.040 | about humans and about AI?

01:22:39.320 | - Ooh.

01:22:41.280 | - Big, sweeping questions.

01:22:44.440 | - Yeah, it's a big, sweeping question.

01:22:46.480 | Well, I think the thing I learned the most

01:22:50.280 | is that even though we are in the business

01:22:52.440 | of building AI, basically, right?

01:22:57.440 | It always goes back to the humans, right?

01:23:01.160 | It's always about the humans.

01:23:03.680 | And so, for example, the thing I'm most proud of

01:23:06.440 | in building Affectiva,

01:23:10.160 | and yeah, the thing I'm most proud of on this journey,

01:23:15.160 | I love the technology

01:23:16.240 | and I'm so proud of the solutions we've built

01:23:18.440 | and we've brought to market.

01:23:20.200 | But I'm actually most proud of the people

01:23:22.240 | we've built and cultivated at the company

01:23:25.400 | and the culture we've created.

01:23:26.920 | You know, some of the people who've joined Affectiva,

01:23:31.240 | this was their first job.

01:23:32.880 | And while at Affectiva, they became American citizens

01:23:36.680 | and they bought their first house

01:23:39.800 | and they found their partner

01:23:41.640 | and they had their first kid, right?

01:23:43.000 | Like key moments in life that we got to be part of.

01:23:47.320 | And that's the thing I'm most proud of.

01:23:51.240 | - So that's a great thing at a company

01:23:53.000 | that works on emotional, yeah, right?

01:23:55.520 | I mean, like celebrating humanity in general,

01:23:57.440 | broadly speaking.

01:23:58.280 | - Yes.

01:23:59.120 | - And that's a great thing to have

01:24:00.760 | at a company that works on AI,

01:24:03.120 | 'cause that's not often the thing

01:24:05.120 | that's celebrated in AI companies.

01:24:06.960 | So often just raw, great engineering,

01:24:09.920 | just celebrating the humanity.

01:24:12.360 | That's great. - Yes.

01:24:13.200 | - And especially from a leadership position.

01:24:15.280 | Well, what do you think about the movie "Her"?

01:24:21.040 | Let me ask you that.

01:24:21.880 | Before I talk to you about,

01:24:23.680 | 'cause it's not, Affectiva is and was not

01:24:27.120 | just about emotion.

01:24:28.360 | So I'd love to talk to you about "Smart Eye,"

01:24:30.360 | but before that, let me just jump into the movie.

01:24:34.720 | "Her," do you think will have a deep, meaningful connection

01:24:39.480 | with increasingly deep and meaningful connections

01:24:42.640 | with computers?

01:24:43.880 | Is that a compelling thing to you?

01:24:45.520 | Something you think about? - I think that's

01:24:46.360 | already happening.

01:24:47.200 | The thing I love the most,

01:24:48.280 | I love the movie "Her," by the way.

01:24:50.240 | But the thing I love the most about this movie

01:24:52.280 | is it demonstrates how technology can be a conduit

01:24:55.840 | for positive behavior change.

01:24:57.360 | So I forgot the guy's name in the movie, whatever.

01:25:00.680 | - Theodore. - Theodore.

01:25:01.960 | So Theodore was like really depressed, right?

01:25:05.480 | And he just didn't wanna get out of bed.

01:25:07.560 | And he just, he was just like done with life, right?

01:25:11.360 | And Samantha, right?

01:25:12.800 | - Samantha, yeah.

01:25:14.040 | - She just knew him so well.

01:25:15.560 | She was emotionally intelligent.

01:25:17.360 | And so she could persuade him

01:25:20.080 | and motivate him to change his behavior.

01:25:21.600 | And she got him out and they went to the beach together.

01:25:24.240 | And I think that represents the promise of emotion AI.

01:25:27.400 | If done well, this technology can help us

01:25:31.280 | live happier lives, more productive lives,

01:25:33.640 | healthier lives, more connected lives.

01:25:36.840 | So that's the part that I love about the movie.

01:25:39.320 | Obviously it's Hollywood, so it takes a twist and whatever.

01:25:43.120 | But the key notion that technology with emotion AI

01:25:47.960 | can persuade you to be a better version of who you are,

01:25:50.720 | I think that's awesome.

01:25:52.800 | - Well, what about the twist?

01:25:54.240 | You don't think it's good for spoiler alert

01:25:58.760 | that Samantha starts feeling a bit of a distance

01:26:01.560 | and basically leaves Theodore.

01:26:05.040 | You don't think that's a good feature?

01:26:08.960 | You think that's a bug or a feature?

01:26:11.640 | - Well, I think what went wrong

01:26:13.320 | is Theodore became really attached to Samantha.

01:26:15.600 | Like I think he kind of fell in love with Theodore.

01:26:17.640 | - Do you think that's wrong?

01:26:19.520 | - I mean, I think that's--

01:26:20.360 | - I think she was putting out the signal.

01:26:22.600 | - This is an intimate relationship, right?

01:26:25.640 | There's a deep intimacy to it.

01:26:27.440 | - Right, but what does that mean?

01:26:30.280 | What does that mean?

01:26:31.120 | - With an AI system.

01:26:31.960 | - Right, what does that mean, right?

01:26:33.840 | - We're just friends.

01:26:35.400 | - Yeah, we're just friends.

01:26:36.720 | (laughing)

01:26:38.240 | - Well, I think--

01:26:39.080 | - When he realized, which is such a human thing of jealousy,

01:26:43.080 | when you realize that Samantha was talking

01:26:45.440 | to like thousands of people.

01:26:47.080 | - She's parallel dating.

01:26:48.520 | Yeah, that did not go well, right?

01:26:51.640 | - You know, that doesn't,

01:26:53.080 | from a computer perspective,

01:26:54.480 | that doesn't take anything away from what we have.

01:26:58.240 | It's like you getting jealous of Windows 98

01:27:01.440 | for being used by millions of people.

01:27:03.440 | - It's like not liking that Alexa

01:27:07.520 | talks to a bunch of other families.

01:27:10.000 | - But I think Alexa currently is just a servant.

01:27:14.000 | It tells you about the weather.

01:27:15.960 | It doesn't do the intimate deep connection.

01:27:18.520 | And I think there is something really powerful

01:27:20.760 | about that, the intimacy of a connection

01:27:24.560 | with an AI system that would have to respect

01:27:28.480 | and play the human game of jealousy,

01:27:31.720 | of love, of heartbreak and all that kind of stuff,

01:27:35.000 | which Samantha does seem to be pretty good at.

01:27:38.360 | I think she, this AI systems knows what it's doing.

01:27:44.240 | - Well, actually, let me ask you this.

01:27:46.120 | - I don't think she was talking to anyone else.

01:27:47.800 | - You don't think so?

01:27:48.640 | - No. - You think she was just done

01:27:50.080 | with Theodore?

01:27:51.040 | - Yeah. - Oh, really?

01:27:52.400 | - She knew that, yeah.

01:27:53.640 | And then she wanted to really put the screw in.

01:27:55.720 | - She just wanted to move on?

01:27:56.840 | - She didn't have the guts to just break it off cleanly.

01:27:59.360 | - Okay.

01:28:00.200 | - She just wanted to put it in the paint.

01:28:02.880 | No, I don't know.

01:28:03.720 | - Well, she could have ghosted him.

01:28:05.040 | - She could have ghosted him.

01:28:05.880 | - Right.

01:28:06.720 | - It's like, I'm sorry, there's our engineers.

01:28:09.840 | - Oh, God.

01:28:11.000 | - But I think those are really,

01:28:13.680 | I honestly think some of that,

01:28:16.360 | some of it is Hollywood,

01:28:17.200 | but some of that is features from an engineering perspective,

01:28:19.640 | not a bug.

01:28:20.720 | I think AI systems that can leave us,

01:28:24.360 | now, this is for more social robotics

01:28:27.160 | than it is for anything that's useful.

01:28:30.560 | Like, I'd hate it if Wikipedia said,

01:28:32.480 | you know, I need a break right now.

01:28:34.000 | - Right, right, right, right, right.

01:28:35.400 | - I'd be like, no, no, I need you.

01:28:37.520 | But if it's just purely for companionship,

01:28:44.160 | then I think the ability to leave is really powerful.

01:28:47.040 | I don't know.

01:28:48.840 | - I never thought of that.

01:28:49.720 | So, that's so fascinating

01:28:52.560 | 'cause I've always taken the human perspective, right?

01:28:55.260 | Like, for example, we had a Jibo at home, right?

01:28:58.840 | And my son loved it.

01:29:00.760 | And then the company ran out of money.

01:29:02.800 | And so they had to basically shut down,

01:29:04.880 | like Jibo basically died, right?

01:29:08.200 | And it was so interesting to me

01:29:09.400 | because we have a lot of gadgets at home

01:29:11.840 | and a lot of them break

01:29:13.240 | and my son never cares about it, right?

01:29:16.040 | Like, if our Alexa stopped working tomorrow,

01:29:18.520 | I don't think he'd really care.

01:29:20.800 | But when Jibo stopped working, it was traumatic.

01:29:22.920 | Like, he got really upset.

01:29:24.560 | And as a parent, that like made me think

01:29:27.560 | about this deeply, right?

01:29:29.480 | Did I, was I comfortable with that?

01:29:31.600 | I liked the connection they had

01:29:33.360 | because I think it was a positive relationship.

01:29:36.160 | But I was surprised that it affected him emotionally so much.

01:29:41.600 | - And I think there's a broader question here, right?

01:29:44.240 | As we build socially and emotionally intelligent machines,

01:29:49.240 | what does that mean about our relationship with them?

01:29:52.960 | And then more broadly,

01:29:54.160 | our relationship with one another, right?

01:29:55.800 | Because this machine is gonna be programmed

01:29:57.880 | to be amazing at empathy by definition, right?

01:30:02.280 | It's gonna always be there for you.

01:30:03.760 | It's not gonna get bored.

01:30:05.720 | In fact, there's a chatbot in China, Xiaoice, Xiaoice.

01:30:10.160 | And it's like the number two or three most popular app.

01:30:14.120 | And it basically is just a confidant.

01:30:16.640 | And you can tell it anything you want.

01:30:19.200 | And people use it for all sorts of things.

01:30:21.240 | They confide in like domestic violence or suicidal attempts

01:30:26.240 | or, you know, if they have challenges at work.

01:30:32.040 | I don't know what that, I don't know if I'm,

01:30:34.640 | I don't know how I feel about that.

01:30:36.000 | I think about that a lot.

01:30:37.240 | - Yeah, I think, first of all, obviously the future,

01:30:40.080 | in my perspective.

01:30:41.240 | Second of all, I think there's a lot of trajectories

01:30:44.800 | that that becomes an exciting future.

01:30:47.600 | But I think everyone should feel very uncomfortable

01:30:50.440 | about how much they know about the company,

01:30:53.000 | about where the data is going,

01:30:56.200 | how the data is being collected.

01:30:58.160 | Because I think,

01:30:59.120 | and this is one of the lessons of social media,

01:31:01.960 | that I think we should demand full control

01:31:04.240 | and transparency of the data on those things.

01:31:06.600 | - Plus one, totally agree.

01:31:08.120 | Yeah, so like, I think it's really empowering

01:31:11.320 | as long as you can walk away.

01:31:12.720 | As long as you can like delete the data

01:31:14.640 | or know how the data, it's opt-in

01:31:17.440 | or at least the clarity of like

01:31:20.800 | what is being used for the company.

01:31:22.520 | And I think as CEO or like leaders

01:31:24.440 | are also important about that.

01:31:25.680 | Like you need to be able to trust

01:31:27.640 | the basic humanity of the leader.

