Dan Kokotov: Speech Recognition with AI and Humans

Lex Fridman (02:04.880)

It probably goes without saying, but I should say it anyway, that you

Dan Kokotov (02:08.640)

can not buy your way onto this podcast.

Lex Fridman (02:11.600)

I don't know why you would want to.

Dan Kokotov (02:13.720)

I wanted to bring this up to make a specific point that no sponsor

Lex Fridman (02:17.480)

will ever influence what I do on this podcast, or to the best of

Dan Kokotov (02:21.240)

my ability, influence what I think.

Lex Fridman (02:23.480)

I wasn't really thinking about this.

Dan Kokotov (02:25.480)

For example, when I interviewed Jack Dorsey, who is the CEO of Square that

Lex Fridman (02:30.520)

happens to be sponsoring this podcast, but I should really make it explicit.

Dan Kokotov (02:34.960)

I will never take money for bringing a guest on.

Dan Kokotov (02:37.920)

Every guest on this podcast is someone I genuinely am curious to talk to or just

Dan Kokotov (02:43.240)

genuinely love something they've created.

Lex Fridman (02:45.440)

As I sometimes get criticized for, I'm just a fan of people.

Lex Fridman (02:49.760)

And that's who I talk to.

Lex Fridman (02:51.440)

As I also talk about way too much, money is really never a consideration.

Dan Kokotov (02:56.120)

In general, no amount of money can buy my integrity.

Lex Fridman (03:00.240)

That's true for this podcast, and that's true for anything else I do.

Dan Kokotov (03:05.160)

If you enjoy this thing, subscribe on YouTube, review on the Apple podcast,

Lex Fridman (03:09.760)

follow on Spotify, support on Patreon, a podcast on YouTube, and

Dan Kokotov (03:14.120)

support on Patreon, or connect with me on Twitter at Lex Friedman.

Lex Fridman (03:18.280)

And now here's my conversation with Dan Kokotov.

Dan Kokotov (03:23.080)

You mentioned science fiction on the phone.

Lex Fridman (03:25.600)

So let's go with the ridiculous first.

Lex Fridman (03:28.040)

What's the greatest sci fi novel of all time in your view?

Lex Fridman (03:32.360)

And maybe what ideas do you find philosophically fascinating about it?

Dan Kokotov (03:37.800)

The greatest sci fi novel of all time is Dune.

Lex Fridman (03:40.960)

And the second greatest is The Children of Dune.

Lex Fridman (03:43.880)

And the third greatest is The God Emperor of Dune.

Lex Fridman (03:47.080)

So I'm a huge fan of the whole series.

Dan Kokotov (03:50.120)

I mean, it's just an incredible world that he created.

Lex Fridman (03:53.720)

And I don't know if you've read the book or not.

Dan Kokotov (03:55.480)

No, I have not.

Lex Fridman (03:56.280)

It's one of my biggest regrets, especially because a new movie is coming out.

Dan Kokotov (04:01.480)

Everyone's super excited about it.

Lex Fridman (04:02.920)

I used to, it's ridiculous to say, and sorry to interrupt, is that I

Dan Kokotov (04:07.600)

used to play the video game.

Lex Fridman (04:09.600)

It used to be Dune.

Dan Kokotov (04:11.280)

I guess you would call that real time strategy.

Lex Fridman (04:13.720)

Right.

Dan Kokotov (04:14.080)

I think I remember that game.

Lex Fridman (04:15.240)

Yeah, it was kind of awesome.

Dan Kokotov (04:16.400)

Nineties or something.

Lex Fridman (04:17.400)

I think I played it actually when I was in Russia.

Dan Kokotov (04:19.800)

I definitely remember it.

Lex Fridman (04:21.240)

I was not in Russia anymore.

Dan Kokotov (04:22.880)

I think at the time that I used to live in Russia, I think video games

Lex Fridman (04:26.800)

were about like the suspicion of Pong.

Dan Kokotov (04:29.080)

I think Pong was pretty much like the greatest game I ever got to play in Russia,

Lex Fridman (04:33.280)

which was still a privilege right in that age.

Lex Fridman (04:35.240)

So you didn't get color?

Lex Fridman (04:36.520)

You didn't get like, uh, so I left Russia.

Lex Fridman (04:39.120)

I left Russia in 1991, right?

Lex Fridman (04:40.840)

Okay.

Lex Fridman (04:41.680)

So I was one of the few lucky kids because my mom was a programmer.

Lex Fridman (04:45.200)

So I would go to her work, right?

Dan Kokotov (04:47.120)

I would take the, the Metro.

Lex Fridman (04:49.200)

I've got our work and play like on, I guess the equivalent of like a

Dan Kokotov (04:52.160)

286 PC, you know, nice floppy disks.

Lex Fridman (04:56.160)

Yes.

Dan Kokotov (04:56.880)

So, okay.

Lex Fridman (04:57.280)

Put back to doing what you get back to doing.

Lex Fridman (04:59.960)

And by the way, the new movie I'm pretty interested in, but the

Lex Fridman (05:04.200)

skeptical, I'm a little skeptical.

Dan Kokotov (05:06.600)

I'm a little skeptical.

Lex Fridman (05:07.200)

I saw the trailer.

Dan Kokotov (05:08.120)

Uh, I don't know.

Lex Fridman (05:09.320)

So there's, there's a David Lynch movie doing as you may know, I'm

Dan Kokotov (05:12.680)

a huge David Lynch fan, by the way.

Lex Fridman (05:14.200)

So the movie is somewhat controversial, but it's a little confusing, but it

Dan Kokotov (05:20.200)

captures kind of the mood of the book better than I would say like most any

Lex Fridman (05:24.600)

adaptation and like doing so much about kind of mood and the world, right.

Lex Fridman (05:28.640)

But back to the philosophical point.

Lex Fridman (05:29.920)

So in the fourth book, God, emperor of doing, there's a sort of setting where

Dan Kokotov (05:36.400)

Leto, one of the characters, he's become this weird sort of God emperor.

Lex Fridman (05:41.280)

He's turned into a gigantic worm.

Dan Kokotov (05:42.720)

I mean, you kind of have to read the book to understand what that means.

Lex Fridman (05:44.920)

So the worms are involved, the worms are involved.

Dan Kokotov (05:47.120)

You probably saw the worms in the trailer, right.

Lex Fridman (05:49.200)

And in the video, you kind of like merges with the swarm, um, and becomes

Dan Kokotov (05:53.200)

this tyrant of the world and like oppresses the people for a long time.

Lex Fridman (05:56.640)

Right.

Lex Fridman (05:56.840)

But he has a purpose and the purpose is to kind of, uh, break through kind of

Lex Fridman (06:01.640)

a stagnation period in civilization.

Dan Kokotov (06:03.840)

Right.

Lex Fridman (06:04.120)

Um, but people have gotten too comfortable, right.

Lex Fridman (06:06.040)

And so you kind of oppresses them so that they explode and like go on to

Lex Fridman (06:11.720)

colonize new worlds and kind of renew the forward momentum of humanity.

Dan Kokotov (06:15.720)

Right.

Lex Fridman (06:16.520)

And so to me, that's kind of fascinating, right.

Dan Kokotov (06:18.520)

You need a little bit of pressure and suffering, right.

Lex Fridman (06:21.640)

To kind of like make progress, not, not, not get too comfortable.

Dan Kokotov (06:28.120)

Maybe that's a bit of a cruel philosophy to take away, but that seems to be

Lex Fridman (06:33.120)

the case, unfortunately, obviously, I'm a huge fan of, uh, suffering.

Lex Fridman (06:39.040)

So one of the reasons we're talking today is that a bunch of people requested

Lex Fridman (06:46.360)

that I do transcripts for this podcast and do captioning.

Dan Kokotov (06:51.280)

I used to make all kinds of YouTube videos and I would go on up work, I

Lex Fridman (06:56.240)

think, and I would hire folks to do transcription and it was always a pain

Dan Kokotov (07:01.440)

in the ass, if I'm being honest, and then I don't know how I discovered Rev.

Lex Fridman (07:07.400)

But when I did, it was this feeling of like, Holy shit, somebody figured

Dan Kokotov (07:13.960)

out how to do it just really easily.

Dan Kokotov (07:16.760)

I I'm, I'm such a fan of just when people take a problem and they just make it easy.

Dan Kokotov (07:25.840)

Right.

Lex Fridman (07:26.000)

You know, like just, uh, there's so many, uh, there's so many,

Dan Kokotov (07:31.280)

it's like, there's so many things in life that you might not even

Lex Fridman (07:35.240)

be aware of that are painful.

Dan Kokotov (07:37.720)

Then Rev, you just like give the audio, give the video, you can

Lex Fridman (07:43.320)

actually give a YouTube link.

Lex Fridman (07:45.920)

And then it comes back like a day later or, uh, two days later, whatever

Lex Fridman (07:52.320)

the hell it is with the captions, you know, all in a standardized format.

Dan Kokotov (07:57.200)

It was, I dunno, it was, it was, it was, it was truly a joy.

Lex Fridman (08:00.280)

So I thought I had, you know, just for the hell of it, uh, talk to you

Dan Kokotov (08:04.920)

that one other product just made my soul feel good.

Lex Fridman (08:08.200)

One other product that I've used like that is, uh, for people who might

Dan Kokotov (08:12.280)

be familiar is called isotope RX.

Lex Fridman (08:15.840)

It's for audio editing and like, and that's another one where it was

Dan Kokotov (08:21.000)

like, you just drop it.

Lex Fridman (08:24.080)

I dropped into the audio and it just cleans everything up really nicely.

Dan Kokotov (08:28.240)

All the stupid, like the mouth sounds and sometimes there's a background

Lex Fridman (08:35.400)

like sounds due to the malfunction of the equipment.

Dan Kokotov (08:39.000)

It can clean that stuff up.

Lex Fridman (08:40.400)

It can, it has a general voice denoising.

Dan Kokotov (08:43.040)

It has like automation capabilities where you can do batch processing

Lex Fridman (08:47.400)

and you can put a bunch of effects.

Dan Kokotov (08:49.600)

I mean, it just, I dunno, everything else sucked for like voice based

Lex Fridman (08:55.480)

cleanup that I've ever used.

Dan Kokotov (08:57.800)

They've used audition, Adobe audition, and he's all kinds of other things

Lex Fridman (09:01.120)

with plugins and you have to kind of figure it all out.

Dan Kokotov (09:04.480)

You have to do it manually here.

Lex Fridman (09:05.840)

It's just, it just worked.

Lex Fridman (09:07.520)

So that's another one in this whole pipeline.

Lex Fridman (09:09.640)

It just brought joy to my, to my heart.

Dan Kokotov (09:12.640)

Anyway, all that to say is, uh, uh, Rev put a smile to my face.

Lex Fridman (09:18.760)

So can you maybe take a step back and say, what is Rev and how does it work?

Lex Fridman (09:24.560)

And Rev or Rev.com?

Lex Fridman (09:26.040)

Rev, Rev.com, the same thing, I guess, uh, that way we do have Rev.ai now as

Dan Kokotov (09:31.240)

well, which we can talk about later.

Lex Fridman (09:33.760)

Like, do you have the actual domain or is it just, uh, the actual domain,

Lex Fridman (09:37.160)

but we also use it kind of as a, as a sub brand.

Lex Fridman (09:41.120)

Oh, so we've, so we use Rev.ai to denote our ASR services, right?

Lex Fridman (09:46.040)

And Rev.com is kind of our more human and to the end user services.

Lex Fridman (09:49.800)

So it's like wordpress.com and wordpress.org, they actually have separate

Dan Kokotov (09:53.400)

brands that like, I don't know if you're familiar with what those are.

Lex Fridman (09:56.680)

Yeah, they provide almost like a separate branch of a little bit.

Lex Fridman (10:00.360)

I think with that, it's like wordpress.org is kind of their open source, right?

Lex Fridman (10:04.080)

And, uh, wordpress.com is sort of their hosted commercial offering.

Dan Kokotov (10:07.600)

Yes.

Lex Fridman (10:08.160)

Um, and with us, the differential is a little bit different,

Lex Fridman (10:10.200)

but maybe a similar idea.

Lex Fridman (10:11.720)

Yep.

Dan Kokotov (10:12.800)

Okay.

Lex Fridman (10:12.960)

So what is Rev?

Lex Fridman (10:13.760)

Before I launch into, uh, what is Rev?

Lex Fridman (10:16.360)

I was going to say, you know, like you, you were talking about like

Dan Kokotov (10:18.080)

Rev was music to your ears, your, your, your field was music to my ears.

Lex Fridman (10:21.840)

To us, the founders of Rev, because, um, Rev was kind of founded to improve

Dan Kokotov (10:27.720)

on the model of Upwork that was kind of the original, um, or part of their

Lex Fridman (10:32.600)

original impetus, like our CEO, Jason, was a early employee of Upwork.

Lex Fridman (10:38.200)

So he was very familiar with their work, the company Upwork company.

Lex Fridman (10:41.400)

Um, and so he was very familiar with that model and he wanted to make the whole

Dan Kokotov (10:46.360)

experience better because he knew like, when you go at that time, Upwork was

Lex Fridman (10:49.600)

primarily programmers, so the main thing they offered us, if you want to hire,

Dan Kokotov (10:54.040)

you know, someone to help you code a little site, right.

Dan Kokotov (10:56.880)

Um, you could go on Upwork, um, you could like browse through a list of freelancers,

Dan Kokotov (11:01.480)

pick a programmer, you know, have a contract with them and have them do some

Lex Fridman (11:04.720)

work, but it was kind of a difficult experience because, uh, for the, for you,

Dan Kokotov (11:11.400)

you would kind of have to browse through all these people, right.

Lex Fridman (11:13.600)

And you have to decide, okay, like, well, is this guy good as, um, or somebody

Dan Kokotov (11:18.040)

else better and naturally, you know, you're going to Upwork because you're not

Lex Fridman (11:21.520)

an expert, right?

Dan Kokotov (11:22.600)

If you're an expert, you probably wouldn't be like getting a programmer

Lex Fridman (11:24.640)

from Upwork, uh, so, so how can you really tell?

Lex Fridman (11:27.840)

So there's a kind of like a lot of potential regret, right?

Lex Fridman (11:31.160)

What if I choose a bad person, they're like, going to be late on the work.

Dan Kokotov (11:34.560)

It's going to be a painful experience.

Lex Fridman (11:36.240)

And for the freelancer, it was also painful because, you know, half the time

Dan Kokotov (11:39.920)

they spent not on actually doing the work, but kind of figuring out how can I make

Lex Fridman (11:43.640)

my profile most attractive to the buyer, right?

Lex Fridman (11:47.560)

And they're not an expert on that either.

Lex Fridman (11:49.520)

So like Grav's idea was let's remove the barrier, right?

Dan Kokotov (11:52.560)

Like, let's make it simple where we'll pick a few, uh, verticals

Lex Fridman (11:56.600)

that are fairly standardizable.

Dan Kokotov (11:58.520)

Now we actually started with translation, um, and then we added

Lex Fridman (12:01.760)

audio transcription a bit later and we'll just make it a website.

Dan Kokotov (12:05.200)

You go give us your files.

Lex Fridman (12:06.920)

We'll give you back, uh, the results, you know, as soon as possible.

Dan Kokotov (12:11.880)

You know, originally maybe it was 48 hours.

Lex Fridman (12:13.800)

Then we made it shorter and shorter and shorter.

Dan Kokotov (12:15.680)

Um, yeah, there's a rush processing too.

Dan Kokotov (12:17.880)

There's a rush processing now, uh, and, uh, we'll hide all the details from you.

Dan Kokotov (12:22.920)

Right.

Lex Fridman (12:23.520)

Yeah.

Lex Fridman (12:24.240)

And like, that's kind of exactly what you're experiencing, right?

Dan Kokotov (12:27.000)

You don't, you don't need to worry about the details of how the sausage is made.

Dan Kokotov (12:29.840)

That's really cool.

Lex Fridman (12:30.480)

The, so you picked like a vertical by vertically, you mean basically a

Dan Kokotov (12:35.160)

service, a service category.

Lex Fridman (12:36.760)

Why translation is Rev thinking of potentially going into other verticals

Dan Kokotov (12:41.600)

in the future, or is this like the focus now is a translation transcription, like

Dan Kokotov (12:46.280)

language, the focus now is, is language or, uh, speech services, generally speech

Dan Kokotov (12:52.200)

to text language services, you can kind of group them however you want.

Lex Fridman (12:56.000)

Um, so, but we, uh, originally the categorization was work from home.

Lex Fridman (13:01.960)

So when, uh, work that was done by people on a computer, you know, we weren't trying

Lex Fridman (13:06.280)

to get into, you know, uh, task rabbit type of things and something that could

Dan Kokotov (13:11.720)

be relatively standard, not a lot of options.

Lex Fridman (13:14.080)

So we could kind of present the simplified interface, right?

Lex Fridman (13:16.800)

So programming wasn't like a good fit because each programming

Lex Fridman (13:20.240)

project is kind of unique, right?

Dan Kokotov (13:21.640)

We're looking for something that, uh, Transcription is, you know, you have five

Lex Fridman (13:25.920)

hours of idea, it's five hours of audio, right?

Dan Kokotov (13:27.680)

Translation is somewhat similar.

Lex Fridman (13:29.200)

In that, you know, you can have a five page document, you know, and then you

Dan Kokotov (13:33.600)

just can price it by that and then you pick the language you want and that

Lex Fridman (13:37.480)

that's mostly all that is to it.

Lex Fridman (13:39.320)

So those were a few criteria.

Lex Fridman (13:40.760)

We started with translation because we saw the need, um, and, uh, we picked up

Dan Kokotov (13:48.880)

kind of a specialty of translation, um, where we would translate things like

Lex Fridman (13:52.800)

board certificates, uh, uh, immigration documents, things like that.

Lex Fridman (13:58.480)

And so they were fairly, um, even more well defined and easy to

Lex Fridman (14:03.640)

kind of tell if we did a good job.

Lex Fridman (14:05.200)

So you can literally charge per type of document.

Lex Fridman (14:07.680)

Was that, was, was that the, so what, what is it now?

Lex Fridman (14:10.960)

Is it per word or something like that?

Lex Fridman (14:12.600)

Like, how do you, like, how do you measure the effort

Lex Fridman (14:16.280)

involved in a particular thing?

Lex Fridman (14:18.320)

So now it looks like for audio translation, it's like,

Lex Fridman (14:21.080)

so now it looks a for audio transcription, right?

Lex Fridman (14:23.400)

It's a per audio minute.

Dan Kokotov (14:24.960)

Well, that, that yes, for, for, for our translation, we don't really,

Lex Fridman (14:28.120)

uh, actually focus it on anymore.

Dan Kokotov (14:30.240)

Uh, but you know, back when it was still a main business of Revit was per page,

Lex Fridman (14:35.000)

right.

Dan Kokotov (14:35.240)

Or per word, depending on the kind of, uh, cause you can also do translation

Lex Fridman (14:38.600)

now on the audio, right?

Dan Kokotov (14:40.600)

Mm hmm.

Lex Fridman (14:40.880)

So like subtitles.

Lex Fridman (14:41.960)

So it would be both, uh, transcription and translation.

Lex Fridman (14:45.040)

That's right.

Dan Kokotov (14:45.560)

I wanted to test the system to see how good it is to see like how, how, uh,

Lex Fridman (14:50.280)

well, is Russian supported?

Dan Kokotov (14:52.680)

I think so.

Lex Fridman (14:53.240)

Yeah.

Lex Fridman (14:54.040)

And it'd be interesting to try it out.

Lex Fridman (14:55.720)

I mean, one of the, now it's only in like the one direction, right?

Lex Fridman (14:58.080)

So you start with English and then you can have subtitles in Russian.

Lex Fridman (15:00.880)

Not really, not really the other way.

Dan Kokotov (15:02.800)

Got it.

Lex Fridman (15:03.200)

Because it's, um, I'm deeply curious about this.

Dan Kokotov (15:05.480)

Um, when COVID opens up a little bit, when the economy, when the

Lex Fridman (15:08.840)

world opens up a little bit, you want to build your brand in Russia?

Dan Kokotov (15:12.720)

No, I don't.

Lex Fridman (15:13.920)

First of all, I'm allergic to the word brand.

Dan Kokotov (15:15.720)

All right, I'm definitely not building, uh, any brands in Russia, but I'm going to

Lex Fridman (15:21.720)

Paris to talk to the translators of Dostoevsky and Tolstoy, there's this

Dan Kokotov (15:26.880)

famous couple that does translation.

Lex Fridman (15:29.600)

And, you know, I'm more and more thinking of how is it possible to have a

Lex Fridman (15:35.120)

conversation with a Russian speaker?

Lex Fridman (15:37.640)

Cause I have just some number of famous Russian speakers that

Dan Kokotov (15:42.640)

I'm interested in talking to, and my Russian is not strong

Lex Fridman (15:47.560)

enough to be witty and funny.

Dan Kokotov (15:49.720)

I'm already an idiot in English.

Lex Fridman (15:51.920)

I'm an extra level of like awkward idiot in Russian, but I can understand it.

Dan Kokotov (15:57.600)

Right.

Lex Fridman (15:58.200)

And I also like wonder how can I create a compelling English Russian

Lex Fridman (16:04.480)

experience for an English speaker?

Lex Fridman (16:06.480)

Like if I, there's a guy named Grigori Perlman, who's a mathematician who,

Dan Kokotov (16:11.640)

uh, obviously doesn't speak any English.

Lex Fridman (16:14.480)

So I would probably incorporate like a Russian translator into the picture.

Lex Fridman (16:21.440)

And then it would be like a, not to use a weird term, but like a three, like a

Dan Kokotov (16:25.320)

three, three person thing where it's like a dance of, like, I understand it one way.

Dan Kokotov (16:32.960)

They don't understand the other way, but I'll be asking questions in English.

Lex Fridman (16:38.200)

I don't know.

Dan Kokotov (16:38.840)

I don't know the right way.

Lex Fridman (16:39.800)

It's complicated.

Dan Kokotov (16:40.400)

It's complicated, but I feel like it's worth the effort for certain kinds of

Lex Fridman (16:43.920)

people, one of whom I'm confident of Vladimir Putin, I'm for sure talking to.

Dan Kokotov (16:48.880)

I really want to make it happen.

Lex Fridman (16:50.400)

Cause I think I could do a good job with, but the, the right, you know,

Dan Kokotov (16:54.240)

understanding the fundamentals of translation is something I'm really

Lex Fridman (16:58.000)

interested in.

Lex Fridman (16:58.920)

So that's why I'm starting with, um, the actual translators of like Russian

Lex Fridman (17:03.920)

literature, because they understand the nuance and the beauty of the

Dan Kokotov (17:06.640)

language and how it goes back and forth.

Lex Fridman (17:08.880)

But I also want to see, like in speech, how can we do it in real time?

Lex Fridman (17:13.240)

So that's, that's like a little bit of a baby project that I hope to push forward.

Lex Fridman (17:18.240)

But anyway, it's a challenging thing.

Lex Fridman (17:19.920)

So just to share, uh, my dad, um, actually does translation, um, not, not

Lex Fridman (17:25.840)

professional, he's a, uh, he writes poetry.

Dan Kokotov (17:28.160)

That was kind of always his, uh, not a hobby, but he's, uh, he, you know, he

Lex Fridman (17:33.480)

had a job, like a day job, but his passion was always writing poetry.

Dan Kokotov (17:36.840)

Uh, and then when I got to America, like he started also translating, um, first

Lex Fridman (17:43.080)

he was translating English poetry to Russia.

Dan Kokotov (17:45.120)

Now he also like goes the other, uh, the other way, you kind of gain some small

Lex Fridman (17:50.400)

fame in that world anyways, because, uh, recently this poet like Lewis

Dan Kokotov (17:54.840)

clock, I don't know if you know of, uh, some American poet, um, she was

Lex Fridman (17:59.080)

awarded the Nobel prize for literature.

Dan Kokotov (18:01.240)

Uh, and so my dad had translated, uh, one of her books of poetry in

Lex Fridman (18:05.840)

to Russian, and he was like one of the few.

Lex Fridman (18:07.640)

So he kind of like, they asked him and gave an interview to Radiosvoboda,

Lex Fridman (18:12.160)

if you know what that is.

Lex Fridman (18:13.000)

And he kind of talked about some of the intricacies of translating poetry.

Lex Fridman (18:17.000)

So that's like an extra level of difficulty, right?

Dan Kokotov (18:18.640)

Because translating poetry is even more challenging than

Lex Fridman (18:21.600)

translating just, you know, interviews.

Lex Fridman (18:24.280)

Do you remember any, any experiences and challenges to having to do the

Lex Fridman (18:28.560)

translation that, that's the God to you, like something he's talked about?

Lex Fridman (18:32.640)

I mean, a lot of it, I think is word choice, right?

Lex Fridman (18:35.400)

It's the way Russian is structured is first of all, quite different

Lex Fridman (18:38.160)

than, um, the way English is structured, right?

Dan Kokotov (18:40.200)

Just there is inflections in Russian and genders, and they don't exist in English.

Dan Kokotov (18:43.880)

That's just one of the reasons actually why, um, machine translation is quite

Lex Fridman (18:47.760)

difficult for English to Russian and Russian to English, because there's

Dan Kokotov (18:50.840)

such different languages, but then English has like a huge number of words.

Lex Fridman (18:55.240)

Um, many more than Russian, actually, I think.

Lex Fridman (18:57.040)

So it's often difficult to find the right word to convey the same emotional

Lex Fridman (19:01.400)

meaning, yeah, Russian language.

Dan Kokotov (19:03.440)

They play with words much more.

Lex Fridman (19:06.240)

So you, you're mentioning that, uh, Rev was kind of born out of, um, trying to

Dan Kokotov (19:12.080)

take a vertical on the upwork and then standardize it.

Lex Fridman (19:15.760)

So we're just trying to make the, the freelancer marketplace idea better, right?

Dan Kokotov (19:21.000)

Um, better for both customers and better for the freelancers themselves.

Lex Fridman (19:27.080)

Is there something else to the story of Rev finding the right word?

Lex Fridman (19:30.840)

Rev, finding Rev, like what, what did it take to bring it to actually to life?

Lex Fridman (19:35.720)

Was there any pain points?

Dan Kokotov (19:37.200)

Um, plenty of, plenty of pain points.

Lex Fridman (19:39.840)

I mean, uh, as, as often the case it's with scaling it up, right?

Dan Kokotov (19:43.920)

Um, and in this case, you know, the scaling is kind of scaling the,

Lex Fridman (19:48.080)

the marketplace, so to speak, right?

Lex Fridman (19:49.280)

Rev is essentially a two sided marketplace, right?

Lex Fridman (19:51.600)

Because, you know, there's the customers and then there's the reverse.

Dan Kokotov (19:55.360)

Um, if there's not enough Revers, Revers are world color freelancers.

Lex Fridman (19:59.000)

So if there's not enough Revers, then customers have a bad experience, right?

Dan Kokotov (1:00:01.980)

Yeah, I'm with you.

Lex Fridman (1:00:02.820)

I tend to believe in the intelligence of people

Lex Fridman (1:00:04.520)

and we should trust them.

Lex Fridman (1:00:06.960)

But I also do think it's the responsibility of platforms

Dan Kokotov (1:00:10.940)

to encourage more love in the world,

Lex Fridman (1:00:12.640)

more kindness to each other.

Lex Fridman (1:00:14.120)

And I don't always think that they're great

Lex Fridman (1:00:16.920)

at doing that particular thing.

Lex Fridman (1:00:19.240)

So that, there's a nice balance there.

Lex Fridman (1:00:25.400)

And I think philosophically, I think about that a lot.

Dan Kokotov (1:00:28.240)

Where's the balance between free speech

Lex Fridman (1:00:32.040)

and like encouraging people,

Dan Kokotov (1:00:35.000)

even though they have the freedom of speech

Lex Fridman (1:00:37.880)

to not be an asshole.

Dan Kokotov (1:00:39.520)

Yeah, right.

Lex Fridman (1:00:41.020)

That's not a constitutional, like...

Lex Fridman (1:00:44.720)

So you have the right for free speech,

Lex Fridman (1:00:48.100)

but like, just don't be an asshole.

Dan Kokotov (1:00:50.640)

Like you can't really put that in the constitution

Lex Fridman (1:00:52.640)

that the Supreme Court can't be like,

Dan Kokotov (1:00:54.160)

eh, just don't be a dick.

Lex Fridman (1:00:56.040)

But I feel like platforms have a role to be like,

Dan Kokotov (1:00:59.520)

just be nicer.

Lex Fridman (1:01:00.800)

Maybe do the carrot, like encourage people to be nicer

Dan Kokotov (1:01:04.160)

as opposed to the stake of censorship.

Lex Fridman (1:01:06.760)

But I think it's an interesting machine learning problem.

Dan Kokotov (1:01:11.020)

Just be nicer.

Lex Fridman (1:01:12.960)

Machine, yeah, machine learning for niceness.

Dan Kokotov (1:01:15.760)

It is, I mean, that's...

Lex Fridman (1:01:16.600)

Responsible, yeah, I mean, it is.

Dan Kokotov (1:01:17.880)

It is a thing, for sure.

Lex Fridman (1:01:20.080)

Jack Dorsey kind of talks about it

Dan Kokotov (1:01:22.080)

as a vision for Twitter is,

Lex Fridman (1:01:23.880)

how do we increase the health of conversations?

Dan Kokotov (1:01:26.760)

I don't know how seriously

Lex Fridman (1:01:28.060)

they're actually trying to do that though.

Dan Kokotov (1:01:30.720)

Which is one of the reasons that I'm in part considering

Lex Fridman (1:01:35.780)

entering that space a little bit.

Lex Fridman (1:01:37.240)

It's difficult for them, right?

Lex Fridman (1:01:38.480)

Because, you know, it's kind of like well known

Dan Kokotov (1:01:40.520)

that people are kind of driven by rage

Lex Fridman (1:01:45.520)

and you know, outrage maybe is a better word, right?

Dan Kokotov (1:01:49.400)

Outrage drives engagement.

Lex Fridman (1:01:51.840)

And well, these companies are judged by engagement, right?

Lex Fridman (1:01:55.560)

So it's...

Lex Fridman (1:01:56.400)

In the short term, but this goes to the metrics thing

Dan Kokotov (1:01:58.160)

that we were talking about earlier.