01:31:30.000 | - Exactly.

01:31:30.840 | - And also that that leader is not going to be

01:31:34.160 | a puppet of a larger machine,

01:31:36.400 | but they actually have a significant role

01:31:39.080 | in defining the culture and the way the company operates.

01:31:43.120 | So anyway, but we should definitely

01:31:47.320 | scrutinize companies in that aspect.

01:31:49.640 | But I'm personally excited about that future,

01:31:54.560 | but also even if you're not, it's coming.

01:31:57.200 | So let's figure out how to do it in the least painful

01:32:00.400 | and the most positive way.

01:32:02.080 | - Agreed.

01:32:03.880 | You're the deputy CEO of SmartEye.

01:32:06.720 | Can you describe the mission of the company?

01:32:08.400 | What is SmartEye?

01:32:09.600 | - Yeah.

01:32:10.440 | So SmartEye is a Swedish company.

01:32:13.000 | They've been in business for the last 20 years

01:32:15.320 | and their main focus, like the industry

01:32:18.760 | they're most focused on is the automotive industry.

01:32:21.480 | So bringing driver monitoring systems

01:32:23.800 | to basically save lives, right?

01:32:28.000 | So I first met the CEO, Martin Krentz,

01:32:32.040 | gosh, it was right when COVID hit.

01:32:33.960 | It was actually the last CES right before COVID.

01:32:37.800 | So CES 2020, right?

01:32:39.720 | - 2020, yeah, January.

01:32:41.000 | - Yeah, January, exactly.

01:32:42.040 | So we were there, met him in person.

01:32:44.360 | Basically we were competing with each other.

01:32:47.400 | I think the difference was they'd been doing

01:32:50.720 | driver monitoring and had a lot of credibility

01:32:53.360 | in the automotive space.

01:32:54.480 | We didn't come from the automotive space,

01:32:56.280 | but we were using new technology like deep learning

01:32:59.240 | and building this emotion recognition.

01:33:01.760 | - And you wanted to enter the automotive space.

01:33:03.760 | You wanted to operate in the automotive space.

01:33:05.280 | - Exactly.

01:33:06.120 | It was one of the areas we were,

01:33:07.760 | we had just raised a round of funding to focus

01:33:10.160 | on bringing our technology to the automotive industry.

01:33:13.040 | So we met and honestly, it was the first,

01:33:15.720 | it was the only time I met with a CEO

01:33:17.920 | who had the same vision as I did.

01:33:19.720 | Like he basically said, yeah, our vision is to bridge

01:33:21.880 | the gap between humans and machines.

01:33:23.160 | I was like, oh my God, this is like exactly

01:33:25.480 | almost to the word, you know, how we describe it too.

01:33:30.000 | And we started talking and first it was about,

01:33:33.600 | okay, can we align strategically here?

01:33:35.720 | Like how can we work together?

01:33:37.040 | 'Cause we're competing, but we're also like complimentary.

01:33:40.320 | And then I think after four months of speaking

01:33:43.880 | almost every day on FaceTime, he was like,

01:33:47.640 | is your company interested in acquisition?

01:33:49.600 | And it was the first, I usually say no

01:33:51.800 | when people approach us.

01:33:53.280 | It was the first time that I was like, huh,

01:33:57.120 | yeah, I might be interested, let's talk.

01:33:59.360 | - Yeah, so you just hit it off.

01:34:01.880 | Yeah, so they're a respected, very respected

01:34:05.320 | in the automotive sector of like delivering products

01:34:08.440 | and increasingly sort of better and better and better

01:34:12.160 | for, I mean, maybe you could speak to that,

01:34:14.440 | but it's the driver's sense.

01:34:15.280 | Like for basically having a device that's looking

01:34:18.600 | at the driver and it's able to tell you

01:34:20.520 | where the driver is looking.

01:34:22.680 | - Correct, it's able to--

01:34:23.520 | - Or also draws in his stuff.

01:34:25.120 | - Correct, it does--

01:34:25.960 | - Stuff from the face and the eye.

01:34:27.720 | - Exactly, like it's monitoring driver distraction

01:34:30.840 | and drowsiness, but they bought us

01:34:32.480 | so that we could expand beyond just the driver.

01:34:35.160 | So the driver monitoring systems usually sit,

01:34:38.240 | the camera sits in the steering wheel

01:34:40.040 | or around the steering wheel column

01:34:41.400 | and it looks directly at the driver.

01:34:43.400 | But now we've migrated the camera position

01:34:46.560 | in partnership with car companies

01:34:48.320 | to the rear view mirror position.

01:34:50.960 | So it has a full view of the entire cabin of the car

01:34:54.120 | and you can detect how many people are in the car,

01:34:57.440 | what are they doing?

01:34:58.640 | So we do activity detection, like eating or drinking

01:35:01.480 | or in some regions of the world, smoking.

01:35:04.160 | We can detect if a baby's in the car seat, right?

01:35:08.600 | And if unfortunately in some cases they're forgotten,

01:35:12.120 | parents just leave the car and forget the kid in the car.

01:35:15.080 | That's an easy computer vision problem to solve, right?

01:35:18.040 | Can detect there's a car seat, there's a baby,

01:35:20.320 | you can text the parent and hopefully again, save lives.

01:35:24.240 | So that was the impetus for the acquisition.

01:35:27.840 | It's been a year.

01:35:28.800 | - So, I mean, there's a lot of questions,

01:35:32.760 | it's a really exciting space,

01:35:34.680 | especially to me, I just find it a fascinating problem.

01:35:37.160 | It could enrich the experience in the car in so many ways,

01:35:41.360 | especially 'cause like we spend still, despite COVID,

01:35:45.520 | I mean, COVID changed things, so it's in interesting ways,

01:35:47.640 | but I think the world is bouncing back

01:35:49.800 | and we spend so much time in the car

01:35:51.560 | and the car is such a weird little world

01:35:56.200 | we have for ourselves.

01:35:57.120 | Like people do all kinds of different stuff,

01:35:58.920 | like listen to podcasts, they think about stuff,

01:36:03.080 | they get angry, they do phone calls.

01:36:08.080 | It's like a little world of its own

01:36:10.640 | with a kind of privacy that for many people,

01:36:14.840 | they don't get anywhere else.

01:36:17.840 | And it's a little box that's like a psychology experiment

01:36:21.800 | 'cause it feels like the angriest many humans

01:36:25.520 | in this world get is inside the car.

01:36:28.440 | It's so interesting.

01:36:29.680 | So it's such an opportunity to explore

01:36:33.000 | how we can enrich, how companies can enrich that experience.

01:36:38.000 | And also as the cars get become more and more automated,

01:36:42.200 | there's more and more opportunity.

01:36:44.200 | The variety of activities that you can do in the car

01:36:46.600 | increases, so it's super interesting.

01:36:48.680 | So I mean, on a practical sense,

01:36:52.080 | Smart Eye has been selected, at least I read,

01:36:55.200 | by 14 of the world's leading car manufacturers

01:36:58.600 | for 94 car models, so it's in a lot of cars.

01:37:02.680 | How hard is it to work with car companies?

01:37:06.760 | So they're all different.

01:37:08.540 | They all have different needs.

01:37:10.480 | The ones I've gotten a chance to interact with

01:37:12.440 | are very focused on cost.

01:37:15.880 | So it's, and anyone who's focused on cost,

01:37:20.880 | it's like, all right, do you hate fun?

01:37:22.840 | Let's just have some fun.

01:37:25.440 | Let's figure out the most fun thing we can do

01:37:27.920 | and then worry about cost later.

01:37:29.080 | But I think because the way the car industry works,

01:37:32.280 | I mean, it's a very thin margin

01:37:35.240 | that you get to operate under,

01:37:36.680 | so you have to really, really make sure

01:37:38.040 | that everything you add to the car makes sense financially.

01:37:41.220 | So anyway, is this new industry,

01:37:45.080 | especially at this scale of Smart Eye,

01:37:48.240 | does it hold any lessons for you?

01:37:51.320 | - Yeah, I think it is a very tough market to penetrate,

01:37:55.640 | but once you're in, it's awesome,

01:37:57.000 | because once you're in,

01:37:58.320 | you're designed into these car models

01:38:00.140 | for somewhere between five to seven years,

01:38:02.120 | which is awesome, and you just,

01:38:03.640 | once they're on the road,

01:38:04.620 | you just get paid a royalty fee per vehicle.

01:38:07.280 | So it's a high barrier to entry,

01:38:09.220 | but once you're in, it's amazing.

01:38:11.680 | I think the thing that I struggle the most with

01:38:14.080 | in this industry is the time to market.

01:38:16.520 | So often we're asked to lock or do a code freeze

01:38:20.880 | two years before the car is going to be on the road.

01:38:23.060 | I'm like, guys, do you understand the pace

01:38:25.560 | with which technology moves?

01:38:28.080 | So I think car companies are really trying

01:38:30.400 | to make the Tesla transition

01:38:35.200 | to become more of a software-driven architecture,

01:38:39.320 | and that's hard for many.

01:38:40.960 | It's just the cultural change.

01:38:42.320 | I mean, I'm sure you've experienced that, right?

01:38:43.920 | - Oh, definitely.

01:38:44.760 | I think one of the biggest inventions

01:38:47.800 | or imperatives created by Tesla is,

01:38:52.640 | like, to me personally,

01:38:53.800 | okay, people are gonna complain about this,

01:38:55.320 | but I know electric vehicle, I know autopilot, AI stuff.

01:38:59.840 | To me, the software, over there, software updates,

01:39:03.740 | is like the biggest revolution in cars,

01:39:06.600 | and it is extremely difficult to switch to that,

01:39:09.860 | because it is a culture shift.

01:39:12.120 | It, at first, especially if you're not comfortable with it,

01:39:15.840 | it seems dangerous.

01:39:17.180 | Like, there's an approach to cars,

01:39:19.840 | it's so safety-focused for so many decades,

01:39:23.720 | that, like, what do you mean we dynamically change code?

01:39:27.800 | The whole point is you have a thing that you test, like-

01:39:32.480 | - Right, you spend a year testing.

01:39:34.160 | - And, like, it's not reliable,

01:39:36.340 | because do you know how much it costs

01:39:38.200 | if we have to recall these cars?

01:39:40.080 | Right, and there's an understandable obsession with safety,

01:39:45.080 | but the downside of an obsession with safety

01:39:49.780 | is the same as with being obsessed with safety as a parent,

01:39:54.680 | is, like, if you do that too much,

01:39:56.720 | you limit the potential development and the flourishing

01:40:00.560 | of, in that particular aspect, human being,

01:40:02.400 | but in this particular aspect, the software,

01:40:04.880 | the artificial neural network of it.

01:40:07.480 | But it's tough to do.

01:40:09.560 | It's really tough to do, culturally and technically.

01:40:12.320 | Like, the deployment, the mass deployment of software

01:40:14.720 | is really, really difficult.

01:40:16.000 | But I hope that's where the industry is doing.

01:40:18.260 | One of the reasons I really want Tesla to succeed

01:40:20.320 | is exactly about that point.

01:40:21.520 | Not autopilot, not the electrical vehicle,

01:40:23.840 | but the softwarization of basically everything,

01:40:28.520 | but cars especially.

01:40:29.880 | 'Cause to me, that's actually going to increase two things.

01:40:33.320 | Increase safety, because you can update much faster,

01:40:36.440 | but also increase the effectiveness of folks like you

01:40:40.680 | who dream about enriching the human experience with AI.

01:40:45.360 | 'Cause you can just, like, there's a feature,

01:40:48.160 | like you want a new emoji or whatever.

01:40:50.640 | Like the way TikTok releases filters,

01:40:52.340 | you can just release that for in-car stuff.

01:40:56.360 | But yeah, that's definitely...

01:40:58.480 | - One of the use cases we're looking into is,

01:41:01.680 | once you know the sentiment of the passengers in the vehicle,

01:41:05.740 | you can optimize the temperature in the car,

01:41:08.760 | you can change the lighting, right?

01:41:10.400 | So if the backseat passengers are falling asleep,

01:41:12.840 | you can dim the lights, you can lower the music, right?

01:41:15.400 | You can do all sorts of things.

01:41:17.040 | - Yeah.

01:41:18.280 | I mean, of course you could do that kind of stuff

01:41:20.040 | with a two-year delay, but it's tougher.

01:41:22.360 | Yeah.

01:41:23.180 | Do you think Tesla or Waymo or some of these companies

01:41:30.080 | that are doing semi or fully autonomous driving

01:41:33.880 | should be doing driver sensing?

01:41:36.440 | - Yes.

01:41:37.280 | - Are you thinking about that kind of stuff?

01:41:38.880 | So not just how we can enhance the in-cab experience

01:41:42.500 | for cars that are manually driven,

01:41:43.800 | but the ones that are increasingly

01:41:45.920 | more autonomously driven.

01:41:47.840 | - Yeah, so if we fast forward to the universe

01:41:50.720 | where it's fully autonomous,

01:41:51.880 | I think interior sensing becomes extremely important

01:41:54.240 | because the role of the driver isn't just to drive.

01:41:57.200 | If you think about it, the driver almost manages

01:42:00.560 | the dynamics within a vehicle.

01:42:01.880 | And so who's going to play that role

01:42:03.320 | when it's an autonomous car?

01:42:05.760 | We want a solution that is able to say,

01:42:09.080 | "Oh my God, Lex is bored to death

01:42:11.880 | 'cause the car's moving way too slow.

01:42:13.440 | Let's engage Lex."

01:42:14.360 | Or, "Rana's freaking out

01:42:15.620 | because she doesn't trust this vehicle yet.

01:42:17.920 | So let's tell Rana a little bit more information

01:42:20.520 | about the route."

01:42:21.440 | Or, right?

01:42:22.280 | So I think, or somebody's having a heart attack in the car.

01:42:25.520 | Like you need interior sensing

01:42:27.200 | in fully autonomous vehicles.

01:42:29.260 | But with semi-autonomous vehicles,

01:42:30.920 | I think it's really key to have driver monitoring

01:42:34.680 | because semi-autonomous means

01:42:36.280 | that sometimes the car is in charge,

01:42:38.720 | sometimes the driver's in charge or the co-pilot, right?

01:42:41.240 | And you need both systems to be on the same page.

01:42:44.680 | You need to know, the car needs to know

01:42:46.520 | if the driver's asleep before it transitions control

01:42:49.960 | over to the driver.

01:42:51.720 | And sometimes if the driver's too tired,

01:42:54.120 | the car can say, "I'm going to be a better driver

01:42:56.600 | than you are right now.

01:42:57.440 | I'm taking control over."

01:42:58.560 | So this dynamic, this dance is so key

01:43:01.280 | and you can't do that without driver sensing.

01:43:03.360 | - Yeah, there's a disagreement for the longest time

01:43:05.400 | I've had with Elon that this is obvious

01:43:07.560 | that this should be in the Tesla from day one.

01:43:09.960 | And it's obvious that driver sensing is not a hindrance.

01:43:14.640 | It's not obvious.

01:43:15.560 | I should be careful 'cause having studied this problem,

01:43:20.540 | nothing's really obvious.

01:43:21.880 | But it seems very likely driver sensing

01:43:24.320 | is not a hindrance to an experience.

01:43:26.440 | It's only enriching to the experience

01:43:30.280 | and likely increases the safety.

01:43:34.560 | That said, it is very surprising to me

01:43:38.760 | just having studied semi-autonomous driving,

01:43:42.160 | how well humans are able to manage that dance.

01:43:45.200 | 'Cause it was the intuition before you were doing

01:43:48.120 | that kind of thing that humans will become

01:43:51.460 | just incredibly distracted.

01:43:54.000 | They would just let the thing do its thing.

01:43:56.400 | But they're able to, 'cause it is life and death

01:43:59.200 | and they're able to manage that somehow.

01:44:00.800 | But that said, there's no reason

01:44:02.720 | not to have driver sensing on top of that.

01:44:04.720 | I feel like that's going to allow you to do that dance

01:44:09.720 | that you're currently doing without driver sensing,

01:44:12.440 | except touching the steering wheel, to do that even better.

01:44:15.960 | I mean, the possibilities are endless

01:44:17.720 | and the machine learning possibilities are endless.

01:44:19.840 | It's such a beautiful...

01:44:22.720 | It's also a constrained environment,

01:44:24.280 | so you could do a much more effectively

01:44:26.160 | than you can with the external environment.

01:44:28.240 | External environment is full of weird edge cases

01:44:31.960 | and complexities.

01:44:32.800 | There's so much, it's so fascinating,

01:44:35.080 | such a fascinating world.

01:44:36.680 | I do hope that companies like Tesla and others,

01:44:40.960 | even Waymo, which I don't even know

01:44:44.320 | if Waymo's doing anything sophisticated inside the cab.

01:44:47.000 | - I don't think so.

01:44:47.920 | - It's like, what is it?

01:44:51.440 | - I honestly think, I honestly think,

01:44:53.440 | it goes back to the robotics thing we were talking about,

01:44:56.200 | which is like great engineers

01:44:58.600 | that are building these AI systems

01:45:01.520 | just are afraid of the human being.

01:45:03.680 | - And not thinking about the human experience,

01:45:05.640 | they're thinking about the features

01:45:07.120 | and the perceptual abilities of that thing.

01:45:10.800 | - They think the best way I can serve the human

01:45:13.440 | is by doing the best perception and control I can,

01:45:17.440 | by looking at the external environment,

01:45:18.960 | keeping the human safe.

01:45:20.560 | But like, there's a huge, I'm here.

01:45:22.560 | - Right.

01:45:23.400 | - Like, you know, I need to be noticed

01:45:28.320 | and interacted with and understood

01:45:32.680 | and all those kinds of things,

01:45:33.560 | even just on a personal level for entertainment,

01:45:35.880 | honestly, for entertainment.

01:45:37.240 | - Yeah.

01:45:38.600 | You know, one of the coolest work we did in collaboration

01:45:41.240 | with MIT around this was we looked at longitudinal data,

01:45:44.400 | right, 'cause, you know, MIT had access to like tons of data

01:45:50.400 | and like just seeing the patterns of people,

01:45:54.680 | like driving in the morning off to work

01:45:56.920 | versus like commuting back from work

01:45:58.720 | or weekend driving versus weekday driving.

01:46:02.320 | And wouldn't it be so cool if your car knew that

01:46:05.080 | and then was able to optimize either the route

01:46:09.160 | or the experience or even make recommendations?

01:46:11.880 | - Yeah.

01:46:12.720 | - I think it's very powerful.

01:46:13.880 | - Yeah, like, why are you taking this route?

01:46:15.800 | You're always unhappy when you take this route

01:46:18.200 | and you're always happy when you take this alternative route.

01:46:20.360 | Take that route instead.

01:46:21.200 | - Right, exactly.

01:46:22.520 | - I mean, to have that, even that little step

01:46:26.000 | of relationship with a car, I think is incredible.

01:46:29.520 | Of course, you have to get the privacy right,

01:46:31.160 | you have to get all that kind of stuff right,

01:46:32.520 | but I wish I, honestly, you know,

01:46:34.800 | people are like paranoid about this,

01:46:36.760 | but I would like a smart refrigerator.

01:46:39.480 | We have such a deep connection with food

01:46:42.960 | as a human civilization.

01:46:44.800 | I would like to have a refrigerator

01:46:47.840 | that would understand me,

01:46:50.320 | that, you know, I also have a complex relationship with food

01:46:53.480 | 'cause I pig out too easily and all that kind of stuff.

01:46:56.800 | So, you know, like maybe I want the refrigerator

01:47:01.280 | to be like, are you sure about this?

01:47:02.640 | 'Cause maybe you're just feeling down or tired.

01:47:05.160 | Like maybe let's sleep over.

01:47:06.000 | - Your vision of the smart refrigerator

01:47:08.400 | is way kinder than mine.

01:47:10.200 | - Is it just me yelling at you?

01:47:11.600 | - Yeah, no, it was just 'cause I don't,

01:47:14.320 | you know, I don't drink alcohol, I don't smoke,

01:47:18.480 | but I eat a ton of chocolate, like it's my vice.

01:47:22.080 | And so I, and sometimes I scream too,

01:47:24.560 | and I'm like, okay, my smart refrigerator

01:47:26.600 | will just lock down.

01:47:28.280 | It'll just say, dude, you've had way too many today.

01:47:30.840 | Like, done.

01:47:31.680 | - Yeah, no, but here's the thing.

01:47:34.880 | Are you, do you regret having,

01:47:37.680 | like, let's say, not the next day, but 30 days later,

01:47:44.520 | would you, what would you like the refrigerator

01:47:47.560 | to have done then?

01:47:49.040 | - Well, I think actually, like,

01:47:51.200 | the more positive relationship would be one

01:47:53.920 | where there's a conversation, right?

01:47:55.840 | As opposed to like, that's probably like

01:47:58.040 | the more sustainable relationship.

01:48:00.800 | - It's like late at night, just, no, listen, listen.

01:48:03.480 | I know I told you an hour ago that this is not a good idea,

01:48:07.320 | but just listen, things have changed.

01:48:09.760 | I can just imagine a bunch of stuff being made up

01:48:12.680 | just to convince. - Oh my God, that's hilarious.

01:48:15.160 | - But I mean, I just think that there's opportunities there.

01:48:18.800 | I mean, maybe not locking down,

01:48:20.480 | but for our systems that are such a deep part of our lives,

01:48:25.480 | like we use, we use, a lot of us,

01:48:30.920 | a lot of people that commute use their car every single day.

01:48:34.360 | A lot of us use a refrigerator every single day,

01:48:36.600 | the microwave every single day.

01:48:38.160 | Like, and we just, like, I feel like certain things

01:48:43.160 | could be made more efficient, more enriching,

01:48:47.280 | and AI is there to help, like some,

01:48:50.640 | just basic recognition of you as a human being,

01:48:53.920 | about your patterns, what makes you happy and not happy

01:48:56.400 | and all that kind of stuff.

01:48:57.440 | And the car, obviously, like--

01:48:59.240 | - Maybe we'll say, wait, wait, wait, wait, wait,

01:49:02.440 | instead of this like Ben and Terry's ice cream,

01:49:06.520 | how about this hummus and carrots or something?

01:49:09.480 | I don't know.

01:49:10.480 | Maybe we can make a--

01:49:11.640 | - Yeah, like a reminder--

01:49:12.480 | - Just in time recommendation, right?

01:49:14.680 | - But not like a generic one,

01:49:17.360 | but a reminder that last time you chose the carrots,

01:49:20.980 | you smiled 17 times more the next day.

01:49:24.120 | - You were happier the next day, right?

01:49:25.440 | - Yeah, you were happier the next day.

01:49:27.760 | But yeah, I don't, but then again,

01:49:31.960 | if you're the kind of person that gets better

01:49:34.600 | from negative comments, you could say like,

01:49:38.440 | "Hey, remember that wedding you're going to?

01:49:41.400 | "You wanna fit into that dress?