Lex Fridman (1:01:59.320)

I do believe, I have a fundamental belief

Dan Kokotov (1:02:02.240)

that if you have a metric of long term happiness

Lex Fridman (1:02:06.140)

of your users, like not short term engagement,

Lex Fridman (1:02:09.520)

but long term happiness and growth

Lex Fridman (1:02:11.440)

and both like intellectual, emotional health of your users,

Dan Kokotov (1:02:15.520)

you're going to make a lot more money.

Lex Fridman (1:02:17.600)

You're going to have long...

Dan Kokotov (1:02:18.840)

Like you should be able to optimize for that.

Lex Fridman (1:02:21.400)

You don't need to necessarily optimize for engagement.

Lex Fridman (1:02:24.280)

And that'll be good for society too.

Lex Fridman (1:02:26.400)

Yeah, no, I mean, I generally agree with you,

Lex Fridman (1:02:28.800)

but it requires a patient person with, you know,

Lex Fridman (1:02:33.320)

trust from Wall Street to be able

Dan Kokotov (1:02:35.520)

to carry out such a strategy.

Lex Fridman (1:02:36.720)

This is what I believe the Steve Jobs character

Lex Fridman (1:02:39.240)

and Elon Musk character is like,

Lex Fridman (1:02:41.920)

you basically have to be so good at your job.

Dan Kokotov (1:02:45.240)

Right, you got to pass for anything.

Lex Fridman (1:02:46.960)

That you can hold the board

Lex Fridman (1:02:48.680)

and all the investors hostage by saying like,

Lex Fridman (1:02:52.000)

either we do it my way or I leave.

Lex Fridman (1:02:56.360)

And everyone is too afraid of you to leave

Lex Fridman (1:02:59.120)

because they believe in your vision.

Lex Fridman (1:03:01.360)

But that requires being really good at what you do.

Lex Fridman (1:03:04.280)

It requires being Steve Jobs and Elon Musk.

Dan Kokotov (1:03:06.680)

There's kind of a reason why like a third name doesn't

Lex Fridman (1:03:09.280)

come immediately to mind, right?

Dan Kokotov (1:03:10.840)

Like there's maybe a handful of other people,

Lex Fridman (1:03:12.360)

but it's not that many.

Dan Kokotov (1:03:13.440)

It's not many.

Lex Fridman (1:03:14.280)

I mean, people say like, why are you...

Dan Kokotov (1:03:15.480)

Like people say that I'm like a fan of Elon Musk.

Lex Fridman (1:03:18.320)

I'm not, I'm a fan of anybody

Dan Kokotov (1:03:20.960)

who's like Steve Jobs and Elon Musk.

Lex Fridman (1:03:23.080)

And there's just not many of those folks.

Dan Kokotov (1:03:26.320)

It's the guy that made us believe

Lex Fridman (1:03:27.640)

that like we can get to Mars, you know, in 10 years, right?

Dan Kokotov (1:03:31.040)

I mean, that's kind of awesome.

Lex Fridman (1:03:32.480)

And it's kind of making it happen, which is like...

Lex Fridman (1:03:35.280)

And it's kind of gone like that kind of like spirit, right?

Lex Fridman (1:03:40.480)

Like from a lot of our society, right?

Dan Kokotov (1:03:42.240)

You know, like we can get to the moon in 10 years

Lex Fridman (1:03:44.640)

and like we did it, right?

Dan Kokotov (1:03:45.640)

Yeah.

Lex Fridman (1:03:46.480)

Especially in this time of so much kind of existential dread

Dan Kokotov (1:03:52.840)

that people are going through because of COVID,

Lex Fridman (1:03:55.440)

like having rockets that just keep going out there

Dan Kokotov (1:03:58.520)

now with humans.

Lex Fridman (1:04:00.400)

I don't know that it, just like you said,

Dan Kokotov (1:04:03.200)

I mean, it gives you a reason to wake up in the morning

Lex Fridman (1:04:05.520)

and dream, for us engineers too.

Dan Kokotov (1:04:09.960)

It is inspiring as hell, man.

Lex Fridman (1:04:14.360)

Let me ask you this, the worst possible question,

Dan Kokotov (1:04:17.080)

which is, so you're like at the core, you're a programmer,

Lex Fridman (1:04:21.280)

you're an engineer, but now you made the unfortunate choice

Dan Kokotov (1:04:29.040)

or maybe that's the way life goes

Lex Fridman (1:04:30.680)

of basically moving away from the low level work

Lex Fridman (1:04:35.080)

and becoming a manager, becoming an executive,

Lex Fridman (1:04:38.080)

having meetings, what's that transition been like?

Dan Kokotov (1:04:43.200)

It's been interesting.

Lex Fridman (1:04:44.040)

It's been a journey.

Dan Kokotov (1:04:44.880)

Maybe a couple of things to say about that.

Lex Fridman (1:04:46.840)

I mean, I got into this, right?

Dan Kokotov (1:04:49.280)

Because as a kid, I just remember this like incredible

Lex Fridman (1:04:54.440)

amazement at being able to write a program, right?

Lex Fridman (1:04:57.320)

And something comes to life that kind of didn't exist before.

Lex Fridman (1:05:01.200)

I don't think you have that in like many other fields,

Dan Kokotov (1:05:03.920)

like you have that with some other kinds of engineering,

Lex Fridman (1:05:07.880)

but you're maybe a little bit more limited

Lex Fridman (1:05:09.640)

with what you can do, right?

Lex Fridman (1:05:10.680)

But with a computer,

Lex Fridman (1:05:11.640)

you can literally imagine any kind of program, right?

Lex Fridman (1:05:14.760)

So it's a little bit godlike what you do

Dan Kokotov (1:05:16.960)

like when you create it.

Lex Fridman (1:05:19.200)

And so, I mean, that's why I got into it.

Lex Fridman (1:05:21.320)

Do you remember like first program you wrote

Lex Fridman (1:05:23.200)

or maybe the first program that like made you fall in love

Lex Fridman (1:05:25.800)

with computer science?

Lex Fridman (1:05:28.320)

I don't know what was the first program.

Dan Kokotov (1:05:29.400)

It's probably like trying to write one of those games

Lex Fridman (1:05:31.840)

and basic, you know, like emulate the snake game

Dan Kokotov (1:05:34.040)

or whatever.

Lex Fridman (1:05:35.400)

I don't remember to be honest, but I enjoyed like,

Dan Kokotov (1:05:37.800)

that's why I always loved about, you know,

Lex Fridman (1:05:40.000)

being a programmer, it's just the creation process.

Lex Fridman (1:05:41.800)

And it's a little bit different

Lex Fridman (1:05:43.800)

when you're not the one doing the creating.

Dan Kokotov (1:05:47.520)

And, you know, another aspect to it I would say is,

Lex Fridman (1:05:50.480)

you know, when you're a programmer,

Dan Kokotov (1:05:52.040)

when you're a individual contributor,

Lex Fridman (1:05:54.160)

it's kind of very easy to know when you're doing a good job,

Dan Kokotov (1:05:57.800)

when you're not doing a good job,

Lex Fridman (1:05:58.640)

when you're being productive,

Lex Fridman (1:05:59.480)

when you're not being productive, right?

Lex Fridman (1:06:00.360)

You can kind of see like you trying to make something

Lex Fridman (1:06:03.000)

and it's like slowly coming together, right?

Lex Fridman (1:06:05.560)

And when you're a manager, you know, it's more diffuse,

Lex Fridman (1:06:08.880)

right?

Lex Fridman (1:06:09.720)

Like, well, you hope, you know, you're motivating your team

Lex Fridman (1:06:12.760)

and making them more productive and inspiring them, right?

Lex Fridman (1:06:15.920)

But it's not like you get some kind of like dopamine signal

Dan Kokotov (1:06:18.920)

because you like completed X lines of code, you know, today.

Lex Fridman (1:06:22.440)

So kind of like you missed that dopamine rush

Dan Kokotov (1:06:24.040)

a little bit when you first become,

Lex Fridman (1:06:28.040)

but then, you know, slowly you kind of see,

Lex Fridman (1:06:30.440)

yes, your teams are doing amazing work, right?

Lex Fridman (1:06:32.360)

And you can take pride in that.

Lex Fridman (1:06:34.720)

You can get like, what is it?

Lex Fridman (1:06:38.240)

Like a ripple effect of somebody else's dopamine rush.

Dan Kokotov (1:06:41.640)

Yeah, yeah, you live off other people's dopamine.

Lex Fridman (1:06:46.760)

So is there pain points and challenges you had to overcome

Dan Kokotov (1:06:50.800)

from becoming, from going to a programmer to becoming

Lex Fridman (1:06:54.240)

a programmer of humans?

Dan Kokotov (1:06:55.960)

Programmer of humans.

Lex Fridman (1:06:58.360)

I don't know, humans are difficult to understand,

Dan Kokotov (1:07:00.960)

you know, it's like one of those things,

Lex Fridman (1:07:03.640)

like trying to understand other people's motivations

Lex Fridman (1:07:06.720)

and what really drives them.

Lex Fridman (1:07:08.200)

It's difficult, maybe like never really know, right?

Lex Fridman (1:07:10.840)

Do you find that people are different?

Lex Fridman (1:07:13.320)

Yeah.

Dan Kokotov (1:07:14.160)

Like I, one of the things, like I had a group at MIT

Lex Fridman (1:07:18.680)

that, you know, I found that like some people

Dan Kokotov (1:07:24.680)

I could like scream at and criticize like hard

Lex Fridman (1:07:30.080)

and that made them do like much better work

Lex Fridman (1:07:32.800)

and really push them to their limit.

Lex Fridman (1:07:35.160)

And there's some people that I had to nonstop compliment

Dan Kokotov (1:07:39.080)

because like they're so already self critical,

Lex Fridman (1:07:42.800)

like about everything they do that I have to be constantly

Dan Kokotov (1:07:45.840)

like, like I cannot criticize them at all

Lex Fridman (1:07:50.920)

because they're already criticizing themselves

Lex Fridman (1:07:52.840)

and you have to kind of encourage

Lex Fridman (1:07:54.760)

and like celebrate their little victories.

Lex Fridman (1:07:58.080)

And it's kind of fascinating that like how that,

Lex Fridman (1:08:00.960)

the complete difference in people.

Dan Kokotov (1:08:03.480)

Definitely people respond to different motivations

Lex Fridman (1:08:06.400)

and different loads of feedback

Lex Fridman (1:08:07.600)

and you kind of have to figure it out.

Lex Fridman (1:08:10.720)

It was like a pretty good book,

Dan Kokotov (1:08:13.080)

which for some reason now the name escapes me,

Lex Fridman (1:08:15.440)

about management, first break all the rules.

Lex Fridman (1:08:18.800)

First break all the rules?

Lex Fridman (1:08:19.720)

First break all the rules.

Dan Kokotov (1:08:20.800)

It's a book that we generally like ask a lot of

Lex Fridman (1:08:23.880)

like first time managers to read it rough.

Lex Fridman (1:08:26.320)

And like one of the kind of philosophies

Lex Fridman (1:08:28.760)

is managed by exception, right?

Dan Kokotov (1:08:31.080)

Which is, you know, don't like have some standard template

Lex Fridman (1:08:34.440)

like, you know, here's how I, you know,

Dan Kokotov (1:08:36.920)

tell this person to do this or the other thing.

Lex Fridman (1:08:39.160)

Here's how I get feedback, like manage by exception, right?

Dan Kokotov (1:08:41.240)

Every person is a little bit different.

Lex Fridman (1:08:42.800)

You have to try to understand what drives them.

Lex Fridman (1:08:45.280)

And tailor it to them.

Lex Fridman (1:08:47.200)

Since you mentioned books,

Dan Kokotov (1:08:48.880)

I don't know if you can answer this question,

Lex Fridman (1:08:50.800)

but people love it when I ask it, which is,

Dan Kokotov (1:08:53.600)

are there books, technical fiction or philosophical

Lex Fridman (1:08:56.880)

that you enjoyed or had an impact on your life

Lex Fridman (1:08:59.800)

that you would recommend?

Lex Fridman (1:09:01.320)

You already mentioned Dune, like all of the Dune.

Dan Kokotov (1:09:04.400)

All of the Dune.

Lex Fridman (1:09:05.360)

The second one was probably the weakest, but anyway,

Lex Fridman (1:09:07.080)

so yeah, all of the Dune is good.

Lex Fridman (1:09:09.760)

I mean, yeah, can you just slow little tangent on that?

Lex Fridman (1:09:13.440)

Is, how many Dune books are there?

Lex Fridman (1:09:16.280)

Like, do you recommend people start with the first one

Lex Fridman (1:09:18.440)

if that was?

Lex Fridman (1:09:19.880)

Yeah, you gotta have to read them all.

Lex Fridman (1:09:21.040)

I mean, it is a complete story, right?

Lex Fridman (1:09:23.240)

So you start with the first one,

Dan Kokotov (1:09:25.440)

you gotta read all of them.

Lex Fridman (1:09:27.520)

So it's not like a tree, like a creation of like

Lex Fridman (1:09:31.640)

the universe that you should go in sequence?

Lex Fridman (1:09:33.880)

You should go in sequence, yeah.

Dan Kokotov (1:09:35.200)

It's kind of a chronological storyline.

Lex Fridman (1:09:38.000)

There's six books in all.

Dan Kokotov (1:09:40.120)

Then there's like many kind of books

Lex Fridman (1:09:43.720)

that were written by Frank Herbert's son,

Lex Fridman (1:09:47.880)

but those are not as good.

Lex Fridman (1:09:48.840)

So you don't have to bother with those.

Dan Kokotov (1:09:50.880)

Shots fired.

Lex Fridman (1:09:51.720)

Shots fired.

Dan Kokotov (1:09:52.920)

Okay.

Lex Fridman (1:09:53.760)

But the main sequence is good.

Lex Fridman (1:09:56.280)

So what are some other books?

Lex Fridman (1:09:58.840)

Maybe there's a few.

Lex Fridman (1:09:59.680)

So I don't know that like, I would say there's a book

Lex Fridman (1:10:02.000)

that kind of, I don't know, turned my life around

Dan Kokotov (1:10:05.440)

or anything like that, but here's a couple

Lex Fridman (1:10:07.880)

that I really love.

Lex Fridman (1:10:08.960)

So one is Brave New World by Aldous Huxley.

Lex Fridman (1:10:16.040)

And it's kind of incredible how prescient he was

Lex Fridman (1:10:20.160)

about like what a brave new world might be like, right?

Lex Fridman (1:10:25.440)

You know, you kind of see genetic sorting in this book,

Dan Kokotov (1:10:28.200)

right, where there's like these alphas and epsilons

Lex Fridman (1:10:30.720)

and how from like the earliest time of society,

Dan Kokotov (1:10:34.920)

like they're sort of like, you can kind of see it

Lex Fridman (1:10:36.760)

in a slightly similar way today where,

Dan Kokotov (1:10:40.640)

well, one of the problems with society is people

Lex Fridman (1:10:42.720)

are kind of genetically sorting a little bit, right?

Dan Kokotov (1:10:45.960)

Like there's much less, like most marriages, right,

Lex Fridman (1:10:48.960)

are between people of similar kind of intellectual level

Dan Kokotov (1:10:53.360)

or socioeconomic status, more so these days than in the past.

Lex Fridman (1:10:57.560)

And you kind of see some effects of it

Dan Kokotov (1:10:59.080)

in stratifying society and kind of he illustrated

Lex Fridman (1:11:03.720)

what that could be like in the extreme.