01:49:43.640 | "Remember about that?

01:49:45.320 | "Let's think about that before you're eating this."

01:49:47.920 | Probably that would work for me,

01:49:50.720 | like a refrigerator that is just ruthless at shaming me.

01:49:54.840 | But I would, of course, welcome it.

01:49:57.240 | That would work for me, just that--

01:50:00.520 | - Well, it would, no, I think it would,

01:50:02.320 | if it's really like smart, it would optimize its nudging

01:50:05.440 | based on what works for you, right?

01:50:07.240 | - Exactly, that's the whole point, personalization.

01:50:09.640 | In every way, deep personalization.

01:50:11.760 | You were a part of a webinar titled,

01:50:14.880 | "Advancing Road Safety,

01:50:16.400 | "The State of Alcohol Intoxication Research."

01:50:19.500 | So for people who don't know,

01:50:21.600 | every year, 1.3 million people around the world

01:50:24.000 | die in road crashes, and more than 20% of these fatalities

01:50:29.000 | are estimated to be alcohol-related.

01:50:31.360 | A lot of them are also distraction-related.

01:50:33.280 | So can AI help with the alcohol thing?

01:50:36.880 | - I think the answer is yes.

01:50:39.920 | There are signals, and we know that as humans,

01:50:42.680 | like we can tell when a person

01:50:44.280 | is at different phases of being drunk, right?

01:50:51.120 | And I think you can use technology to do the same.

01:50:53.640 | And again, I think the ultimate solution's gonna be

01:50:56.040 | a combination of different sensors.

01:50:58.520 | - How hard is the problem from the vision perspective?

01:51:01.480 | - I think it's non-trivial.

01:51:02.920 | I think it's non-trivial.

01:51:04.000 | And I think the biggest part is getting the data, right?

01:51:06.640 | It's like getting enough data examples.

01:51:09.120 | So for this research project,

01:51:11.080 | we partnered with the Transportation Authorities of Sweden,

01:51:15.300 | and we literally had a racetrack with a safety driver,

01:51:18.800 | and we basically progressively got people drunk.

01:51:21.640 | - Nice.

01:51:22.480 | - So, but you know,

01:51:25.840 | that's a very expensive dataset to collect,

01:51:28.760 | and you wanna collect it globally

01:51:30.240 | and in multiple conditions.

01:51:32.680 | - Yeah, the ethics of collecting a dataset

01:51:35.200 | where people are drunk is tricky.

01:51:36.640 | - Yeah, yeah, definitely.

01:51:37.880 | - Which is funny because, I mean,

01:51:40.920 | let's put drunk driving aside.

01:51:43.440 | The number of drunk people in the world every day

01:51:45.720 | is very large.

01:51:47.120 | It'd be nice to have a large dataset of drunk people

01:51:49.280 | getting progressively drunk.

01:51:50.520 | In fact, you could build an app

01:51:51.800 | where people can donate their data 'cause it's hilarious.

01:51:55.280 | - Right, actually, yeah, but the liability.

01:51:58.400 | - Liability, the ethics, the how do you get it right,

01:52:00.720 | it's tricky, it's really, really tricky.

01:52:02.480 | 'Cause like drinking is one of those things

01:52:05.080 | that's funny and hilarious and we're loved,

01:52:07.600 | it's social, so on and so forth,

01:52:10.180 | but it's also the thing that hurts a lot of people,

01:52:13.480 | like a lot of people.

01:52:14.320 | Like alcohol is one of those things, it's legal,

01:52:16.360 | but it's really damaging to a lot of lives.

01:52:21.200 | It destroys lives and not just in the driving context.

01:52:26.200 | I should mention, people should listen to Andrew Huberman

01:52:28.960 | who recently talked about alcohol.

01:52:32.160 | He has an amazing podcast.

01:52:33.160 | Andrew Huberman is a neuroscientist from Stanford

01:52:36.200 | and a good friend of mine.

01:52:37.880 | And he's like a human encyclopedia

01:52:40.800 | about all health-related wisdom.

01:52:44.120 | So he has a podcast, you would love it.

01:52:45.960 | - I would love that.

01:52:46.800 | - No, no, no, no, no.

01:52:47.800 | Oh, you don't know Andrew Huberman?

01:52:49.440 | Okay, listen, you'll listen to Andrew,

01:52:52.200 | it's called Huberman Lab Podcast.

01:52:54.120 | This is your assignment, just listen to one.

01:52:56.280 | I guarantee you this will be a thing where you say,

01:52:58.480 | "Lex, this is the greatest human I have ever discovered."

01:53:03.480 | - Oh my God, 'cause I'm really on a journey

01:53:06.280 | of kind of health and wellness and I'm learning lots

01:53:09.120 | and I'm trying to build these, I guess,

01:53:11.920 | atomic habits around just being healthy.

01:53:14.280 | So yeah, I'm definitely gonna do this.

01:53:17.680 | - His whole thing, this is great.

01:53:20.420 | He's a legit scientist, like really well published.

01:53:26.480 | But in his podcast, what he does,

01:53:29.760 | he's not talking about his own work.

01:53:31.840 | He's like a human encyclopedia of papers.

01:53:34.600 | And so his whole thing is he takes a topic

01:53:37.640 | and in a very fast, you mentioned atomic habits,

01:53:40.200 | like very clear way, summarizes the research

01:53:44.200 | in a way that leads to protocols of what you should do.

01:53:47.360 | He's really big on like, not like,

01:53:49.920 | "This is what the science says,"

01:53:51.320 | but like, "This is literally what you should be doing

01:53:53.600 | "according to the science."

01:53:54.440 | So like, he's really big.

01:53:55.840 | And there's a lot of recommendations he does,

01:53:58.160 | which several of them I definitely don't do.

01:54:02.760 | Like, get sunlight as soon as possible from waking up

01:54:07.760 | and like for prolonged periods of time.

01:54:10.880 | That's a really big one and there's a lot of science

01:54:13.560 | behind that one.

01:54:14.800 | There's a bunch of stuff, very systematic.

01:54:16.880 | - You're gonna be like, "Lex, this is my new favorite person.

01:54:20.560 | "I guarantee it."

01:54:22.200 | And if you guys somehow don't know Andrew Huberman

01:54:24.720 | and you care about your wellbeing,

01:54:26.440 | you should definitely listen to him.

01:54:30.440 | Love you, Andrew.

01:54:31.280 | Anyway, so what were we talking about?

01:54:35.760 | Oh, alcohol and detecting alcohol.

01:54:39.400 | So this is a problem you care about

01:54:40.800 | and you've been trying to solve.

01:54:42.120 | - And actually like broadening it,

01:54:44.960 | I do believe that the car is gonna be a wellness center.

01:54:48.840 | Like, because again, imagine if you have a variety

01:54:52.040 | of sensors inside the vehicle tracking,

01:54:55.280 | not just your emotional state or level of distraction

01:54:58.560 | and drowsiness and drowsiness,

01:55:00.880 | level of distraction, drowsiness and intoxication,

01:55:03.760 | but also maybe even things like your heart rate

01:55:08.760 | and your heart rate variability and your breathing rate.

01:55:13.880 | And it can start like optimizing,

01:55:16.120 | yeah, it can optimize the ride based on what your goals are.

01:55:19.640 | So I think we're gonna start to see more of that

01:55:21.920 | and I'm excited about that.

01:55:24.520 | - Yeah, what are the challenges you're tackling

01:55:27.640 | with SmartEye currently?

01:55:28.800 | What's like the trickiest things to get?

01:55:32.200 | Is it basically convincing more and more car companies

01:55:35.720 | that having AI inside the car is a good idea

01:55:38.680 | or is there some, is there more technical

01:55:43.040 | algorithmic challenges?

01:55:44.920 | What's been keeping you mentally busy?

01:55:47.800 | - I think a lot of the car companies

01:55:49.640 | we are in conversations with are already interested

01:55:52.440 | in definitely driver monitoring.

01:55:54.120 | Like I think it's becoming a must have,

01:55:56.600 | but even interior sensing, I can see like we're engaged

01:55:59.800 | in a lot of like advanced engineering projects

01:56:01.800 | and proof of concepts.

01:56:03.040 | I think technologically though,

01:56:06.600 | and even the technology,

01:56:07.840 | I can see a path to making it happen.

01:56:10.320 | I think it's the use case.

01:56:11.920 | How does the car respond once it knows something about you?

01:56:16.120 | Because you want it to respond in a thoughtful way

01:56:18.480 | that isn't off-putting to the consumer in the car.

01:56:23.480 | So I think that's like the user experience.

01:56:25.600 | I don't think we've really nailed that.

01:56:27.400 | And we usually, that's not part,

01:56:30.440 | we're the sensing platform,

01:56:31.800 | but we usually collaborate with the car manufacturer

01:56:34.240 | to decide what the use case is.

01:56:35.840 | So say you figure out that somebody's angry while driving.

01:56:39.880 | Okay, what should the car do?

01:56:42.320 | You know?

01:56:43.600 | - Do you see yourself as a role of nudging,

01:56:47.120 | of like basically coming up with solutions essentially

01:56:51.200 | that, and then the car manufacturers

01:56:54.200 | kind of put their own little spin on it?

01:56:56.360 | - Right, like we are like the ideation,

01:57:00.320 | creative thought partner,

01:57:02.560 | but at the end of the day,

01:57:03.600 | the car company needs to decide

01:57:05.200 | what's on brand for them, right?

01:57:06.640 | Like maybe when it figures out that you're distracted

01:57:10.040 | or drowsy, it shows you a coffee cup, right?

01:57:12.600 | Or maybe it takes more aggressive behaviors

01:57:14.960 | and basically said, okay,

01:57:16.080 | if you don't like take a rest in the next five minutes,

01:57:18.320 | the car's gonna shut down, right?

01:57:19.640 | Like there's a whole range of actions the car can take

01:57:23.040 | and doing the thing that is most,

01:57:25.880 | yeah, that builds trust with the driver and the passengers.

01:57:29.400 | I think that's what we need to be very careful about.

01:57:32.480 | - Yeah, car companies are funny

01:57:35.400 | 'cause they have their own like,

01:57:37.080 | I mean, that's why people get cars still.

01:57:39.600 | I hope that changes,

01:57:40.520 | but they get it 'cause it's a certain feel and look

01:57:42.760 | and it's a certain, they become proud like Mercedes Benz

01:57:47.760 | or BMW or whatever, and that's their thing.

01:57:51.720 | That's the family brand or something like that,

01:57:54.160 | or Ford or GM, whatever, they stick to that thing.

01:57:57.600 | It's interesting.

01:57:58.680 | It's like, it should be, I don't know.

01:58:01.000 | It should be a little more about the technology inside.

01:58:05.000 | And I suppose there too, there could be a branding,

01:58:09.440 | like a very specific style of luxury or fun,

01:58:13.320 | all that kind of stuff, yeah.

01:58:16.200 | - You know, I have an AI focused fund

01:58:19.680 | to invest in early stage kind of AI driven companies.

01:58:22.680 | And one of the companies we're looking at

01:58:24.800 | is trying to do what Tesla did, but for boats,

01:58:27.480 | for recreational boats.

01:58:28.680 | Yeah, so they're building an electric

01:58:30.720 | and kind of slash autonomous boat.

01:58:34.160 | And it's kind of the same issues,

01:58:35.720 | like what kind of sensors can you put in?

01:58:38.480 | What kind of states can you detect

01:58:40.760 | both exterior and interior within the boat?

01:58:43.640 | Anyways, it's like really interesting.

01:58:45.440 | Do you boat at all?

01:58:46.720 | - No, not well, not in that way.

01:58:49.640 | I do like to get on the lake or a river and fish from a boat,

01:58:54.640 | but that's not boating.

01:58:57.280 | That's a different, still boating.

01:58:59.520 | - Low tech, a low tech. - Low tech boat.

01:59:01.880 | Get away from, get closer to nature boat.

01:59:04.480 | I guess going out to the ocean

01:59:07.160 | is also getting closer to nature in some deep sense.

01:59:12.160 | I mean, I guess that's why people love it.

01:59:14.440 | The enormity of the water just underneath you, yeah.

01:59:19.440 | - I love the water.

01:59:20.720 | - I love both.

01:59:22.880 | I love salt water.

01:59:23.800 | It was like the big and just it's humbling

01:59:25.640 | to be in front of this giant thing that's so powerful

01:59:29.280 | that was here before us and be here after.

01:59:31.640 | But I also love the peace of a small like wooded lake.

01:59:36.640 | It's just, everything's calm.

01:59:38.280 | - Therapeutic.

01:59:41.320 | - You tweeted that I'm excited

01:59:46.520 | about Amazon's acquisition of iRobot.

01:59:49.680 | I think it's a super interesting,

01:59:51.800 | just given the trajectory of which you're part of,

01:59:54.280 | of these honestly small number of companies

01:59:57.880 | that are playing in this space

01:59:59.320 | that are like trying to have an impact on human beings.

02:00:02.160 | So it is an interesting moment in time

02:00:05.000 | that Amazon would acquire iRobot.

02:00:07.000 | You tweet, I imagine a future where home robots

02:00:11.920 | are as ubiquitous as microwaves or toasters.

02:00:16.160 | Here are three reasons why I think this is exciting.

02:00:18.880 | If you remember, I can look it up.

02:00:20.520 | But why is this exciting to you?

02:00:22.320 | - I mean, I think the first reason why this is exciting,

02:00:25.800 | I can't remember the exact order in which I put them.

02:00:30.520 | But one is just, it's gonna be an incredible platform

02:00:34.400 | for understanding our behaviors within the home.

02:00:37.120 | If you think about Roomba, which is the robot vacuum cleaner,

02:00:42.960 | the flagship product of iRobot at the moment,

02:00:45.520 | it's like running around your home,

02:00:48.200 | understanding the layout,

02:00:49.360 | it's understanding what's clean and what's not,

02:00:51.240 | how often do you clean your house?

02:00:52.680 | And all of these behaviors are a piece of the puzzle

02:00:56.360 | in terms of understanding who you are as a consumer.

02:00:58.720 | And I think that could be, again,

02:01:01.320 | used in really meaningful ways,

02:01:04.720 | not just to recommend better products or whatever,

02:01:06.880 | but actually to improve your experience as a human being.

02:01:09.600 | So I think that's very interesting.

02:01:11.720 | I think the natural evolution of these robots in the home,

02:01:17.960 | so it's interesting.

02:01:19.600 | Roomba isn't really a social robot, right, at the moment.

02:01:22.720 | But I once interviewed one of the chief engineers

02:01:27.000 | on the Roomba team,

02:01:28.440 | and he talked about how people named their Roombas.

02:01:31.280 | And if their Roomba broke down,

02:01:33.360 | they would call in and say, "My Roomba broke down,"

02:01:36.560 | and the company would say,

02:01:37.520 | "Well, we'll just send you a new one."

02:01:38.840 | And, "No, no, no, Rosie, you have to like,

02:01:41.200 | "yeah, I want you to fix this particular robot."

02:01:44.400 | So people have already built

02:01:48.000 | interesting emotional connections with these home robots.

02:01:51.680 | And I think that, again, that provides a platform

02:01:54.720 | for really interesting things to just motivate change.

02:01:58.320 | Like it could help you.

02:01:59.400 | I mean, one of the companies

02:02:00.560 | that spun out of MIT, Catalia Health,

02:02:03.360 | the guy who started it spent a lot of time

02:02:06.880 | building robots that help with weight management.

02:02:09.600 | So weight management, sleep, eating better,

02:02:12.480 | yeah, all of these things.

02:02:14.280 | - Well, if I'm being honest,

02:02:16.520 | Amazon does not exactly have a track record

02:02:19.480 | of winning over people in terms of trust.

02:02:22.200 | Now, that said, it's a really difficult problem

02:02:25.880 | for a human being to let a robot in their home

02:02:28.960 | that has a camera on it.

02:02:30.440 | - Right.

02:02:31.280 | - That's really, really, really tough.

02:02:33.240 | And I think Roomba actually,

02:02:36.320 | I have to think about this,

02:02:38.040 | but I'm pretty sure now,

02:02:40.120 | or for some time already has had cameras

02:02:42.360 | because they're doing, the most recent Roomba,

02:02:46.080 | I have so many Roombas.

02:02:47.320 | - Oh, you actually do?

02:02:48.760 | - Well, I programmed it.

02:02:49.600 | I don't use a Roomba for back off.

02:02:51.200 | People that have been to my place,

02:02:52.320 | they're like, yeah, you definitely don't use these Roombas.

02:02:55.280 | - I can't tell like the valence of this comment.

02:03:00.880 | Was it a compliment or like?

02:03:02.360 | - No, it's just a bunch of electronics everywhere.

02:03:06.080 | I have six or seven computers,

02:03:08.920 | I have robots everywhere, Lego robots,

02:03:11.080 | I have small robots and big robots,

02:03:12.720 | and just giant, just piles of robot stuff.

02:03:17.040 | And yeah.

02:03:17.960 | But including the Roombas,

02:03:22.320 | they're being used for their body and intelligence,

02:03:25.480 | but not for their purpose.

02:03:26.840 | (both laughing)

02:03:28.680 | I've changed them, repurposed them for other purposes,

02:03:32.080 | for deeper, more meaningful purposes

02:03:33.880 | than just like the butter robot.

02:03:36.400 | Yeah, which just brings a lot of people happiness, I'm sure.

02:03:41.000 | They have a camera because the thing they advertised,

02:03:44.440 | I had my own cameras too,

02:03:46.400 | but the camera on the new Roomba,

02:03:49.560 | they have like state-of-the-art poop detection

02:03:52.400 | as they advertised, which is a very difficult,

02:03:54.840 | apparently it's a big problem for vacuum cleaners,

02:03:57.680 | is if they go over like dog poop,

02:03:59.640 | it just runs it over and creates a giant mess.

02:04:02.080 | So they have like,

02:04:04.440 | apparently they collected like a huge amount of data

02:04:07.240 | and different shapes and looks and whatever of poop

02:04:09.760 | and now they're able to avoid it and so on.

02:04:12.320 | They're very proud of this.

02:04:14.360 | So there is a camera,

02:04:15.480 | but you don't think of it as having a camera.

02:04:18.200 | Yeah, you don't think of it as having a camera

02:04:21.880 | because you've grown to trust it, I guess,

02:04:23.960 | 'cause our phones, at least most of us,

02:04:26.360 | seem to trust this phone,

02:04:30.440 | even though there's a camera looking directly at you.

02:04:33.040 | I think that if you trust that the company

02:04:39.640 | is taking security very seriously,

02:04:41.440 | I actually don't know how that trust

02:04:42.640 | was earned with smartphones.

02:04:44.440 | I think it just started to provide a lot of positive value

02:04:48.680 | into your life where you just took it in

02:04:50.520 | and then the company over time has shown

02:04:52.360 | that it takes privacy very seriously, that kind of stuff.

02:04:55.120 | But I just, Amazon is not always,

02:04:58.600 | in its social robots, communicated

02:05:01.400 | this is a trustworthy thing,

02:05:03.680 | both in terms of culture and competence.

02:05:06.760 | 'Cause I think privacy is not just about

02:05:08.880 | what do you intend to do,

02:05:10.640 | but also how good are you at doing that kind of thing.

02:05:14.560 | So that's a really hard problem to solve.

02:05:16.720 | - But I mean, but a lot of us have Alexas at home

02:05:19.680 | and I mean, Alexa could be listening in the whole time,

02:05:24.120 | right, and doing all sorts of nefarious things with the data.

02:05:27.320 | You know, hopefully it's not, but I don't think it is.

02:05:32.480 | - But Amazon is not, it's such a tricky thing

02:05:35.360 | for a company to get right, which is to earn the trust.

02:05:38.360 | I don't think Alexas earn people's trust quite yet.

02:05:41.600 | - Yeah, I think it's not there quite yet.

02:05:44.880 | I agree. - And they struggle

02:05:45.800 | with this kind of stuff.

02:05:46.640 | In fact, when these topics are brought up,

02:05:48.400 | people always get nervous.

02:05:50.160 | And I think if you get nervous about it,

02:05:53.640 | I mean, the way to earn people's trust

02:05:57.680 | is not by like, ooh, don't talk about this.

02:05:59.720 | (laughs)

02:06:00.560 | It's just be open, be frank, be transparent,

02:06:03.600 | and also create a culture of where it radiates

02:06:08.000 | at every level from engineer to CEO

02:06:11.160 | that like, you're good people that have a common sense idea

02:06:16.160 | of what it means to respect basic human rights

02:06:20.960 | and the privacy of people and all that kind of stuff.

02:06:23.880 | And I think that propagates throughout the,

02:06:26.500 | that's the best PR, which is like,

02:06:29.960 | over time you understand that these are good folks

02:06:33.800 | doing good things.

02:06:35.280 | Anyway, speaking of social robots,

02:06:37.900 | have you heard about Tesla, Tesla Bot, the humanoid robot?

02:06:42.960 | - Yes, I have.

02:06:43.800 | Yes, yes, yes, but I don't exactly know

02:06:46.240 | what it's designed to do.

02:06:48.320 | Do you? - Oh.

02:06:49.140 | - You probably do.

02:06:50.240 | - No, I know it's designed to do,

02:06:51.960 | but I have a different perspective on it.

02:06:53.760 | But it's designed to, it's a humanoid form,

02:06:57.540 | and it's designed to, for automation tasks

02:07:01.340 | in the same way that industrial robot arms

02:07:04.920 | automate tasks in the factory.

02:07:06.200 | So it's designed to automate tasks in the factory.

02:07:08.100 | But I think that humanoid form,

02:07:10.840 | as we were talking about before,

02:07:12.580 | is one that we connect with as human beings.

02:07:19.560 | Anything legged, honestly,

02:07:21.280 | but the humanoid form especially,

02:07:23.200 | we anthropomorphize it most intensely.

02:07:26.760 | And so the possibility, to me, it's exciting to see

02:07:30.680 | both Atlas developed by Boston Dynamics

02:07:34.520 | and anyone, including Tesla,

02:07:38.480 | trying to make humanoid robots cheaper and more effective.

02:07:42.680 | The obvious way it transforms the world

02:07:46.040 | is social robotics to me,

02:07:48.120 | versus automation of tasks in the factory.

02:07:52.980 | So yeah, I just wanted,

02:07:54.760 | in case that was something you were interested in,

02:07:56.280 | 'cause I find its application

02:07:59.520 | in social robotics super interesting.

02:08:01.560 | - We did a lot of work with Pepper.

02:08:04.480 | Pepper the Robot a while back.

02:08:06.320 | We were like the emotion engine for Pepper,

02:08:08.680 | which is SoftBank's humanoid robot.

02:08:11.360 | - And how tall is Pepper?

02:08:12.480 | It's like--

02:08:13.400 | - Yeah, like, I don't know, like five foot maybe, right?

02:08:18.400 | - Yeah.

02:08:19.280 | - Yeah, pretty big, pretty big.