Dan Kokotov (1:11:05.840)

There's different versions of it on social media as well.

Lex Fridman (1:11:07.920)

It's not just like marriages and so on.

Dan Kokotov (1:11:09.840)

Like it's genetic sorting in terms of what Dawkins called

Lex Fridman (1:11:13.440)

memes as ideas being put into these bins

Dan Kokotov (1:11:17.320)

of these little echo chambers and so on.

Lex Fridman (1:11:20.000)

Yeah, I know, so that's the book

Dan Kokotov (1:11:21.440)

that's I think a worthwhile read for everyone.

Lex Fridman (1:11:23.560)

I mean, 1984 is good, of course, as well.

Dan Kokotov (1:11:25.240)

Like if you're talking about, you know,

Lex Fridman (1:11:26.520)

dystopian novels of the future.

Lex Fridman (1:11:28.200)

Yeah, it's a slightly different view of the future, right?

Lex Fridman (1:11:30.520)

But I kind of like identify with Brave New World a bit more.

Dan Kokotov (1:11:33.600)

Yeah, speaking of not a book,

Lex Fridman (1:11:38.040)

but my favorite kind of dystopian science fiction

Dan Kokotov (1:11:41.480)

is a movie called Brazil,

Lex Fridman (1:11:42.600)

which I don't know if you've heard of.

Dan Kokotov (1:11:44.160)

I've heard of and I know I need to watch it,

Lex Fridman (1:11:46.320)

but yeah, because it's in, is it in English or no?

Dan Kokotov (1:11:50.520)

It's an English movie, yeah.

Lex Fridman (1:11:52.120)

And it's a sort of like dystopian movie

Lex Fridman (1:11:55.800)

of authoritarian incompetence, right?

Lex Fridman (1:11:58.640)

It's like nothing really works very well, you know,

Dan Kokotov (1:12:03.640)

the system is creaky, you know,

Lex Fridman (1:12:05.720)

but no one is kind of like willing to challenge it,

Dan Kokotov (1:12:07.800)

you know, just things kind of ample along

Lex Fridman (1:12:10.040)

and kind of strikes me as like a very plausible future

Dan Kokotov (1:12:13.680)

of like, you know, what authoritarianism might look like.

Lex Fridman (1:12:16.840)

It's not like this, you know,

Dan Kokotov (1:12:19.240)

super efficient evil dictatorship of 1984.

Lex Fridman (1:12:21.880)

It's just kind of like this badly functioning, you know,

Lex Fridman (1:12:25.240)

but it's status quo, so it just goes on.

Lex Fridman (1:12:30.080)

Yeah, that's one funny thing that stands out to me

Dan Kokotov (1:12:33.520)

is in whether it's authoritarian, dystopian stuff,

Lex Fridman (1:12:37.120)

or just basic like, you know,

Dan Kokotov (1:12:39.480)

if you look at the movie Contagion,

Lex Fridman (1:12:42.400)

it seems in the movies,

Dan Kokotov (1:12:44.480)

government is almost always exceptionally competent.

Lex Fridman (1:12:49.600)

Like it's like used as a storytelling tool

Dan Kokotov (1:12:53.200)

of like extreme competence.

Lex Fridman (1:12:55.480)

Like, you know, you use it whether it's good or evil,

Lex Fridman (1:12:58.360)

but it's competent.

Lex Fridman (1:12:59.680)

It's very interesting to think about

Dan Kokotov (1:13:01.840)

where much more realistically is it's incompetence

Lex Fridman (1:13:06.440)

and that incompetence isn't itself has consequences

Dan Kokotov (1:13:11.280)

that are difficult to predict.

Lex Fridman (1:13:13.200)

Like bureaucracy has a very boring way of being evil,

Dan Kokotov (1:13:19.040)

of just, you know, if you look at the show,

Lex Fridman (1:13:21.400)

HBO show at Chernobyl,

Dan Kokotov (1:13:23.160)

it's a really good story of how bureaucracy, you know,

Lex Fridman (1:13:30.200)

leads to catastrophic events,

Lex Fridman (1:13:32.800)

but not through any kind of evil

Lex Fridman (1:13:34.320)

in any one particular place,

Lex Fridman (1:13:36.200)

but more just like the...

Lex Fridman (1:13:37.720)

It's just the system kind of system.

Dan Kokotov (1:13:40.040)

Distorting information as it travels up the chain

Lex Fridman (1:13:43.280)

that people unwilling to take responsibility for things

Lex Fridman (1:13:46.040)

and just kind of like this laziness resulting in evil.

Lex Fridman (1:13:50.960)

There's a comedic version of this,

Dan Kokotov (1:13:52.320)

I don't know if you've seen this movie,

Lex Fridman (1:13:53.600)

it's called The Death of Stalin.

Dan Kokotov (1:13:54.960)

Yeah, I liked that.

Lex Fridman (1:13:58.080)

I wish it wasn't so...

Dan Kokotov (1:14:00.080)

There's a movie called Inglourious Bastards

Lex Fridman (1:14:02.360)

about, you know, Hitler and so on.

Dan Kokotov (1:14:07.760)

For some reason, those movies pissed me off.

Lex Fridman (1:14:09.640)

I know a lot of people love them,

Lex Fridman (1:14:11.120)

but like, I just feel like there's not enough good movies,

Lex Fridman (1:14:16.640)

even about Hitler.

Dan Kokotov (1:14:18.520)

There's good movies about the Holocaust,

Lex Fridman (1:14:21.400)

but even Hitler, there's a movie called Dawnfall

Dan Kokotov (1:14:23.640)

that people should watch.

Lex Fridman (1:14:24.480)

I think it's the last few days of Hitler.

Dan Kokotov (1:14:26.080)

That's a good movie, turned into a meme, but it's good.

Lex Fridman (1:14:30.440)

But on Stalin, I feel like I may be wrong on this,

Lex Fridman (1:14:33.760)

but at least in the English speaking world,

Lex Fridman (1:14:35.520)

there's not good movies about the evil of Stalin.

Dan Kokotov (1:14:38.760)

That's true.

Lex Fridman (1:14:39.600)

Let's try to see that.

Dan Kokotov (1:14:40.640)

Actually, so I agree with you on Inglourious Bastard.

Lex Fridman (1:14:43.320)

I didn't love the movie

Lex Fridman (1:14:46.000)

because I felt like kind of the stylizing of it, right?

Lex Fridman (1:14:50.000)

The whole Tarantino kind of Tarantinoism, if you will,

Dan Kokotov (1:14:55.040)

kind of detracted from it

Lex Fridman (1:14:56.080)

and made it seem like unserious a little bit.

Lex Fridman (1:15:00.280)

But Death of Stalin, I felt differently.

Lex Fridman (1:15:02.240)

Maybe it's because it's a comedy to begin with.

Dan Kokotov (1:15:03.800)

This is not like I'm expecting seriousness,

Lex Fridman (1:15:06.560)

but it kind of depicted the absurdity

Lex Fridman (1:15:10.760)

of the whole situation in a way, right?

Lex Fridman (1:15:13.360)

I mean, it was funny, so maybe it does make light of it,

Lex Fridman (1:15:15.280)

but something goes probably like this, right?

Lex Fridman (1:15:18.200)

Like a bunch of kind of people,

Lex Fridman (1:15:20.440)

they're like, oh shit, right?

Lex Fridman (1:15:22.440)

You're right.

Lex Fridman (1:15:23.280)

But like the thing is,

Lex Fridman (1:15:25.440)

it was so close to like what probably was reality.

Dan Kokotov (1:15:32.720)

It was caricaturing reality

Lex Fridman (1:15:35.480)

to where I think an observer might think that this is not,

Dan Kokotov (1:15:39.320)

like they might think it's a comedy.

Lex Fridman (1:15:41.560)

Well, in reality, that's the absurdity

Dan Kokotov (1:15:45.560)

of how people act with dictators.

Lex Fridman (1:15:48.800)

I mean, that's, I guess it was too close to reality for me.

Dan Kokotov (1:15:54.000)

The kind of banality of like what were eventually

Lex Fridman (1:15:57.720)

like fairly evil acts, right?

Lex Fridman (1:15:59.440)

But like, yeah, they're just a bunch of people

Lex Fridman (1:16:02.280)

trying to survive.

Dan Kokotov (1:16:04.400)

Cause I think there's a good,

Lex Fridman (1:16:05.440)

I haven't watched it yet, the good movie on,

Dan Kokotov (1:16:07.560)

the movie on Churchill with Gary Oldman,

Lex Fridman (1:16:12.440)

I think it's Gary Oldman.

Dan Kokotov (1:16:14.000)

I may be making that up.

Lex Fridman (1:16:15.400)

But I think he won,

Dan Kokotov (1:16:16.240)

like he was nominated for an Oscar or something.

Lex Fridman (1:16:17.960)

So I like, I love these movies about these humans

Lex Fridman (1:16:20.960)

and Stalin, like Chernobyl made me realize the HBO show

Lex Fridman (1:16:26.000)

that there's not enough movies about Russia

Dan Kokotov (1:16:28.760)

that capture that spirit.

Lex Fridman (1:16:33.040)

I'm sure it might be in Russian there is,

Lex Fridman (1:16:35.640)

but the fact that some British dude that like did comedy,

Lex Fridman (1:16:39.280)

I feel like he did like hangover or some shit like that.

Dan Kokotov (1:16:42.120)

I don't know if you're familiar

Lex Fridman (1:16:43.080)

with the person who created Chernobyl,

Lex Fridman (1:16:44.400)

but he was just like some guy

Lex Fridman (1:16:45.720)

that doesn't know anything about Russia.

Lex Fridman (1:16:47.320)

And he just went in and just studied it,

Lex Fridman (1:16:49.760)

like did a good job of creating it

Lex Fridman (1:16:51.880)

and then got it so accurate, like poetically.

Lex Fridman (1:16:56.080)

And the facts that you need to get accurate,

Dan Kokotov (1:16:58.880)

he got accurate, just the spirit of it

Lex Fridman (1:17:01.120)

down to like the bowls that pets use,

Dan Kokotov (1:17:03.760)

just the whole feel of it.

Lex Fridman (1:17:05.080)

It was incredible.

Dan Kokotov (1:17:05.920)

It was good, yeah, I saw the series.

Lex Fridman (1:17:07.480)

Yeah, it's incredible.

Dan Kokotov (1:17:08.640)

It's made me wish that somebody did a good,

Lex Fridman (1:17:10.920)

like 1930s, like starvation that Stalin led to,

Dan Kokotov (1:17:17.880)

like leading up to World War II

Lex Fridman (1:17:20.160)

and in World War II itself, like Stalingrad and so on.

Dan Kokotov (1:17:23.680)

Like, I feel like that story needs to be told.

Lex Fridman (1:17:27.520)

Millions of people died.

Lex Fridman (1:17:30.080)

And to me, it's so much more fascinating than Hitler

Lex Fridman (1:17:32.800)

because Hitler is like a caricature of evil almost

Dan Kokotov (1:17:37.560)

that it's so, especially with the Holocaust,

Lex Fridman (1:17:41.800)

it's so difficult to imagine that something like that

Dan Kokotov (1:17:45.520)

is possible ever again.

Lex Fridman (1:17:47.520)

Stalin to me represents something that is possible.

Dan Kokotov (1:17:52.720)

Like the so interesting, like the bureaucracy of it

Lex Fridman (1:17:56.720)

is so fascinating that it potentially might be happening

Dan Kokotov (1:18:01.200)

in the world now, like that we're not aware of,

Lex Fridman (1:18:03.200)

like with North Korea, another one that,

Dan Kokotov (1:18:06.120)

like there should be a good film on.

Lex Fridman (1:18:08.240)

And like the possible things that could be happening

Dan Kokotov (1:18:10.640)

in China with overreach of government.

Lex Fridman (1:18:13.080)

I don't know, there's a lot of possibilities there.

Dan Kokotov (1:18:15.600)

I suppose.

Lex Fridman (1:18:16.440)

Yeah, I wonder how much, you know,

Dan Kokotov (1:18:18.320)

I guess the archives should be maybe more open nowadays,

Lex Fridman (1:18:20.440)

right, I mean, for a long time, they just, we didn't know,

Dan Kokotov (1:18:22.960)

right, or anyways, no one in the West knew for sure.

Lex Fridman (1:18:25.920)

Well, there's a, I don't know if you know him,

Dan Kokotov (1:18:27.560)

there's a guy named Stephen Kotkin.

Lex Fridman (1:18:29.440)

He is a historian of Stalin that I spoke to on this podcast.

Dan Kokotov (1:18:33.080)

I'll speak to him again.

Lex Fridman (1:18:34.760)

The guy knows his shit on Stalin.

Dan Kokotov (1:18:38.040)

He like read everything and it's so fascinating

Lex Fridman (1:18:44.160)

to talk to somebody, like he knows Stalin better

Dan Kokotov (1:18:50.280)

than Stalin himself, it's crazy.

Lex Fridman (1:18:53.000)

Like you have, so he's, I think he's a Princeton,

Dan Kokotov (1:18:55.440)

he is basically, his whole life is Stalin.

Lex Fridman (1:18:58.840)

Fighting Stalin.

Dan Kokotov (1:18:59.760)

Yeah, it's great.

Lex Fridman (1:19:00.920)

And in that context, he also talks about

Lex Fridman (1:19:03.640)

and writes about Putin a little bit.

Lex Fridman (1:19:06.040)

I've also read at this point,

Dan Kokotov (1:19:07.880)

I think every biography of Putin, English biography of Putin,

Lex Fridman (1:19:12.360)

I need to read some Russians.

Dan Kokotov (1:19:14.000)

Obviously, I'm mentally preparing

Lex Fridman (1:19:15.400)

for a possible conversation with Putin.

Lex Fridman (1:19:17.000)

So what is your first question to Putin

Lex Fridman (1:19:19.400)

when you have him on the podcast?

Dan Kokotov (1:19:22.480)

I, it's interesting you bring that up.

Lex Fridman (1:19:26.400)

First of all, I wouldn't tell you, but.

Dan Kokotov (1:19:27.920)

You can't give it away now.

Lex Fridman (1:19:30.680)

But I actually haven't even thought about that.

Lex Fridman (1:19:34.360)

So my current approach, and I do this with interviews often,

Lex Fridman (1:19:38.760)

obviously that's a special one,

Lex Fridman (1:19:40.160)

but I try not to think about questions until last minute.

Lex Fridman (1:19:45.120)

I'm trying to sort of get into the mindset.

Lex Fridman (1:19:50.000)

And so that's why I'm soaking in a lot of stuff,

Lex Fridman (1:19:52.280)

not thinking about questions, just learning about the man.

Lex Fridman (1:19:56.120)

But in terms of like human to human,

Lex Fridman (1:19:59.680)

it's like, I would say it's,

Dan Kokotov (1:20:01.360)

I don't know if you're a fan of mob movies,

Lex Fridman (1:20:03.440)

but like the mafia, which I am, like Goodfellas and so on,

Dan Kokotov (1:20:07.000)

he's much closer to like mob morality, which is like.

Lex Fridman (1:20:11.960)

Mob morality, maybe, I could see that.

Lex Fridman (1:20:14.000)

But I like your approach anyways of this,

Lex Fridman (1:20:16.600)

the extreme empathy, right?

Lex Fridman (1:20:18.160)

It's a little bit like Hannibal, right?