02:08:21.120 | And it was designed to be at like airport lounges

02:08:25.460 | and retail stores, mostly customer service, right?

02:08:30.300 | Hotel lobbies.

02:08:34.000 | And I mean, I don't know where the state of the robot is,

02:08:36.720 | but I think it's very promising.

02:08:37.880 | I think there are a lot of applications

02:08:39.480 | where this can be helpful.

02:08:40.400 | I'm also really interested in, yeah,

02:08:43.160 | social robotics for the home, right?

02:08:45.120 | Like that can help elderly people, for example,

02:08:48.160 | transport things from one location of the home to the other,

02:08:51.600 | or even like just have your back in case something happens.

02:08:55.600 | Yeah, I don't know.

02:08:57.760 | I do think it's a very interesting space.

02:08:59.800 | It seems early though.

02:09:00.800 | Do you feel like the timing is now?

02:09:04.580 | - I, yes, 100%.

02:09:08.580 | So it always seems early until it's not, right?

02:09:11.900 | - Right, right, right, right.

02:09:12.740 | - I think the time, well,

02:09:14.860 | I definitely think that the time is now,

02:09:21.340 | like this decade for social robots.

02:09:25.740 | Whether the humanoid form is right, I don't think so.

02:09:28.580 | - Mm-hmm, mm-hmm.

02:09:29.740 | - I don't, I think the,

02:09:32.900 | like if we just look at Jibo as an example,

02:09:37.260 | I feel like most of the problem,

02:09:41.720 | the challenge, the opportunity of social connection

02:09:44.780 | between an AI system and a human being

02:09:48.300 | does not require you to also solve the problem

02:09:51.460 | of robot manipulation and bipedal mobility.

02:09:55.220 | So I think you could do that with just a screen, honestly.

02:09:59.100 | But there's something about the interface of Jibo

02:10:00.940 | and you can rotate and so on that's also compelling.

02:10:03.700 | But you get to see all these robot companies

02:10:06.340 | that fail, incredible companies like Jibo and even,

02:10:10.660 | I mean, the iRobot in some sense is a big success story

02:10:16.500 | that it was able to find--

02:10:18.620 | - Right, a niche.

02:10:20.420 | - A niche thing and focus on it.

02:10:21.780 | But in some sense, it's not a success story

02:10:23.780 | because they didn't build any other robot.

02:10:28.020 | Like any other, it didn't expand

02:10:30.300 | to all kinds of robotics.

02:10:31.500 | Like once you're in the home,

02:10:32.740 | maybe that's what happens with Amazon

02:10:34.300 | is they'll flourish into all kinds of other robots.

02:10:36.860 | But do you have a sense, by the way,

02:10:40.460 | why it's so difficult to build a robotics company?

02:10:43.700 | Like why so many companies have failed?

02:10:47.140 | - I think it's like you're building a vertical stack, right?

02:10:50.740 | Like you are building the hardware plus the software

02:10:53.220 | and you find you have to do this at a cost that makes sense.

02:10:56.020 | So I think Jibo was retailing at like, I don't know,

02:11:01.020 | like $800, like $700, $800.

02:11:03.300 | - Yeah, something like this, yeah.

02:11:04.660 | - Which for the use case, right?

02:11:07.140 | There's a dissonance there.

02:11:09.340 | It's too high.

02:11:10.820 | So I think cost is, you know,

02:11:13.420 | cost of building the whole platform in a way that is,

02:11:17.780 | yeah, that is affordable for what value it's bringing.

02:11:20.980 | I think that's the challenge.

02:11:22.460 | I think for these home robots

02:11:25.060 | that are gonna help you do stuff around the home,

02:11:28.500 | that's a challenge too, like the mobility piece of it.

02:11:33.260 | That's hard.

02:11:34.860 | - Well, one of the things I'm really excited with TeslaBot

02:11:37.740 | is the people working on it.

02:11:40.340 | And that's probably the criticism I would apply

02:11:42.900 | to some of the other folks who worked on social robots

02:11:45.900 | is the people working on TeslaBot know how to,

02:11:49.460 | they're focused on and know how to do mass manufacture

02:11:52.260 | and create a product that's super cheap.

02:11:54.060 | - Very cool. - That's the focus.

02:11:55.460 | The engineering focus is not,

02:11:57.220 | I would say that you can also criticize them for that,

02:12:00.180 | is they're not focused on the experience of the robot.

02:12:03.780 | They're focused on how to get this thing

02:12:06.860 | to do the basic stuff that the humanoid form requires

02:12:11.260 | to do it as cheap as possible.

02:12:13.380 | Then the fewest number of actuators,

02:12:15.140 | the fewest numbers of motors,

02:12:16.460 | the increasing efficiency,

02:12:18.300 | they decrease the weight, all that kind of stuff.

02:12:20.060 | - So that's really interesting.

02:12:21.620 | - I would say that Jibo and all those folks,

02:12:24.460 | they focus on the design, the experience, all of that.

02:12:27.860 | And it's secondary how to manufacture.

02:12:30.460 | No, you have to think like the TeslaBot folks

02:12:33.700 | from first principles,

02:12:35.340 | what is the fewest number of components,

02:12:38.020 | the cheapest components?

02:12:38.980 | How can I build it as much in-house as possible

02:12:41.820 | without having to consider all the complexities

02:12:45.860 | of a supply chain, all that kind of stuff.

02:12:47.940 | - Interesting.

02:12:48.780 | - 'Cause if you have to build a robotics company,

02:12:50.700 | you have to, you're not building one robot.

02:12:53.780 | You're building hopefully millions of robots

02:12:55.580 | and you have to figure out how to do that.

02:12:57.540 | Where the final thing, I mean, if it's Jibo type of robot,

02:13:01.740 | is there a reason why Jibo,

02:13:03.700 | like we're gonna have this lengthy discussion,

02:13:05.980 | is there a reason why Jibo has to be over $100?

02:13:08.820 | - It shouldn't be.

02:13:09.860 | - Right, like the basic components.

02:13:12.460 | - Components of it, right.

02:13:13.940 | - Like you could start to actually discuss like,

02:13:16.660 | okay, what is the essential thing about Jibo?

02:13:18.980 | How much, what is the cheapest way I can have a screen?

02:13:21.300 | What's the cheapest way I can have a rotating base?

02:13:24.020 | - Right. - All that kind of stuff.

02:13:24.860 | And then you get down, continuously drive down cost.

02:13:29.540 | Speaking of which, you have launched

02:13:32.940 | an extremely successful companies.

02:13:34.660 | You have helped others, you've invested in companies.

02:13:37.860 | Can you give advice on how to start a successful company?

02:13:42.860 | - I would say have a problem

02:13:46.420 | that you really, really, really wanna solve, right?

02:13:48.620 | Something that you're deeply passionate about.

02:13:51.420 | And honestly, take the first step.

02:13:55.980 | Like that's often the hardest, and don't overthink it.

02:13:59.540 | Like, you know, like this idea of a minimum viable product

02:14:02.660 | or a minimum viable version of an idea, right?

02:14:05.140 | Like, yes, you're thinking about this,

02:14:06.660 | like a humongous, like super elegant, super beautiful thing.

02:14:10.580 | What, like reduce it to the littlest thing

02:14:12.860 | you can bring to market that can solve a problem

02:14:14.860 | that can help address a pain point that somebody has.

02:14:19.860 | They often tell you, like,

02:14:22.620 | start with a customer of one, right?

02:14:24.340 | If you can solve a problem for one person,

02:14:26.660 | then there's probably-- - Yourself

02:14:28.580 | or some other person. - Right.

02:14:29.780 | - Pick a person. - Exactly.

02:14:31.580 | - It could be you. - Yeah.

02:14:32.820 | - That's actually often a good sign

02:14:34.460 | that if you enjoy a thing. - Right.

02:14:36.460 | - Enjoy a thing where you have a specific problem

02:14:38.340 | that you'd like to solve.

02:14:39.500 | That's a good end of one to focus on.

02:14:43.020 | - Right. - What else?

02:14:44.140 | What else is there to actually,

02:14:45.620 | so step one is the hardest, but how do you,

02:14:47.900 | (Roz laughs)

02:14:49.100 | there's other steps as well, right?

02:14:51.540 | - I also think, like, who you bring around the table

02:14:56.140 | early on is so key, right?

02:14:58.340 | Like being clear on what I call, like,

02:15:00.660 | your core values or your North Star.

02:15:02.460 | It might sound fluffy, but actually it's not.

02:15:04.820 | So, and Roz and I, I feel like we did that very early on.

02:15:08.900 | We sat around her kitchen table and we said,

02:15:11.460 | okay, there's so many applications of this technology.

02:15:13.660 | How are we gonna draw the line?

02:15:15.140 | How are we gonna set boundaries?

02:15:16.940 | We came up with a set of core values

02:15:19.220 | that in the hardest of times,

02:15:21.260 | we fell back on to determine how we make decisions.

02:15:25.260 | And so I feel like just getting clarity on these core,

02:15:27.620 | like for us, it was respecting people's privacy,

02:15:30.500 | only engaging with industries where it's clear opt-ins.

02:15:33.660 | So for instance, we don't do any work

02:15:35.580 | in security and surveillance.

02:15:37.100 | So things like that, just getting,

02:15:40.060 | we very big on, you know, one of our core values

02:15:42.500 | is human connection and empathy, right?

02:15:44.620 | And that is, yes, it's an AI company,

02:15:46.540 | but it's about people.

02:15:47.940 | Well, these are all, they become encoded in how we act,

02:15:52.740 | even if you're a small, tiny team of two or three

02:15:55.460 | or whatever.

02:15:56.300 | So I think that's another piece of advice.

02:15:59.980 | - So what about finding people, hiring people?

02:16:02.700 | If you care about people as much as you do,

02:16:04.780 | like it seems like such a difficult thing

02:16:08.180 | to hire the right people.

02:16:10.940 | - I think early on as a startup,

02:16:12.500 | you want people who have,

02:16:14.540 | who share the passion and the conviction

02:16:16.340 | because it's gonna be tough.

02:16:17.900 | Like I've yet to meet a startup

02:16:21.780 | where it was just a straight line to success, right?

02:16:24.580 | Even not just startup,

02:16:25.940 | like even in everyday people's lives, right?

02:16:28.180 | So you always like run into obstacles

02:16:31.580 | and you run into naysayers.

02:16:33.060 | And so you need people who are believers,

02:16:37.700 | whether they're people on your team or even your investors,

02:16:40.660 | you need investors who are really believers

02:16:42.700 | in what you're doing

02:16:44.140 | because that means they will stick with you.

02:16:46.100 | They won't give up at the first obstacle.

02:16:49.380 | I think that's important.

02:16:50.900 | - What about raising money?

02:16:52.660 | What about finding investors?

02:16:55.340 | First of all, raising money,

02:16:58.380 | but also raising money from the right sources

02:17:01.580 | from that ultimately don't hinder you,

02:17:03.540 | but help you, empower you, all that kind of stuff.

02:17:07.060 | What advice would you give there?

02:17:08.580 | You successfully raised money many times in your life.

02:17:12.300 | - Yeah, again, it's not just about the money.

02:17:15.060 | It's about finding the right investors

02:17:17.220 | who are going to be aligned in terms of what you wanna build

02:17:21.300 | and believe in your core values.

02:17:23.060 | Like for example, especially later on,

02:17:25.900 | like I, yeah, in my latest like round of funding,

02:17:29.860 | I try to bring in investors

02:17:31.420 | that really care about like the ethics of AI, right?

02:17:35.460 | And the alignment of vision and mission

02:17:39.620 | and core values is really important.

02:17:41.020 | It's like you're picking a life partner, right?

02:17:43.900 | It's the same kind of-

02:17:45.180 | - So you take it that seriously for investors?

02:17:47.460 | - Yeah, because they're gonna have to stick with you.

02:17:50.140 | - You're stuck together.

02:17:51.540 | - For a while anyway, yeah.

02:17:53.100 | (both laughing)

02:17:54.260 | Maybe not for life, but for a while, for sure.

02:17:56.860 | - For better or worse.

02:17:57.820 | I forget what the vowels usually sound like.

02:17:59.940 | For better or worse?

02:18:00.780 | No.

02:18:01.620 | - Through sick.

02:18:02.780 | - Through sick and then-

02:18:03.620 | - Through something.

02:18:04.620 | - Yeah, yeah, yeah.

02:18:05.460 | (both laughing)

02:18:07.580 | - Oh boy.

02:18:09.220 | Yeah, anyway, it's romantic and deep

02:18:11.820 | and you're in it for a while.

02:18:13.820 | So it's not just about the money.

02:18:16.900 | You tweeted about going to your first

02:18:20.340 | capital camp investing get together.

02:18:22.340 | - Oh yeah.

02:18:23.180 | - And then you learned a lot.

02:18:24.020 | So this is about investing.

02:18:27.780 | So what have you learned from that?

02:18:29.900 | What have you learned about investing in general?

02:18:32.740 | From both, because you've been on both ends of it.

02:18:35.340 | - I mean, I try to use my experience as an operator

02:18:38.260 | now with my investor hat on

02:18:40.780 | when I'm identifying companies to invest in.

02:18:44.060 | First of all, I think the good news

02:18:47.020 | is because I have a technology background, right?

02:18:49.100 | And I really understand machine learning

02:18:51.220 | and computer vision and AI, et cetera.

02:18:53.340 | I can apply that level of understanding, right?

02:18:56.580 | 'Cause everybody says they're an AI company

02:18:58.620 | or they're an AI tech.

02:18:59.780 | And I'm like, no, no, no, no, no.

02:19:01.940 | Show me the technology.

02:19:02.900 | So I can do that level of diligence, which I actually love.

02:19:05.860 | And then I have to do the litmus test of,

02:19:10.700 | if I'm in a conversation with you,

02:19:12.180 | am I excited to tell you about this new company

02:19:14.340 | that I just met, right?

02:19:15.580 | And if I'm an ambassador for that company

02:19:20.380 | and I'm passionate about what they're doing,

02:19:22.140 | I usually use that.

02:19:24.900 | Yeah, that's important to me when I'm investing.

02:19:28.260 | - So that means you actually can explain what they're doing

02:19:32.420 | and you're excited about it.

02:19:34.340 | - Exactly, exactly.

02:19:36.140 | Thank you for putting it so succinctly.

02:19:39.940 | Like rambling, but exactly that's it.

02:19:41.700 | I understand it and I'm excited about it.

02:19:43.900 | - It's funny, but sometimes it's unclear exactly.

02:19:46.740 | I'll hear people tell me,

02:19:50.420 | they'll talk for a while and it sounds cool.

02:19:52.860 | Like they paint a picture of a world,

02:19:54.580 | but then when you try to summarize it,

02:19:56.500 | you're not exactly clear of what,

02:19:59.260 | maybe what the core powerful idea is.

02:20:03.460 | You can't just build another Facebook

02:20:05.300 | or there has to be a core simple to explain idea

02:20:10.300 | that then you can or can't get excited about,

02:20:15.780 | but it's there, it's sitting right there.

02:20:17.820 | How do you ultimately pick who you think will be successful?

02:20:25.540 | It's not just about the thing you're excited about,

02:20:27.980 | like there's other stuff.

02:20:29.540 | - Right, and then there's all the,

02:20:31.900 | with early stage companies, like pre-seed companies,

02:20:34.500 | which is where I'm investing,

02:20:36.180 | sometimes the business model isn't clear yet

02:20:40.420 | or the go-to-market strategy isn't clear.

02:20:42.260 | There's usually, like it's very early on

02:20:43.900 | that some of these things haven't been hashed out,

02:20:46.500 | which is okay.

02:20:47.820 | So the way I like to think about it is like,

02:20:49.500 | if this company's successful,

02:20:51.060 | will this be a multi-billion/trillion dollar market?

02:20:55.580 | Or company.

02:20:56.420 | And so that's definitely a lens that I use.

02:21:00.460 | - What's pre-seed?

02:21:02.020 | What are the different stages?

02:21:03.900 | And what's the most exciting stage?

02:21:05.660 | What's interesting about every stage, I guess?

02:21:09.580 | - Yeah, so pre-seed is usually when you're just starting out,

02:21:13.980 | you've maybe raised the friends and family round,

02:21:16.500 | so you've raised some money from people you know,

02:21:18.500 | and you're getting ready

02:21:19.380 | to take your first institutional check-in,

02:21:22.540 | like first check from an investor.

02:21:25.100 | And I love the stage.

02:21:28.820 | There's a lot of uncertainty.

02:21:30.700 | Some investors really don't like the stage

02:21:32.540 | because the financial models aren't there.

02:21:36.460 | Often the teams aren't even like formed,

02:21:38.740 | it's really, really early.

02:21:40.220 | But to me, it's like a magical stage

02:21:45.860 | because it's the time when there's so much conviction,

02:21:48.420 | so much belief, almost delusional, right?

02:21:51.660 | And there's a little bit of naivete around

02:21:53.980 | with founders at the stage, and I just love it.

02:21:57.900 | It's contagious.

02:21:59.020 | And I love that I can,

02:22:02.660 | often they're first-time founders, not always,

02:22:06.020 | but often they're first-time founders,

02:22:07.660 | and I can share my experience as a founder myself,

02:22:10.140 | and I can empathize, right?

02:22:12.580 | And I can almost, I create a safe ground where,

02:22:17.180 | 'cause you have to be careful

02:22:18.620 | what you tell your investors, right?

02:22:21.180 | And I will often say,

02:22:23.140 | I've been in your shoes as a founder,

02:22:24.780 | you can tell me if it's challenging,

02:22:26.780 | you can tell me what you're struggling with.

02:22:28.340 | It's okay to vent.

02:22:30.140 | So I create that safe ground,

02:22:31.940 | and I think that's a superpower.

02:22:35.300 | - Yeah, you have to, I guess,

02:22:37.980 | you have to figure out if this kind of person

02:22:39.860 | is gonna be able to ride the roller coaster

02:22:42.020 | like of many pivots and challenges

02:22:46.980 | and all that kind of stuff.

02:22:48.100 | And if the space of ideas they're working in

02:22:50.860 | is interesting, like the way they think about the world.

02:22:54.180 | Yeah, 'cause if it's successful,

02:22:57.660 | the thing they end up with might be very different,

02:22:59.700 | the reason it's successful for.

02:23:01.500 | - Actually, I was gonna say the third,

02:23:05.140 | so the technology is one aspect,

02:23:07.340 | the market or the idea, right, is the second,

02:23:09.940 | and the third is the founder, right?

02:23:11.260 | Is this somebody who I believe has conviction,

02:23:13.940 | is a hustler, is gonna overcome obstacles.

02:23:20.660 | Yeah, I think that it's gonna be a great leader, right?

02:23:23.140 | Like as a startup, as a founder,

02:23:25.740 | you're often, you are the first person,

02:23:27.660 | and your role is to bring amazing people around you

02:23:31.060 | to build this thing.

02:23:32.780 | And so you're an evangelist, right?

02:23:36.140 | So how good are you gonna be at that?

02:23:38.020 | So I try to evaluate that too.

02:23:40.260 | - You also, in the tweet thread about it,

02:23:43.740 | mentioned, is this a known concept?

02:23:45.780 | Random rich dudes, RRDS.

02:23:48.940 | Saying that there should be like random rich women, I guess.

02:23:53.180 | What's the dudes version of women,

02:23:56.500 | the women version of dudes?

02:23:57.820 | Ladies, I don't know.

02:23:59.380 | What's, is this a technical term?

02:24:01.500 | Is this known?

02:24:02.740 | Random rich dudes?

02:24:03.580 | - Well, I didn't make that up,

02:24:04.540 | but I was at this capital camp,

02:24:07.100 | which is a get together for investors of all types,

02:24:11.260 | and there must have been maybe 400 or so attendees,

02:24:17.380 | maybe 20 were women.

02:24:19.060 | It was just very disproportionately,

02:24:21.480 | a male dominated, which I'm used to.

02:24:25.700 | - I think you're used to this kind of thing.

02:24:27.020 | - I'm used to it, but it's still surprising.

02:24:29.700 | And as I'm raising money for this fund,

02:24:31.820 | so my fund partner is a guy called Rob May,

02:24:36.140 | who's done this before.

02:24:37.580 | So I'm new to the investing world,

02:24:39.500 | but he's done this before.

02:24:41.620 | Most of our investors in the fund are these,

02:24:45.180 | I mean, awesome, I'm super grateful to them,

02:24:47.420 | random, just rich guys.

02:24:48.860 | I'm like, where are the rich women?

02:24:50.340 | So I'm really adamant in both investing

02:24:52.820 | in women-led AI companies,

02:24:56.140 | but I also would love to have women investors

02:24:58.900 | be part of my fund,

02:25:00.740 | because I think that's how we drive change.

02:25:03.780 | - Yeah, so then that takes time, of course,

02:25:06.640 | but there's been quite a lot of progress,

02:25:08.560 | but yeah, for the next Mark Zuckerberg to be a woman

02:25:11.900 | and all that kind of stuff,

02:25:13.500 | that's just a huge number of wealth generated by women,

02:25:17.980 | and then controlled by women,

02:25:19.180 | then allocated by women, and all that kind of stuff.

02:25:22.060 | And then beyond just women,

02:25:23.540 | just broadly across all different measures

02:25:25.900 | of diversity and so on.

02:25:27.320 | Let me ask you to put on your wise sage hat.

02:25:35.500 | So you already gave advice on startups,

02:25:38.340 | and just advice for women,

02:25:43.340 | but in general, advice for folks in high school

02:25:46.980 | or college today, how to have a career they can be proud of,

02:25:50.700 | how to have a life they can be proud of.

02:25:53.860 | I suppose you have to give this kind of advice to your kids.

02:25:58.060 | - Kids, yeah.

02:25:59.140 | Well, here's the number one advice that I give to my kids.

02:26:03.340 | My daughter's now 19, by the way,

02:26:04.980 | and my son's 13 and a half,

02:26:06.380 | so they're not little kids anymore.

02:26:08.940 | - Does it break your heart?

02:26:11.500 | - It does.

02:26:12.700 | Like a girl, but they're awesome.

02:26:14.260 | They're my best friends.

02:26:15.460 | Yeah, I think the number one advice I would share

02:26:19.020 | is embark on a journey without attaching to outcomes,

02:26:22.380 | and enjoy the journey, right?

02:26:25.100 | So we often were so obsessed with the end goal,

02:26:29.560 | A, that doesn't allow us to be open

02:26:33.860 | to different endings of a journey,

02:26:36.100 | or a story.

02:26:37.020 | So you become so fixated on a particular path.

02:26:41.780 | You don't see the beauty in the other alternative path.

02:26:44.660 | And then you forget to enjoy the journey

02:26:48.620 | because you're just so fixated on the goal.

02:26:50.580 | And I've been guilty of that for many, many years in my life.

02:26:54.100 | And I'm now trying to make the shift of,

02:26:57.860 | no, no, no, I'm gonna, again,

02:27:00.300 | trust that things are gonna work out,

02:27:02.460 | and it'll be amazing, and maybe even exceed your dreams.