Lex Fridman (1:20:21.360)

Like if you ever watched the show Hannibal, right?

Dan Kokotov (1:20:22.960)

They had that guy, well, you know Hannibal of course, like.

Lex Fridman (1:20:27.560)

Yeah, Silence of the Lambs.

Lex Fridman (1:20:30.200)

But there were those TV shows as well,

Lex Fridman (1:20:31.760)

and they focused on this guy, Will Durant,

Lex Fridman (1:20:34.040)

who's a character like extreme empath, right?

Lex Fridman (1:20:36.240)

So in the way he like catches all these killers,

Lex Fridman (1:20:38.120)

as he pretty much, he can empathize with them, right?

Lex Fridman (1:20:42.520)

Like he can understand why they're doing

Lex Fridman (1:20:44.080)

the things they're doing, right?

Lex Fridman (1:20:45.560)

It's a pretty excruciating thing, right?

Dan Kokotov (1:20:48.160)

Like, because you're pretty much like spending

Lex Fridman (1:20:49.760)

half your time in the head of evil people, right?

Dan Kokotov (1:20:52.360)

Like, but.

Lex Fridman (1:20:54.200)

I mean, I definitely try to do that with others.

Lex Fridman (1:20:57.040)

So you should do that in moderation,

Lex Fridman (1:20:59.080)

but I think it's a pretty safe place, safe place to be.

Dan Kokotov (1:21:04.760)

One of the cool things with this podcast,

Lex Fridman (1:21:06.480)

and I know you didn't sign up to hear me

Dan Kokotov (1:21:08.760)

listen to this bullshit, but.

Lex Fridman (1:21:10.160)

That was interesting.

Lex Fridman (1:21:11.920)

I, and what's his name?

Lex Fridman (1:21:15.480)

Chris Latner, who's a Google,

Dan Kokotov (1:21:17.640)

oh, he's not Google anymore, SciFi.

Lex Fridman (1:21:19.120)

He's legit, he's one of the most legit engineers

Dan Kokotov (1:21:21.040)

I talk with, I talk with him again on this podcast.

Lex Fridman (1:21:23.360)

And one of the, he gives me private advice a lot.

Lex Fridman (1:21:27.200)

And he said for this podcast, I should like interview,

Lex Fridman (1:21:31.240)

like I should widen the range of people

Dan Kokotov (1:21:34.640)

because that gives you much more freedom to do stuff.

Lex Fridman (1:21:38.200)

Like, so his idea, which I think I agree with Chris

Dan Kokotov (1:21:41.560)

is that you go to the extremes.

Lex Fridman (1:21:44.040)

You just like cover every extreme base

Lex Fridman (1:21:46.080)

and then it gives you freedom to then go

Lex Fridman (1:21:48.200)

to the more nuanced conversations.

Lex Fridman (1:21:50.440)

And it's kind of, I think there's a safe place for that.

Lex Fridman (1:21:53.920)

There's certainly a hunger for that nuanced conversation,

Dan Kokotov (1:21:56.640)

I think, amongst people where like on social media,

Lex Fridman (1:22:00.400)

you get canceled for anything slightly tense,

Dan Kokotov (1:22:04.040)

that there's a hunger to go full.

Lex Fridman (1:22:06.040)

Right, you go so far to the opposite side.

Lex Fridman (1:22:08.360)

And that's like demystifies it a little bit, right?

Lex Fridman (1:22:10.720)

Yeah, that's.

Dan Kokotov (1:22:11.560)

There is a person behind all of these things.

Lex Fridman (1:22:15.080)

And that's the cool thing about podcasting,

Dan Kokotov (1:22:17.280)

like three, four hour conversations

Lex Fridman (1:22:19.280)

that it's very different than a clickbait journalism,

Dan Kokotov (1:22:24.120)

it's like the opposite, that there's a hunger for that.

Lex Fridman (1:22:26.680)

There's a willingness for that.

Dan Kokotov (1:22:28.040)

Yeah, especially now, I mean,

Lex Fridman (1:22:29.440)

how many people do you even see face to face anymore?

Lex Fridman (1:22:31.840)

Right, like this, you know?

Lex Fridman (1:22:33.320)

It's like not that many people like in my day today,

Dan Kokotov (1:22:36.080)

aside from my own family that like I sit across.

Lex Fridman (1:22:39.240)

It's sad, but it's also beautiful.

Dan Kokotov (1:22:41.520)

Like I've gotten the chance to like,

Lex Fridman (1:22:44.200)

like our conversation now, there's somebody,

Dan Kokotov (1:22:47.080)

I guarantee you there's somebody in Russia

Lex Fridman (1:22:50.000)

listening to this now, like jogging.

Dan Kokotov (1:22:52.240)

There's somebody who is just smoked some weed,

Lex Fridman (1:22:55.280)

sit back on a couch and just like enjoying.

Dan Kokotov (1:22:58.440)

I guarantee you that we'll write in the comments right now

Lex Fridman (1:23:00.720)

that yes, I'm in St. Petersburg, I'm in Moscow, whatever.

Lex Fridman (1:23:05.000)

And we're in their head and they have a friendship with us.

Lex Fridman (1:23:10.960)

I'm the same way, I'm a huge fan of podcasting.

Dan Kokotov (1:23:14.520)

It's a beautiful thing.

Lex Fridman (1:23:15.560)

I mean, it's a weird one way human connection.

Dan Kokotov (1:23:18.160)

Like before I went on Joe Rogan and still,

Lex Fridman (1:23:22.200)

I'm just a huge fan of his.

Lex Fridman (1:23:24.120)

So it was like surreal.

Lex Fridman (1:23:25.760)

I've been friends with Joe Rogan for 10 years, but one way.

Dan Kokotov (1:23:28.960)

Yeah, from this way, from the St. Petersburg way.

Lex Fridman (1:23:31.760)

Yeah, the St. Petersburg way and it's a real friendship.

Dan Kokotov (1:23:34.960)

I mean, now it's like two way, but it's still surreal.

Lex Fridman (1:23:38.760)

And that's the magic of podcasting.

Dan Kokotov (1:23:40.480)

I'm not sure what to make of it.

Lex Fridman (1:23:42.040)

That voice, it's not even the video part.

Dan Kokotov (1:23:45.320)

It's the audio that's magical.

Lex Fridman (1:23:48.560)

I don't know what to do with it,

Lex Fridman (1:23:50.200)

but it's people listen to three, four hours.

Lex Fridman (1:23:53.080)

Yeah, we evolved over millions of years, right?

Lex Fridman (1:23:57.440)

To be very fine tuned to things like that, right?

Lex Fridman (1:24:00.480)

Oh, expressions as well, of course, right?

Lex Fridman (1:24:02.520)

But back in the day on the Savannah,

Lex Fridman (1:24:06.960)

you had to be very attuned to whether

Dan Kokotov (1:24:09.360)

you had a good relationship with the rest of your tribe

Lex Fridman (1:24:11.840)

or a very bad relationship, right?

Dan Kokotov (1:24:13.440)

Because if you had a very bad relationship,

Lex Fridman (1:24:15.160)

you were probably gonna be left behind

Lex Fridman (1:24:17.400)

and eaten by the lions.

Lex Fridman (1:24:18.800)

Yeah, but it's weird that the tribe is different now.

Dan Kokotov (1:24:22.600)

Like you could have a one way connection with Joe Rogan

Lex Fridman (1:24:26.040)

as opposed to the tribe of your physical vicinity.

Lex Fridman (1:24:30.520)

But that's why it works with the podcasting,

Lex Fridman (1:24:33.280)

but it's the opposite of what happens on Twitter, right?

Lex Fridman (1:24:35.960)

Because all those nuances are removed, right?

Lex Fridman (1:24:38.080)

You're not connecting with the person

Dan Kokotov (1:24:40.760)

because you don't hear the voice.

Lex Fridman (1:24:42.720)

You're connecting with like an abstraction, right?

Lex Fridman (1:24:44.400)

It's like some stream of tweets, right?

Lex Fridman (1:24:48.400)

And it's very easy to assign to them

Dan Kokotov (1:24:52.560)

any kind of evil intent or dehumanize them,

Lex Fridman (1:24:56.720)

which it's much harder to do when it's a real voice, right?

Dan Kokotov (1:24:59.120)

Because you realize it's a real person behind the voice.

Lex Fridman (1:25:02.720)

Let me try this out on you.

Dan Kokotov (1:25:05.000)

I sometimes ask about the meaning of life.

Lex Fridman (1:25:07.160)

Do you, your father now, an engineer,

Dan Kokotov (1:25:12.320)

you're building up a company.

Lex Fridman (1:25:14.040)

Do you ever zoom out and think like,

Lex Fridman (1:25:16.840)

what the hell is this whole thing for?

Lex Fridman (1:25:19.360)

Like why are we descended to vapes even on this planet?

Lex Fridman (1:25:24.000)

What's the meaning of it all?

Lex Fridman (1:25:26.120)

That's a pretty big question.

Dan Kokotov (1:25:29.320)

I think I don't allow myself to think about it too often,

Lex Fridman (1:25:32.240)

or maybe like life doesn't allow me

Dan Kokotov (1:25:34.320)

to think about it too often.

Lex Fridman (1:25:36.720)

But in some ways, I guess the meaning of life

Dan Kokotov (1:25:39.080)

is kind of contributing to this kind of weird thing

Lex Fridman (1:25:44.040)

we call humanity, right?

Dan Kokotov (1:25:45.320)

Like it's in a way, you can think of humanity

Lex Fridman (1:25:47.640)

as like a living and evolving organism, right?

Dan Kokotov (1:25:50.200)

That like we all contributing in a sway way,

Lex Fridman (1:25:52.560)

but just by existing, by having our own unique set

Lex Fridman (1:25:55.640)

of desires and drives, right?

Lex Fridman (1:25:58.560)

And maybe that means like creating something great.

Lex Fridman (1:26:01.640)

And it's bringing up kids who are unique and different

Lex Fridman (1:26:06.200)

and seeing like, they can join what they do.

Lex Fridman (1:26:09.680)

But I mean, to me, that's pretty much it.

Lex Fridman (1:26:11.040)

I mean, if you're not a religious person, right?

Dan Kokotov (1:26:13.160)

Which I guess I'm not, that's the meaning of life.

Lex Fridman (1:26:16.400)

It's in the living and in the creation.

Dan Kokotov (1:26:20.880)

Yeah, there's something magical

Lex Fridman (1:26:22.400)

about that engine of creation.

Dan Kokotov (1:26:24.160)

Like you said, programming, I would say,

Lex Fridman (1:26:27.240)

I mean, it's even just actually what you said

Dan Kokotov (1:26:29.400)

with even just programs.

Lex Fridman (1:26:30.520)

I don't care if it's like some JavaScript thing

Dan Kokotov (1:26:32.760)

on a button on the website.

Lex Fridman (1:26:36.000)

It's like magical that you brought that to life.

Dan Kokotov (1:26:38.960)

I don't know what that is in there, but that seems,

Lex Fridman (1:26:41.520)

that's probably some version of like reproduction

Lex Fridman (1:26:47.200)

and sex, whatever that's in evolution.

Lex Fridman (1:26:49.760)

But like creating that HTML button has echoes

Dan Kokotov (1:26:55.600)

of that feeling and it's magical.

Lex Fridman (1:26:58.880)

Right, well, I mean, if you're a religious person,

Dan Kokotov (1:27:00.640)

maybe you could even say, all right,

Lex Fridman (1:27:01.720)

like we were created in God's image, right?

Dan Kokotov (1:27:04.360)

Well, I mean, I guess part of that is the drive

Lex Fridman (1:27:07.240)

to create something ourselves, right?

Dan Kokotov (1:27:09.200)

I mean, that's part of it.

Lex Fridman (1:27:11.760)

Yeah, that HTML button is the creation in God's image.

Dan Kokotov (1:27:15.800)

Maybe hopefully it'll be something a little more.

Lex Fridman (1:27:18.880)

So dynamic, maybe some JavaScript.

Dan Kokotov (1:27:20.960)

Yeah, maybe some JavaScript, some React and so on.

Lex Fridman (1:27:25.400)

But no, I mean, I think that's what differentiates us

Dan Kokotov (1:27:29.400)

from the apes, so to speak.

Lex Fridman (1:27:32.160)

Yeah, we did a pretty good job.

Dan Kokotov (1:27:34.240)

Dan, it was an honor to talk to you.

Lex Fridman (1:27:36.960)

Thank you so much for being part of creating

Dan Kokotov (1:27:38.760)

one of my favorite services and products.

Lex Fridman (1:27:42.000)

This is actually a little bit of an experiment.

Dan Kokotov (1:27:45.040)

Allow me to sort of fanboy over some of the things I love.

Lex Fridman (1:27:49.840)

So thanks for wasting your time with me today.

Dan Kokotov (1:27:52.280)

It was really fun.

Lex Fridman (1:27:53.120)

Well, it was awesome.

Dan Kokotov (1:27:53.960)

Thanks for having me on and giving me a chance

Lex Fridman (1:27:55.520)

to try this out.

Dan Kokotov (1:27:57.200)

Awesome.

Lex Fridman (1:27:59.360)

Thanks for listening to this conversation

Dan Kokotov (1:28:00.720)

with Dan Kokotov and thank you to our sponsors,

Lex Fridman (1:28:03.560)

Athletic Greens, Only One Nutrition Drink,

Dan Kokotov (1:28:06.200)

Blinkist app that summarizes books,

Lex Fridman (1:28:09.080)

Business Wars podcast and Cash App.

Lex Fridman (1:28:13.040)

So the choice is health, wisdom or money.

Lex Fridman (1:28:16.800)

Choose wisely, my friends.

Lex Fridman (1:28:18.320)

And if you wish, click the sponsor links below

Lex Fridman (1:28:21.920)

to get a discount and to support this podcast.

Lex Fridman (1:28:25.280)

And now let me leave you with some words

Lex Fridman (1:28:27.320)

from Ludwig Wittgenstein.

Dan Kokotov (1:28:29.840)

The limits of my language means the limits of my world.

Lex Fridman (1:28:33.720)

Thank you for listening and hope to see you next time.

Dan Kokotov (20:03.400)

You know, it takes longer to get your work done.

Lex Fridman (20:06.120)

Um, things like that, you know, if there's too many done, the Revers have

Dan Kokotov (20:09.800)

a bad experience because they might log on to see like what work is available

Lex Fridman (20:12.880)

and there's not very much work, right?

Dan Kokotov (20:15.040)

Uh, so kind of keeping that balance, um, is, is, is a quite challenging problem.

Lex Fridman (20:20.000)

And, you know, that's, that's like a problem we've been working on for many

Lex Fridman (20:22.920)

years and we're still like refining our methods, right?

Lex Fridman (20:25.360)

If you can kind of talk to this gig economy idea, I did a bunch of different

Dan Kokotov (20:30.240)

psychology experiments on mechanical Turk, for example, I've asked to do

Lex Fridman (20:33.680)

different kinds of very tricky computer vision annotation on mechanical Turk.

Lex Fridman (20:37.960)

And it's connecting, connecting people in a more systematized way.

Dan Kokotov (20:43.560)

I would say, you know, between task and, and, uh, what would you call that worker

Dan Kokotov (20:50.760)

is what mechanical Turk calls it.