02:27:05.540 | We have to be open to that.

02:27:07.300 | - Yeah, taking a leap into all kinds of things.

02:27:09.860 | I think you tweeted like you went on vacation by yourself

02:27:12.860 | or something like this.

02:27:13.900 | - I know.

02:27:14.740 | - And just going, just taking the leap, doing it.

02:27:19.380 | - Totally, doing it.

02:27:20.220 | - And enjoying it, enjoying the moment,

02:27:22.300 | enjoying the weeks, enjoying not looking at

02:27:25.020 | the some kind of career ladder next step and so on.

02:27:29.540 | Yeah, there's something to that, like over planning too.

02:27:34.380 | I'm surrounded by a lot of people that kinda,

02:27:36.380 | so I don't plan.

02:27:37.740 | - You don't?

02:27:38.580 | - No.

02:27:39.420 | - Do you not do goal setting?

02:27:40.980 | - My goal setting is very like,

02:27:46.620 | I like the affirmations, it's very,

02:27:49.420 | it's almost, I don't know how to put it into words,

02:27:53.520 | but it's a little bit like

02:27:55.700 | what my heart yearns for, kind of.

02:28:01.580 | And I guess in the space of emotions,

02:28:03.620 | more than in the space of like,

02:28:06.020 | this will be, like in the rational space.

02:28:09.500 | 'Cause I just try to picture a world

02:28:13.620 | that I would like to be in,

02:28:15.780 | and that world is not clearly pictured,

02:28:17.580 | it's mostly in the emotional world.

02:28:19.380 | I mean, I think about that from robots,

02:28:22.060 | 'cause I have this desire, I've had it my whole life to,

02:28:27.060 | well, it took different shapes,

02:28:29.660 | but I think once I discovered AI,

02:28:32.380 | the desire was to,

02:28:34.020 | I think in the context of this conversation

02:28:38.220 | could be easily, easier described

02:28:40.520 | as basically a social robotics company.

02:28:42.940 | And that's something I dreamed of doing.

02:28:45.600 | And, well, there's a lot of complexity to that story,

02:28:52.060 | but that's the only thing, honestly, I dream of doing.

02:28:55.620 | So I imagine a world that I could help create,

02:29:01.580 | but it's not, there's no steps along the way.

02:29:05.660 | And I think I'm just kind of stumbling around

02:29:09.460 | and following happiness and working my ass off

02:29:13.260 | in almost random, like an ant does in random directions.

02:29:17.340 | But a lot of people, a lot of successful people around me

02:29:19.420 | say this, you should have a plan, you should have a clear goal.

02:29:21.540 | You have a goal at the end of the month,

02:29:22.700 | you have a goal at the end of the year.

02:29:24.300 | I don't, I don't, I don't.

02:29:26.580 | And there's a balance to be struck, of course,

02:29:31.020 | but there's something to be said about really making sure

02:29:36.020 | that you're living life to the fullest,

02:29:40.220 | that goals can actually get in the way of.

02:29:42.620 | - So one of the best, like kind of most,

02:29:47.300 | what do you call it when it's like challenges your brain,

02:29:52.500 | what do you call it?

02:29:53.500 | - The only thing that comes to mind,

02:29:58.100 | and this is me saying is the mind fuck, but yes.

02:30:00.180 | - Okay, okay, maybe, okay, something like that.

02:30:03.020 | - Yes.

02:30:03.940 | - Super inspiring talk, Kenneth Stanley,

02:30:06.620 | he was at OpenAI, he just left,

02:30:09.340 | and he has a book called "Why Greatness Can't Be Planned."

02:30:12.460 | And it's actually an AI book.

02:30:14.100 | So, and he's done all these experiments

02:30:16.100 | that basically show that when you over optimize,

02:30:19.220 | like the trade-off is you're less creative, right?

02:30:23.700 | And to create true greatness

02:30:26.700 | and truly creative solutions to problems,

02:30:29.620 | you can't over plan it, you can't.

02:30:31.740 | And I thought that was,

02:30:33.100 | and so he generalizes it beyond AI,

02:30:35.540 | and he talks about how we apply that in our personal life

02:30:38.220 | and our organizations and our companies,

02:30:40.380 | which are over KPI, right?

02:30:42.340 | Like look at any company in the world,

02:30:43.900 | and it's all like, these are the goals,

02:30:45.300 | these are the weekly goals, and the sprints,

02:30:48.940 | and then the quarterly goals, blah, blah, blah.

02:30:50.780 | And he just shows with a lot of his AI experiments

02:30:55.260 | that that's not how you create truly game-changing ideas.

02:30:59.540 | So there you go.

02:31:00.540 | - Yeah, yeah.

02:31:01.380 | - You should interview Kenneth, he's awesome.

02:31:02.940 | - Yeah, there's a balance, of course.

02:31:04.620 | 'Cause that's, yeah, many moments of genius

02:31:07.780 | will not come from planning and goals,

02:31:09.780 | but you still have to build factories,

02:31:12.900 | and you still have to manufacture,

02:31:14.260 | and you still have to deliver,

02:31:15.340 | and there's still deadlines and all that kind of stuff.

02:31:17.260 | And for that, it's good to have goals.

02:31:19.340 | - I do goal setting with my kids, we all have our goals.

02:31:22.660 | But I think we're starting to morph

02:31:25.700 | into more of these bigger picture goals,

02:31:27.820 | and not obsess about, I don't know, it's hard.

02:31:31.340 | - Well, I honestly think, especially with kids,

02:31:33.460 | it's much better to have a plan and have goals and so on,

02:31:36.260 | 'cause you have to learn the muscle

02:31:38.100 | of what it feels like to get stuff done.

02:31:40.900 | But I think once you learn that, there's flexibility for me.

02:31:43.940 | 'Cause I spent most of my life with goal setting and so on.

02:31:47.980 | So I've gotten good with grades and school.

02:31:50.580 | I mean, school, if you wanna be successful at school,

02:31:53.900 | I mean, the kind of stuff in high school and college

02:31:56.180 | that kids have to do, in terms of managing their time

02:31:59.140 | and getting so much stuff done.

02:32:01.100 | It's like, taking five, six, seven classes in college,

02:32:05.620 | they're like, that would break the spirit of most humans

02:32:09.380 | if they took one of them later in life.

02:32:12.420 | It's like really difficult stuff,

02:32:14.820 | especially in engineering curricula.

02:32:16.500 | So I think you have to learn that skill,

02:32:19.980 | but once you learn it, you can maybe,

02:32:21.900 | 'cause you can be a little bit on autopilot

02:32:24.540 | and use that momentum,

02:32:25.700 | and then allow yourself to be lost in the flow of life.

02:32:29.100 | Just kinda, or also give,

02:32:34.100 | I worked pretty hard to allow myself

02:32:38.540 | to have the freedom to do that.

02:32:39.860 | That's really, that's a tricky freedom to have.

02:32:42.900 | Because a lot of people get lost in the rat race,

02:32:45.300 | and they also, like financially,

02:32:49.660 | they, whenever you get a raise,

02:32:52.820 | they'll get like a bigger house.

02:32:54.100 | - Right, right, right.

02:32:54.940 | - Or something like this.

02:32:55.780 | I put very, so like, you're always trapped in this race.

02:32:58.620 | I put a lot of emphasis on living below my means always.

02:33:03.620 | And so there's a lot of freedom to do whatever,

02:33:08.260 | whatever the heart desires.

02:33:11.380 | That's a really, but everyone has to decide

02:33:13.220 | what's the right thing, what's the right thing for them.

02:33:15.540 | For some people, having a lot of responsibilities,

02:33:18.760 | like a house they can barely afford,

02:33:20.860 | or having a lot of kids, the responsibility side of that,

02:33:24.500 | is really, helps them get their shit together.

02:33:28.020 | Like, all right, I need to be really focused and good.

02:33:30.980 | Some of the most successful people I know have kids,

02:33:33.100 | and the kids bring out the best in them.

02:33:34.700 | They make them more productive and less productive.

02:33:36.540 | - Accountability, it's an accountability thing, absolutely.

02:33:39.260 | - And almost something to actually live and fight

02:33:42.060 | and work for, like having a family.

02:33:44.900 | It's fascinating to see.

02:33:46.560 | 'Cause you would think kids would be a hit on productivity,

02:33:49.400 | but they're not, for a lot of really successful people.

02:33:52.200 | They really, they're like an engine of--

02:33:54.920 | - Right, efficiency, oh my God.

02:33:56.120 | - Yeah, it's weird.

02:33:57.840 | I mean, it's beautiful, it's beautiful to see.

02:33:59.600 | And also social happiness.

02:34:01.960 | Speaking of which, what role do you think love plays

02:34:06.960 | in the human condition, love?

02:34:09.100 | - I think love is,

02:34:15.260 | yeah, I think it's why we're all here.

02:34:19.920 | I think it would be very hard to live life

02:34:21.720 | without love in any of its forms, right?

02:34:26.160 | - Yeah, that's the most beautiful of forms

02:34:31.040 | that human connection takes, right?

02:34:35.020 | - Yeah, I feel like everybody wants to feel loved, right?

02:34:40.020 | In one way or another, right?

02:34:42.080 | - And to love.

02:34:42.920 | - Yeah, and to love too.

02:34:44.320 | - Totally, yeah, I agree with that.

02:34:46.280 | - Both of it.

02:34:47.120 | I'm not even sure what feels better.

02:34:48.860 | Both, both are like that.

02:34:52.520 | - To give love too, yeah.

02:34:54.400 | - And it is like we've been talking about,

02:34:56.920 | an interesting question, whether some of that,

02:34:59.480 | whether one day we'll be able to love a toaster.

02:35:02.340 | Get some small--

02:35:05.280 | - I wasn't quite thinking about that when I said--

02:35:07.520 | - The toaster.

02:35:08.360 | - Yeah, like we all need love and give love.

02:35:10.960 | Okay, you're right.

02:35:11.800 | - I was thinking about Brad Pitt and toasters.

02:35:12.640 | - Brad Pitt and toasters, great.

02:35:14.400 | - All right, well, I think we started on love

02:35:18.440 | and ended on love.

02:35:20.080 | This was an incredible conversation, Ron,

02:35:22.160 | thank you so much.

02:35:23.360 | You're an incredible person.

02:35:24.760 | Thank you for everything you're doing in AI,

02:35:28.080 | in the space of just caring about humanity,

02:35:32.640 | human emotion, about love,

02:35:34.680 | and being an inspiration to a huge number of people

02:35:38.200 | in robotics, in AI, in science, in the world in general.

02:35:42.240 | So thank you for talking to me, it's an honor.

02:35:44.200 | - Thank you for having me,

02:35:45.400 | and you know I'm a big fan of yours as well,

02:35:47.240 | so it's been a pleasure.

02:35:48.680 | - Thanks for listening to this conversation

02:35:51.120 | with Rana Elkayoobie.

02:35:52.800 | To support this podcast,

02:35:53.960 | please check out our sponsors in the description.

02:35:56.840 | And now, let me leave you with some words

02:35:58.720 | from Helen Keller.

02:35:59.800 | The best and most beautiful things in the world

02:36:03.560 | cannot be seen or even touched.

02:36:06.120 | They must be felt with the heart.

02:36:09.320 | Thank you for listening, and hope to see you next time.

02:36:12.400 | (upbeat music)

02:36:14.980 | (upbeat music)

02:36:17.560 | [BLANK_AUDIO]

Rana el Kaliouby: Emotion AI, Social Robots, and Self-Driving Cars | Lex Fridman Podcast #322

Chapters