Lex Fridman (20:52.200)

What do you think about this world of gig economies, of there being a service

Dan Kokotov (20:58.400)

that connects customers to workers in a way that's like massively distributed,

Lex Fridman (21:06.320)

like potentially scaling to, it could be, it could be scaled to like

Lex Fridman (21:10.080)

tens of thousands of people, right?

Lex Fridman (21:11.800)

Is there something interesting about that world that you can speak to?

Dan Kokotov (21:16.520)

Yeah.

Lex Fridman (21:16.880)

Well, we don't think of it as kind of gig economy, like to some degree,

Lex Fridman (21:20.640)

I don't like the word gig that much, right?

Lex Fridman (21:22.440)

Because to some degree it diminishes the words being done, right?

Dan Kokotov (21:25.960)

It sounds kind of like almost amateurish.

Dan Kokotov (21:28.080)

Well, maybe in like music industry, like gig is a standard term, but in work, it

Dan Kokotov (21:33.760)

kind of sounds like, oh, it's, it's, it's frivolous, um, to us it's, um, improving

Lex Fridman (21:41.400)

the nature of working from home on your own time and on your own terms, right?

Lex Fridman (21:46.160)

And kind of taking away geographical limitations and time limitations, right?

Lex Fridman (21:51.360)

Uh, so, you know, many of our freelancers are maybe work from home moms, right?

Dan Kokotov (21:56.960)

And, you know, they don't want the traditional nine to five job, but they

Dan Kokotov (22:01.160)

want to make some income and rough kind of like allows them to do that and decide

Dan Kokotov (22:05.080)

like exactly how much to work and when to work or by the same token, maybe someone

Lex Fridman (22:10.920)

is, you know, someone wants to live the mountain top, you know, life, right?

Dan Kokotov (22:17.040)

You know, cabin in the woods, but they still want to make some money.

Dan Kokotov (22:20.160)

Um, and like generally that wouldn't be compatible before, before this new world,

Dan Kokotov (22:25.160)

you kind of had to choose, uh, but like with Rev, like, if you like, you don't

Lex Fridman (22:28.760)

have to choose, can you speak to like, what's the demographics like distribution?

Lex Fridman (22:36.360)

Like where do rivers live?

Dan Kokotov (22:38.880)

Is there a way to, to, to, to, to, to, to, to, to, to, to, to, to, to, to, to, to,

Dan Kokotov (22:40.760)

to, to, to, but you really want to teach it how to, how to run, how to track

Lex Fridman (22:46.720)

Once you're out of the bush, 들어가, things like that, you know,

Lex Fridman (22:50.160)

but like in the back of you know, like hard, but you

Lex Fridman (23:02.220)

just as you approach, there's a lot more control over you.

Dan Kokotov (23:05.560)

Like you, you may be Oh, like you know, one day you might go to the

Lex Fridman (23:08.160)

For some years now, we've been doing these little meetings

Dan Kokotov (23:10.400)

where the management team will go to some place

Lex Fridman (23:12.280)

and we'll try to meet Revers.

Lex Fridman (23:13.640)

And pretty much wherever we go, it's

Lex Fridman (23:15.980)

pretty easy to find a large number of Revers.

Dan Kokotov (23:19.520)

The most recent one we did is in Utah.

Lex Fridman (23:24.000)

But anyway, really.

Lex Fridman (23:25.240)

Are they from all walks of life?

Lex Fridman (23:26.680)

Are these young folks, older folks?

Dan Kokotov (23:28.800)

Yeah, all walks of life, really.

Lex Fridman (23:30.220)

Like I said, one category is the work from home.

Dan Kokotov (23:33.760)

Students who want to make some extra income.

Lex Fridman (23:37.040)

There are some people who maybe have some social anxiety,

Lex Fridman (23:41.880)

so they don't want to be in the office.

Lex Fridman (23:43.420)

And this is one way for them to make a living.

Lex Fridman (23:45.380)

So it's really pretty wide variety.

Lex Fridman (23:47.360)

But on the flip side, for example,

Dan Kokotov (23:49.040)

one Rever we were talking to was a person

Lex Fridman (23:52.840)

who had a fairly high powered career before

Lex Fridman (23:54.600)

and was kind of like taking a break.

Lex Fridman (23:57.280)

And she was almost doing this just to explore and learn

Dan Kokotov (24:00.320)

about the gig economy, quote unquote.

Lex Fridman (24:03.400)

So it really is a pretty wide variety of folks.

Dan Kokotov (24:06.280)

Yeah, it's kind of interesting through the captioning

Lex Fridman (24:09.400)

process for me to learn about the Revers

Dan Kokotov (24:13.080)

because like some are clearly like weirdly knowledgeable

Lex Fridman (24:19.840)

about technical concepts.

Dan Kokotov (24:22.960)

Like you can tell by how good they are

Lex Fridman (24:25.280)

at like capitalizing stuff, like technical terms,

Dan Kokotov (24:29.040)

like a machine learning or deep learning.

Lex Fridman (24:30.840)

Like I've used Rev to annotate, to caption

Dan Kokotov (24:35.440)

the deep learning lectures or machine learning lectures

Lex Fridman (24:38.000)

I did at MIT.

Lex Fridman (24:39.880)

And it's funny, like a large number of them were like,

Lex Fridman (24:44.520)

I don't know if they looked it up

Dan Kokotov (24:45.880)

or were already knowledgeable,

Lex Fridman (24:47.280)

but they do a really good job at like, I don't know.

Dan Kokotov (24:50.400)

They invest time into these things.

Lex Fridman (24:52.400)

They will like do research, they will Google things,

Dan Kokotov (24:55.040)

you know, to kind of make sure that they got it right.

Lex Fridman (24:57.380)

But to some of them, it's like,

Dan Kokotov (24:59.120)

it's actually part of the enjoyment of the work.

Lex Fridman (25:01.640)

Like they'll tell us, you know,

Dan Kokotov (25:03.360)

I love doing this because I get paid

Lex Fridman (25:05.760)

to watch a documentary on something, right?

Lex Fridman (25:07.400)

And I learn something while I'm transcribing, right?

Lex Fridman (25:10.040)

Pretty cool.

Dan Kokotov (25:10.880)

Yeah, so what's that captioning transcription process

Lex Fridman (25:14.680)

look like for the Revers?

Lex Fridman (25:16.160)

Can you maybe speak to that to give people a sense,

Lex Fridman (25:18.960)

like how much is automated, how much is manual?

Lex Fridman (25:23.040)

What's the actual interface look like?

Lex Fridman (25:25.120)

All that kind of stuff.

Dan Kokotov (25:26.320)

Yeah, so, you know, we've invested a pretty good amount

Lex Fridman (25:28.920)

of time to give like our Revers the best tools possible.

Lex Fridman (25:33.200)

So typical day of forever,

Lex Fridman (25:34.760)

they might log into their workspace,

Dan Kokotov (25:37.120)

they'll see a list of audios that need to be transcribed.

Lex Fridman (25:41.440)

And we try to give them tools to pick specifically

Lex Fridman (25:43.400)

the ones they want to do, you know?

Lex Fridman (25:44.500)

So maybe some people like to do longer audios

Dan Kokotov (25:47.800)

or shorter audios, people have their preferences.

Lex Fridman (25:52.340)

Some people like to do audios in a particular subject

Dan Kokotov (25:55.080)

or from a particular country.

Lex Fridman (25:56.020)

So we try to give people the tools to control,

Dan Kokotov (25:59.720)

things like that.

Lex Fridman (26:01.100)

And then when they pick what they want to do,

Dan Kokotov (26:04.600)

we'll launch a specialized editor that we've built

Lex Fridman (26:07.520)

to make transcription as efficient as possible.

Dan Kokotov (26:10.240)

They'll start with a speech drag draft.

Lex Fridman (26:12.360)

So, you know, we have our machine learning model

Dan Kokotov (26:15.160)

for automated speech recognition, they'll start with that.

Lex Fridman (26:18.560)

And then our tools are optimized to help them correct that.

Lex Fridman (26:22.780)

So it's basically a process of correction.

Lex Fridman (26:26.000)

Yeah, it depends on, you know, I would say the audio.

Dan Kokotov (26:29.520)

If the audio itself is pretty good,

Lex Fridman (26:31.360)

like probably like our podcast right now

Dan Kokotov (26:33.160)

would be quite good.

Lex Fridman (26:34.120)

So the ASR would do a fairly good job.

Lex Fridman (26:37.840)

But if you imagine someone recorded a lecture,

Lex Fridman (26:41.160)

you know, in the back of a auditorium, right?

Dan Kokotov (26:45.720)

Where like the speaker is really far away

Lex Fridman (26:47.340)

and there's maybe a lot of cross talk and things like that,

Dan Kokotov (26:49.960)

then maybe the ASR wouldn't do a good job.

Lex Fridman (26:52.320)

So the person might say like, you know what,

Dan Kokotov (26:53.720)

I'm just gonna do it from scratch.

Lex Fridman (26:55.080)

Do it from scratch, yeah.

Lex Fridman (26:56.280)

So it kind of really depends.

Lex Fridman (26:57.640)

What would you say is the speed that you can possibly get?

Lex Fridman (27:00.560)

Like, what's the fastest?

Lex Fridman (27:02.080)

Can you get, is it possible to get real time or no?

Lex Fridman (27:05.200)

As you're like listening, can you write as fast as?

Lex Fridman (27:09.120)

Real time would be pretty difficult.

Dan Kokotov (27:10.440)

It's actually a pretty, it's not an easy job, you know.

Lex Fridman (27:13.400)

We actually encourage everyone at the company

Dan Kokotov (27:16.240)

to try to be a transcriber for a day,

Lex Fridman (27:17.680)

transcriptionist for a day.

Lex Fridman (27:20.400)

And it's way harder than you might think it is, right?

Lex Fridman (27:24.120)

Because people talk fast and people have accents

Lex Fridman (27:28.280)

and all this kind of stuff.

Lex Fridman (27:29.160)

So real time is pretty difficult.

Lex Fridman (27:30.920)

Is it possible?

Lex Fridman (27:32.600)

Like there's somebody, we're probably gonna use Rev

Dan Kokotov (27:34.800)

to caption this, they're listening to this right now.

Lex Fridman (27:39.320)

What do you think is the fastest

Lex Fridman (27:42.440)

you could possibly get on this right now?

Lex Fridman (27:44.880)

I think on a good audio, maybe two to three X,

Dan Kokotov (27:47.480)

I would say, real time.

Lex Fridman (27:49.880)

Meaning it takes two to three times longer

Dan Kokotov (27:51.680)

than the actual audio of the podcast.

Lex Fridman (27:55.560)

This is so meta, I could just imagine the Revvers

Dan Kokotov (27:58.600)

working on this right now.

Lex Fridman (27:59.440)

You're like, you're way wrong.

Dan Kokotov (28:01.040)

You're way wrong, this takes way longer.

Lex Fridman (28:03.520)

But yeah, it definitely works.

Dan Kokotov (28:04.360)

Or you doubted me, I could do real time.

Lex Fridman (28:06.440)

Yeah.

Dan Kokotov (28:08.600)

Okay, so you mentioned ASR.

Lex Fridman (28:11.200)

Can you speak to what is ASR, automatic speech recognition?

Lex Fridman (28:15.480)

How much, like what is the gap

Lex Fridman (28:19.320)

between perfect human performance

Lex Fridman (28:22.000)

and perfect or pretty damn good ASR?

Lex Fridman (28:26.680)

Yeah, so ASR, automatic speech recognition,

Lex Fridman (28:28.800)

it's a class of machine learning problem, right?

Lex Fridman (28:31.800)

So take speech like we're talking

Lex Fridman (28:34.200)

and transform it into a sequence of words, essentially.

Lex Fridman (28:37.040)

Audio of people talking.

Dan Kokotov (28:38.800)

Audio, audio to words.

Lex Fridman (28:41.680)

And there's a variety of different approaches

Lex Fridman (28:43.600)

and techniques, which we could talk about later if you want.

Lex Fridman (28:46.920)

So, we think we have pretty much the world's best ASR

Lex Fridman (28:50.400)

for this kind of speech, right?

Lex Fridman (28:52.600)

So there's different kinds of domains, right, for ASR.

Lex Fridman (28:55.600)

Like one domain might be voice assistance, right?

Lex Fridman (28:58.800)

So Siri, very different than what we're doing, right?

Dan Kokotov (29:02.720)

Because Siri, there's fairly limited vocabulary.

Lex Fridman (29:05.680)

You might ask Siri to play a song

Dan Kokotov (29:08.480)

or order a pizza or whatever.

Lex Fridman (29:10.560)

And it's very good at doing that.

Dan Kokotov (29:12.360)

Very different from when we start talking

Lex Fridman (29:14.680)

in a very unstructured way.

Lex Fridman (29:17.040)

And Siri will also generally adapt to your voice

Lex Fridman (29:18.760)

and stuff like this.

Lex Fridman (29:20.080)

So for this kind of audio, we think we have the best.

Lex Fridman (29:22.920)

And our accuracy, right now I think it's maybe 14%

Dan Kokotov (29:29.360)

word error rate on our test suite

Lex Fridman (29:32.920)

that we generally use to measure.

Lex Fridman (29:33.840)

So word error rate is like one way to measure accuracy

Lex Fridman (29:37.040)

for ASR, right?

Lex Fridman (29:37.880)

So what's 14% word error rate?

Lex Fridman (29:39.720)

So 14% means across this test suite,

Dan Kokotov (29:43.360)

of a variety of different audios,

Lex Fridman (29:45.960)

it would be, it would get in some way 14% of the words wrong.

Dan Kokotov (29:53.600)

14% of the words wrong.

Lex Fridman (29:55.400)

So the way you kind of calculate it is,

Dan Kokotov (29:59.680)

you might add up insertions, deletions, and substitutions,

Lex Fridman (30:02.960)

right?

Lex Fridman (30:03.800)

So insertions is like extra words.

Lex Fridman (30:05.960)

Deletions are words that we said,

Lex Fridman (30:07.680)

but weren't in the transcript, right?

Lex Fridman (30:10.840)

Substitutions is, you said Apple, but I said,

Lex Fridman (30:14.200)

but the ASR thought it was able, something like this.

Lex Fridman (30:17.920)

Human accuracy, most people think realistically,

Dan Kokotov (30:21.640)

it's like 3%, 2%, word error rate would be like

Lex Fridman (30:25.680)

the max achievable.

Lex Fridman (30:28.160)

So there's still quite a gap, right?

Lex Fridman (30:29.960)

Would you say that, so YouTube, when I upload videos,

Dan Kokotov (30:33.000)

often generates automatic captions.

Lex Fridman (30:35.440)

Are you sort of from a company perspective,

Dan Kokotov (30:38.120)

from a company perspective, from a tech perspective,

Lex Fridman (30:41.560)

are you trying to beat YouTube, Google?

Dan Kokotov (30:44.720)

It's a hell of a, Google, I mean,

Lex Fridman (30:47.400)

I don't know how seriously they take this task,

Lex Fridman (30:49.720)

but I imagine it's quite serious.

Lex Fridman (30:51.840)

And they, you know, Google is probably up there

Dan Kokotov (30:56.280)

in terms of their teams on, on ASR or just NLP,

Lex Fridman (31:02.080)

natural language processing, different technologies.

Lex Fridman (31:04.440)

So do you think you can beat Google?

Lex Fridman (31:06.680)

On this kind of stuff, yeah, we think so.

Dan Kokotov (31:08.960)

Google just woke up on my phone.

Lex Fridman (31:11.560)

This is hilarious, okay.

Dan Kokotov (31:12.960)

Now Google is listening, sending it back to headquarters.

Lex Fridman (31:17.440)

Who are these rough people?

Lex Fridman (31:19.520)

But that's the goal?

Lex Fridman (31:20.560)

Yeah, I mean, we measure ourselves against like Google,

Dan Kokotov (31:23.120)

Amazon, Microsoft, you know, some of the,

Lex Fridman (31:25.560)

some smaller competitors.

Lex Fridman (31:28.280)

And we use like our internal tests with it,

Lex Fridman (31:30.320)

we try to compose it of a pretty representative

Dan Kokotov (31:32.760)

set of ideas, maybe it's some podcasts, some videos,

Lex Fridman (31:36.360)

some interviews, some lectures, things like that, right?

Lex Fridman (31:39.720)

And we beat them in our own testing.

Lex Fridman (31:42.800)

And actually Rev offers automated,

Dan Kokotov (31:45.920)

like you can actually just do the automated captioning.

Lex Fridman (31:49.280)

So like, I guess it's like way cheaper, whatever it is,

Dan Kokotov (31:52.680)

whatever the rates are.

Lex Fridman (31:54.160)

Yeah, yeah.

Lex Fridman (31:55.000)

So it's a, by the way, it used to be a dollar per minute

Lex Fridman (31:57.880)

for captioning and transcription,

Dan Kokotov (32:00.080)

I think it's like $1.15 or something like that.

Lex Fridman (32:02.320)

$1.25.

Dan Kokotov (32:03.160)

$1.25, yeah.

Lex Fridman (32:08.320)

It's pretty cool.

Dan Kokotov (32:09.240)

That was the other thing that was surprising to me,

Lex Fridman (32:10.920)

it was actually like the cheapest thing you could,

Dan Kokotov (32:15.840)

one of the, I mean, I don't remember it being cheaper.

Lex Fridman (32:18.440)

You could on Upwork get cheaper,

Lex Fridman (32:20.960)

but it was clear to me that this,

Lex Fridman (32:22.520)

that's gonna be really shitty.

Dan Kokotov (32:23.960)

Yeah.

Lex Fridman (32:24.800)

So like, you're also competing on price.

Dan Kokotov (32:26.920)

I think there were services that you can get,

Lex Fridman (32:29.960)

like similar to Rev kind of feel to it,

Lex Fridman (32:34.200)

but it wasn't as automated.

Lex Fridman (32:35.880)

Like the drag and drop, the entirety of the interface,

Dan Kokotov (32:37.920)

it's like the thing we're talking about.

Lex Fridman (32:39.600)

I'm such a huge fan of like frictionless,

Dan Kokotov (32:41.760)

like Amazon's single buy button, whatever.

Lex Fridman (32:47.720)

Yeah, yeah.

Dan Kokotov (32:48.560)

That one click.

Lex Fridman (32:49.400)

The one click, that's genius right there.

Dan Kokotov (32:52.360)

Like that is so important for services.

Lex Fridman (32:54.960)

Yeah.

Lex Fridman (32:55.800)

And simplicity and I mean, Rev is almost there.

Lex Fridman (33:00.480)

I mean, there's like some, I'm trying to think.

Lex Fridman (33:04.400)

So I think I've, I stopped using this pipeline,

Lex Fridman (33:10.520)

but Rev offers it and I like it,

Lex Fridman (33:12.480)

but it was causing me some issues on my side,

Lex Fridman (33:16.240)

which is you can connect it to like Dropbox

Lex Fridman (33:20.320)

and it generates the files in Dropbox.

Lex Fridman (33:22.760)

So like it closes the loop to where

Dan Kokotov (33:25.840)

I don't have to go to Rev at all and I can download it.

Lex Fridman (33:30.280)

Sorry, I don't have to go to Rev at all

Lex Fridman (33:32.720)

and to download the files.

Lex Fridman (33:34.200)

It could just like automatically copy them.

Dan Kokotov (33:36.280)

Right, you're putting your Dropbox in a day later

Lex Fridman (33:39.720)

or maybe a few hours later.

Dan Kokotov (33:41.080)

Yeah, it just shows up.

Lex Fridman (33:42.480)

Just shows up, yeah.

Dan Kokotov (33:43.600)

Yeah, I was trying to do it programmatically too.

Lex Fridman (33:46.520)

Is there an API interface you can,

Dan Kokotov (33:48.920)

I was trying to through like through Python

Lex Fridman (33:51.520)

to download stuff automatically,

Lex Fridman (33:53.440)

but then I realized this is the programmer in me.

Lex Fridman (33:56.160)

Like, dude, you don't need to automate everything

Dan Kokotov (33:58.720)

like in life, like flawlessly,

Lex Fridman (34:01.160)

because I wasn't doing enough captions

Dan Kokotov (34:02.760)

to justify to myself the time investment

Lex Fridman (34:05.640)

into automating everything perfectly.

Dan Kokotov (34:07.800)

Yeah, I would say if you're doing so many interviews

Lex Fridman (34:10.040)

that your biggest roadblock is clicking on the Rev download,

Lex Fridman (34:14.240)

but now you're talking about Elon Musk levels of business.

Lex Fridman (34:18.960)

But for sure, we have like a variety of ways

Dan Kokotov (34:22.040)

to make it easy.

Lex Fridman (34:22.880)

You know, there's the integration.

Dan Kokotov (34:24.200)

You mentioned, I think it's through a company called Zapier,

Lex Fridman (34:26.240)

which kind of can connect Dropbox to Rev and vice versa.

Dan Kokotov (34:31.120)

We have an API if you want to really like customize it,

Lex Fridman (34:33.480)

you know, if you want to create the Lex Friedman,

Dan Kokotov (34:37.160)

you know, CMS or whatever.

Lex Fridman (34:40.920)

For this whole thing.

Dan Kokotov (34:41.760)

Okay, cool.

Lex Fridman (34:42.600)

So can you speak to the ASR a little bit more?

Lex Fridman (34:46.480)

Like, what does it take?

Lex Fridman (34:51.120)

Like approach wise, machine learning wise,

Lex Fridman (34:53.400)

how hard is this problem?

Lex Fridman (34:55.000)

How do you get to the 3% error rate?

Lex Fridman (34:57.720)

Like, what's your vision of all of this?

Lex Fridman (34:59.400)

Yeah, well, the 3% error rate is definitely,

Dan Kokotov (35:03.200)

that's the grand vision.

Lex Fridman (35:06.160)

We'll see what it takes to get there.

Lex Fridman (35:09.840)

But we believe, you know, in ASR,

Lex Fridman (35:13.040)

the biggest thing is the data, right?

Dan Kokotov (35:15.200)

Like, that's true of like a lot of

Lex Fridman (35:16.520)

machine learning problems today, right?

Dan Kokotov (35:18.320)

The more data you have and high quality of the data,

Lex Fridman (35:21.000)

the better label the data.

Dan Kokotov (35:24.840)

Yeah, that's how you get good results.

Lex Fridman (35:26.480)

And we at Rev have kind of like the best data.

Dan Kokotov (35:28.760)

Like we have.

Lex Fridman (35:29.760)

Like you're literally,

Dan Kokotov (35:30.960)

your business model is annotating the data.

Lex Fridman (35:34.000)

Our business model is being paid to annotate the data.

Dan Kokotov (35:36.720)

Being paid to annotate the data.

Lex Fridman (35:39.080)

So it's kind of like a pretty magical flywheel.

Dan Kokotov (35:42.000)

Yeah.

Lex Fridman (35:42.840)

And so we've kind of like written this flywheel

Dan Kokotov (35:44.480)

to this point.

Lex Fridman (35:47.000)

And we think we're still kind of in the early stages

Dan Kokotov (35:50.480)

of figuring out all the parts of the flywheel to use,

Lex Fridman (35:53.080)

you know, because we have the final transcripts

Lex Fridman (35:57.560)

and we have the audios and we train on that.

Lex Fridman (36:01.640)

But we in principle also have all the edits

Lex Fridman (36:05.000)

that the Revvers make, right?

Lex Fridman (36:07.760)

Oh, that's interesting.

Lex Fridman (36:08.600)

How can you use that as data?

Lex Fridman (36:10.520)

Yeah, that's something for us to figure out in the future.

Dan Kokotov (36:13.280)

But, you know, we feel like we're only

Lex Fridman (36:15.080)

in the early stages, right?

Lex Fridman (36:16.200)

So the data is there.

Lex Fridman (36:17.880)

That'd be interesting.

Dan Kokotov (36:18.720)

Like almost like a recurrent neural net

Lex Fridman (36:20.720)

for fixing transcripts.

Dan Kokotov (36:23.960)

I always remember we did a segmentation annotation

Lex Fridman (36:28.680)

for driving data.

Lex Fridman (36:30.200)

So segmenting the scene, like visual data.

Lex Fridman (36:33.160)

And you can get all,

Lex Fridman (36:35.000)

so it was drawing, people were drawing polygons

Lex Fridman (36:37.400)

around different objects and so on.

Lex Fridman (36:39.960)

And it feels like it always felt like

Lex Fridman (36:41.720)

there was a lot of information in the clicking,

Dan Kokotov (36:45.160)

the sequence of clicking that people do,

Lex Fridman (36:46.960)

the kind of fixing of the polygons that they do.

Dan Kokotov (36:50.280)

Now there's a few papers written about

Lex Fridman (36:52.840)

how to draw polygons like with a recurrent neural nets

Dan Kokotov (36:57.720)

to try to learn from the human clicking.

Lex Fridman (37:00.960)

But it was just like experimental,

Dan Kokotov (37:04.320)

you know, it was one of those like CVPR type papers

Lex Fridman (37:06.600)

that people do like a really tiny data set.

Dan Kokotov (37:08.920)

It didn't feel like people really tried to do it seriously.

Lex Fridman (37:13.040)

Yeah, I wonder if there's information in the fixing

Dan Kokotov (37:16.440)

that provides deeper set of signal

Lex Fridman (37:21.520)

than just like the raw data.

Lex Fridman (37:24.400)

The intuition is for sure there must be, right?

Lex Fridman (37:26.120)

There must be.

Lex Fridman (37:26.960)

And in all kinds of signals and how long you took

Lex Fridman (37:29.640)

to make that edit and stuff like that.

Dan Kokotov (37:32.640)

It's gonna be like up to us.

Lex Fridman (37:34.080)

That's why like the next couple of years

Lex Fridman (37:36.760)

is like super exciting for us, right?

Lex Fridman (37:38.280)

So that's what like the focus is now.

Dan Kokotov (37:40.320)

You mentioned rev.ai, that's where you want to.

Lex Fridman (37:43.280)

Yeah, so rev.ai is kind of our way of bringing this ASR

Lex Fridman (37:49.120)

to the rest of the world, right?

Lex Fridman (37:51.520)

So when we started, we were human only.

Dan Kokotov (37:55.560)

Then we kind of created this Temi service.

Lex Fridman (37:59.160)

I think you might've used it,

Lex Fridman (38:00.600)

which was kind of ASR for the consumer, right?

Lex Fridman (38:02.520)

So if you don't wanna pay $1.25, but you wanna pay,

Dan Kokotov (38:06.520)

now it's 25 cents a minute, I think.

Lex Fridman (38:08.040)

And you get the transcript,

Dan Kokotov (38:10.680)

the machine generated transcript and you get an editor

Lex Fridman (38:13.800)

and you can kind of fix it up yourself, right?

Dan Kokotov (38:17.440)

Then we started using ASR

Lex Fridman (38:18.800)

for our own human transcriptionists.

Lex Fridman (38:21.960)

And then the kind of rev.ai is the final step

Lex Fridman (38:23.400)

of the journey, which is, you know,

Dan Kokotov (38:25.120)

we have this amazing engine.

Lex Fridman (38:27.040)

What can people build with it, right?

Lex Fridman (38:28.840)

What kind of new applications could be enabled

Lex Fridman (38:33.600)

if you have SpeedTrack that's that accurate?

Lex Fridman (38:36.320)

Do you have ideas for this

Lex Fridman (38:37.560)

or is it just providing it as a service

Lex Fridman (38:39.240)

and seeing what people come up with?

Lex Fridman (38:40.640)

It's providing it as a service

Lex Fridman (38:41.920)

and seeing what people come up with

Lex Fridman (38:43.480)

and kind of learning from what people do with it.

Lex Fridman (38:45.560)

And we have ideas of our own as well, of course,

Lex Fridman (38:47.160)

but it's a little bit like, you know,

Lex Fridman (38:49.160)

when AWS provided the building blocks, right?

Lex Fridman (38:52.520)

And they saw what people built with it

Lex Fridman (38:53.880)

and they try to make it easier to build those things, right?

Lex Fridman (38:56.960)

And we kind of hope to do the same thing.

Dan Kokotov (38:59.120)

Although AWS kind of does a shitty job of like,

Lex Fridman (39:02.760)

I'm continually surprised, like Mechanical Turk,

Dan Kokotov (39:05.000)

for example, how shitty the interface is.

Lex Fridman (39:07.720)

We're talking about like Rev making me feel good.

Dan Kokotov (39:11.080)

Like when I first discovered Mechanical Turk,

Lex Fridman (39:15.200)

the initial idea of it was like,

Dan Kokotov (39:18.200)

it made me feel like Rev does,

Lex Fridman (39:19.600)

but then the interface is like, come on.

Dan Kokotov (39:22.760)

Yeah, it's horrible.

Lex Fridman (39:24.720)

Why is it so painful?

Lex Fridman (39:27.720)

Does nobody at Amazon want to like seriously invest in it?

Lex Fridman (39:32.480)

It felt like you can make so much money

Dan Kokotov (39:35.000)

if you took this effort seriously.

Lex Fridman (39:37.160)

And it feels like they have a committee

Dan Kokotov (39:39.240)

of like two people just sitting back,

Lex Fridman (39:41.200)

like a meeting, they meet once a month,

Lex Fridman (39:44.000)

like what are we going to do with Mechanical Turk?

Lex Fridman (39:46.520)

It's like two websites making me feel like this,

Dan Kokotov (39:49.240)

that and craiglist.org, whatever the hell it is.

Lex Fridman (39:53.600)

It feels like it's designed in the 90s.

Dan Kokotov (39:55.960)

Well, Craigslist basically hasn't been updated

Lex Fridman (39:59.120)

pretty much since the guy originally built.

Lex Fridman (39:59.960)

Do you seriously think there's a team,

Lex Fridman (40:01.840)

like how big is the team working on Mechanical Turk?

Dan Kokotov (40:04.240)

I don't know.

Lex Fridman (40:05.080)

There's some team, right?

Dan Kokotov (40:06.840)

I feel like there isn't.

Lex Fridman (40:08.360)

I'm skeptical.

Dan Kokotov (40:09.480)

Yeah.

Lex Fridman (40:10.600)

Well, if nothing else, they benefit from the other teams

Dan Kokotov (40:14.680)

like moving things forward in a small way.

Lex Fridman (40:18.400)

But I know what you mean.

Dan Kokotov (40:19.760)

We use Mechanical Turk for a couple of things as well.

Lex Fridman (40:22.280)

And yeah, it's painful UI.

Dan Kokotov (40:23.840)

It's painful, but yeah, it works.

Lex Fridman (40:25.720)

I think most people, the thing is most people

Lex Fridman (40:27.560)

don't really use the UI, right?

Lex Fridman (40:29.160)

Like we, for example, we use it through the API, right?

Lex Fridman (40:33.600)

But even the API documentation and so on,

Lex Fridman (40:36.120)

like it's super outdated.

Dan Kokotov (40:37.520)

Like I don't even know what to...

Lex Fridman (40:42.840)

I mean, the same criticism, as long as we're ranting,

Dan Kokotov (40:47.680)

my same criticism goes to the APIs

Lex Fridman (40:49.960)

of most of these companies.

Dan Kokotov (40:50.920)

Like Google, for example, the API for the different services

Lex Fridman (40:55.120)

is just the documentation is so shitty.

Dan Kokotov (40:59.760)

Like it's not so shitty.

Lex Fridman (41:02.160)

I should actually be...

Dan Kokotov (41:05.280)

I should exhibit some gratitude.

Lex Fridman (41:08.280)

Okay, let's practice some gratitude.

Dan Kokotov (41:10.800)

The documentation is pretty good.

Lex Fridman (41:14.240)

Like most of the things that the API makes available

Dan Kokotov (41:18.720)

is pretty good.

Lex Fridman (41:19.640)

It's just that in the sense that it's accurate,

Dan Kokotov (41:23.040)

sometimes outdated, but like the degree of explanations

Lex Fridman (41:27.200)

with examples is only covering, I would say,

Dan Kokotov (41:31.360)

like 50% of what's possible.

Lex Fridman (41:33.840)

And it just feels a little bit,

Dan Kokotov (41:35.520)

like there's a lot of natural questions

Lex Fridman (41:37.320)

that people would wanna ask that doesn't get covered.

Lex Fridman (41:41.600)

And it feels like it's almost there.

Lex Fridman (41:44.520)

Like it's such a magical thing.

Dan Kokotov (41:46.120)

Like the Maps API, YouTube API, there's a bunch of stuff.

Lex Fridman (41:51.120)

I gotta imagine it's like, there's probably some team

Dan Kokotov (41:54.720)

at Google responsible for writing this documentation

Lex Fridman (41:57.440)

that's probably not the engineers, right?

Lex Fridman (42:00.240)

And probably this team is not where you wanna be.

Lex Fridman (42:04.360)

Well, it's a weird thing.

Dan Kokotov (42:05.800)

I sometimes think about this for somebody

Lex Fridman (42:09.160)

who wants to also build a company.

Dan Kokotov (42:12.120)

I think about this a lot.

Lex Fridman (42:14.920)

YouTube, the service is one of the most magical,

Dan Kokotov (42:21.560)

like I'm so grateful that YouTube exists.

Lex Fridman (42:24.440)

And yet they seem to be quite clueless on so many things

Dan Kokotov (42:30.840)

like that everybody's screaming them at.

Lex Fridman (42:33.400)

Like it feels like whatever the mechanism

Dan Kokotov (42:37.360)

that you use to listen to your quote unquote customers,

Lex Fridman (42:40.080)

which is like the creators, is not very good.

Dan Kokotov (42:44.800)

Like there's literally people that are like screaming why,

Lex Fridman (42:47.880)

like their new YouTube studio, for example.

Dan Kokotov (42:51.040)

There's like features that were like begged for

Lex Fridman (42:55.160)

for a really long time.

Dan Kokotov (42:56.880)

Like being able to upload multiple videos at the same time.

Lex Fridman (43:00.160)

That wasn't missing for a really, really long time.

Dan Kokotov (43:03.880)

Now, like there's probably things that I don't know,

Lex Fridman (43:08.000)

which is maybe for that kind of huge infrastructure,

Dan Kokotov (43:10.960)

it's actually very difficult to build

Lex Fridman (43:12.320)

some of these features.

Lex Fridman (43:13.720)

But the fact that that wasn't communicated

Lex Fridman (43:15.520)

and it felt like you're not being heard.

Dan Kokotov (43:19.120)

Like I remember this experience for me

Lex Fridman (43:21.520)

and it's not a pleasant experience.

Lex Fridman (43:23.800)

And it feels like the company doesn't give a damn about you.

Lex Fridman (43:26.720)

And that's something to think about.

Dan Kokotov (43:28.160)

I'm not sure what that is.

Lex Fridman (43:29.960)

That might have to do with just like small groups

Dan Kokotov (43:32.480)

working on these small features and these specific features.

Lex Fridman (43:35.920)

And there's no overarching like dictator type of human

Dan Kokotov (43:40.280)

that says like, why the hell are we neglecting

Lex Fridman (43:42.440)

like Steve Jobs type of characters?

Dan Kokotov (43:43.920)

Like there's people that we need to speak

Lex Fridman (43:48.240)

to the people that like wanna love our product

Lex Fridman (43:50.640)

and they don't.

Lex Fridman (43:51.680)

Let's fix this shit.

Dan Kokotov (43:52.520)

Maybe at some point you just get so fixated

Lex Fridman (43:54.160)

on the numbers, right?

Lex Fridman (43:55.120)

And it's like, well, the numbers are pretty great, right?

Lex Fridman (43:57.200)

Like people are watching,

Lex Fridman (43:59.680)

doesn't seem to be a problem, right?

Lex Fridman (44:01.960)

And you're not like the person that like build this thing,

Lex Fridman (44:04.080)

right?

Lex Fridman (44:04.920)

So you really care about it.

Lex Fridman (44:05.840)

You're just there, you came in as a product manager, right?

Lex Fridman (44:09.080)

You got hired sometime later,

Dan Kokotov (44:10.680)

your mandate is like increase this number,

Lex Fridman (44:13.520)

like 10%, right?

Lex Fridman (44:15.680)

And you just.

Lex Fridman (44:16.520)

That's brilliantly put.

Dan Kokotov (44:17.520)

Like if you, this is, okay, if there's a lesson in this

Lex Fridman (44:21.320)

is don't reduce your company into a metric

Dan Kokotov (44:24.200)

of like how much, like you said,

Lex Fridman (44:28.280)

how much people watching the videos and so on

Lex Fridman (44:30.800)

and like convince yourself that everything is working

Lex Fridman (44:33.800)

just because the numbers are going up.

Dan Kokotov (44:36.120)

There's something, you have to have a vision.

Lex Fridman (44:39.120)

You have to want people to love your stuff

Dan Kokotov (44:43.400)

because love is ultimately the beginning

Lex Fridman (44:46.120)

of like a successful longterm company

Dan Kokotov (44:49.280)

is that they always should love your product.

Lex Fridman (44:51.320)

You have to be like a creator

Lex Fridman (44:52.640)

and have that like creator's love for your own thing, right?

Lex Fridman (44:55.400)

Like, and you're pained by these comments, right?

Lex Fridman (44:59.560)

And probably like Apple, I think did this generally

Lex Fridman (45:02.480)

like really well.

Dan Kokotov (45:03.840)

They're well known for kind of keeping teams small

Lex Fridman (45:06.880)

even when they were big, right?

Dan Kokotov (45:08.200)

And, you know, he was an engineer,

Lex Fridman (45:10.400)

like there's a book, a creative selection.

Dan Kokotov (45:12.680)

I don't know if you read it by a Apple engineer

Lex Fridman (45:15.440)

named Ken Koscienda.

Dan Kokotov (45:17.280)

It's kind of a great book actually

Lex Fridman (45:18.320)

because unlike most of these business books where it's,

Dan Kokotov (45:21.440)

you know, here's how Steve Jobs ran the company.

Lex Fridman (45:24.600)

It's more like here's how life was like for me, you know,

Dan Kokotov (45:27.240)

an engineer here, the projects I worked on

Lex Fridman (45:29.000)

and here what it was like to pitch Steve Jobs, you know,

Dan Kokotov (45:31.720)

on like, you know, I think it was in charge of like

Lex Fridman (45:34.640)

the keyboard and the auto correction, right?

Lex Fridman (45:36.840)

And at Apple, like Steve Jobs reviewed everything.

Lex Fridman (45:39.440)

And so he was like, this is what it was like

Dan Kokotov (45:41.160)

to show my demos to Steve Jobs and, you know,

Lex Fridman (45:43.720)

to change them because like Steve Jobs didn't like how,

Dan Kokotov (45:46.600)

you know, the shape of the little key was off

Lex Fridman (45:48.760)

because the rounding of the corner was like not quite right

Lex Fridman (45:50.920)

or something like this, right?

Lex Fridman (45:51.760)

He was famously a stickler for this kind of stuff.

Lex Fridman (45:54.640)

But because the teams were small,

Lex Fridman (45:55.800)

he really owned this stuff, right?

Lex Fridman (45:56.920)

So he really cared.

Lex Fridman (45:58.680)

Yeah, Elon Musk does that similar kind of thing with Tesla,

Dan Kokotov (46:01.640)

which is really interesting.

Lex Fridman (46:03.400)

There's another lesson in leadership in that

Dan Kokotov (46:05.880)

is to be obsessed with the details.

Lex Fridman (46:07.640)

And like, he talks to like the lowest level engineers.

Dan Kokotov (46:11.280)

Okay, so we're talking about ASR

Lex Fridman (46:14.600)

and so this is basically where I was saying

Dan Kokotov (46:17.640)

we're gonna take this like ultra seriously.

Lex Fridman (46:20.360)

And then what's the mission?

Dan Kokotov (46:22.640)

To try to keep pushing towards the 3%.

Lex Fridman (46:26.200)

Yeah, and kind of try to build this platform

Dan Kokotov (46:30.320)

where all of your, you know, all of your meetings,

Lex Fridman (46:33.920)

you know, they're as easily accessible as your notes, right?

Dan Kokotov (46:38.440)

Like, so, like, imagine all the meetings

Lex Fridman (46:41.360)

a company might have, right?

Lex Fridman (46:43.080)

You know, now that I'm like no longer a programmer, right?

Lex Fridman (46:46.280)

Then I'm a quote unquote manager.

Lex Fridman (46:49.160)

That's less like my day as in meetings, right?

Lex Fridman (46:51.120)

Yeah.

Dan Kokotov (46:51.960)

And, you know, pretty often I wanna like see

Lex Fridman (46:53.800)

what was said, right?

Lex Fridman (46:55.040)

Who said it, you know?

Lex Fridman (46:56.200)

What's the context?

Lex Fridman (46:57.040)

But it's generally not really something

Lex Fridman (46:59.360)

that you can easily retrieve, right?

Dan Kokotov (47:00.480)

Like imagine if all of those meetings

Lex Fridman (47:03.200)

were indexed, archived, you know, you could go back,

Lex Fridman (47:05.760)

you could share a clip like really easily, right?

Lex Fridman (47:08.280)

So that might change completely.

Dan Kokotov (47:10.040)

It's like everything that's said,

Lex Fridman (47:11.920)

converted to text might change completely

Dan Kokotov (47:14.000)

the dynamics of what we do in this world,

Lex Fridman (47:16.320)

especially now with remote work, right?

Dan Kokotov (47:18.160)

Exactly, exactly.

Lex Fridman (47:19.480)

With Zoom and so on.

Dan Kokotov (47:21.440)

That's fascinating to think about.

Lex Fridman (47:22.760)

I mean, for me, I care about podcasts, right?

Lex Fridman (47:25.560)

And one of the things that was,

Lex Fridman (47:30.400)

you know, I'm torn.

Dan Kokotov (47:31.240)

I know a lot of the engineers at Spotify.

Lex Fridman (47:33.600)

So I love them very much because they dream big

Dan Kokotov (47:39.840)

in terms of like, they wanna empower creators.

Lex Fridman (47:43.400)

So one of my hopes was with Spotify

Dan Kokotov (47:45.080)

that they would use a technology like Rev

Lex Fridman (47:46.840)

or something like that to start converting everything

Dan Kokotov (47:51.720)

into text and make it indexable.

Lex Fridman (47:55.200)

Like one of the things that sucks with podcasts

Dan Kokotov (47:59.560)

is like, it's hard to find stuff.

Lex Fridman (48:01.800)

Like the model is basically subscription.

Dan Kokotov (48:04.480)

Like you find, it's similar to what YouTube used to be like,

Lex Fridman (48:10.600)

which is you basically find a creator that you enjoy

Lex Fridman (48:14.240)

and you subscribe to them.

Lex Fridman (48:15.360)

And like, you just kind of follow what they're doing,

Lex Fridman (48:19.720)

but the search and discovery wasn't a big part of YouTube

Lex Fridman (48:24.320)

like in the early days,

Lex Fridman (48:25.560)

but that's what currently with podcasts,

Lex Fridman (48:28.560)

like is the search and discovery is like non existent.

Dan Kokotov (48:33.880)

You're basically searching for like

Lex Fridman (48:35.280)

the dumbest possible thing,

Dan Kokotov (48:36.400)

which is like keywords in the titles of episodes.

Lex Fridman (48:39.680)

Yeah.

Dan Kokotov (48:40.520)

Even aside from a search and discovery, like all the time.

Lex Fridman (48:42.160)

So I listened to like a number of podcasts

Lex Fridman (48:44.160)

and there's something said,

Lex Fridman (48:46.840)

and I wanna like go back to that later

Dan Kokotov (48:48.560)

because I was trying to, I'm trying to remember,

Lex Fridman (48:49.800)

what do you say?

Dan Kokotov (48:50.640)

Like maybe like recommended some cool product

Lex Fridman (48:52.160)

that I wanna try out.

Lex Fridman (48:53.440)

And like, it's basically impossible.

Lex Fridman (48:54.680)

Maybe like some people have pretty good show notes.

Lex Fridman (48:56.800)

So maybe you'll get lucky and you can find it, right?

Lex Fridman (48:59.040)

But I mean, if everyone had transcripts

Lex Fridman (49:01.600)

and it was all searchable, it would be so much better.

Lex Fridman (49:05.320)

I mean, that's one of the things that I wanted to,

Dan Kokotov (49:08.480)

I mean, one of the reasons we're talking today

Lex Fridman (49:11.040)

is I wanted to take this quite seriously.

Dan Kokotov (49:13.400)

The rough thing, I just been lazy.

Lex Fridman (49:16.840)

So, because I'm very fortunate

Dan Kokotov (49:19.480)

that a lot of people support this podcast,

Lex Fridman (49:21.240)

that there's enough money now to do a transcription and so on.

Lex Fridman (49:24.840)

And it seemed clear to me, especially like CEOs

Lex Fridman (49:29.960)

and sort of like PhDs, like people write to me

Dan Kokotov (49:36.360)

who are like graduate students in computer science

Lex Fridman (49:38.320)

or graduate students in whatever the heck field,

Dan Kokotov (49:41.100)

it's clear that their mind,

Lex Fridman (49:43.120)

like they enjoy podcasts

Dan Kokotov (49:44.320)

when they're doing laundry or whatever,

Lex Fridman (49:46.120)

but they wanna revisit the conversation

Dan Kokotov (49:48.760)

in a much more rigorous way.

Lex Fridman (49:50.720)

And they really wanna transcript.

Dan Kokotov (49:52.920)

Like it's clear that they want to analyze conversations.

Lex Fridman (49:56.120)

Like so many people wrote to me

Dan Kokotov (49:58.400)

about a transcript for Yosha Bach conversation.

Lex Fridman (50:01.080)

I had just a bunch of conversations.

Lex Fridman (50:03.720)

And then on the Elon Musk side,

Lex Fridman (50:05.840)

like reporters, they wanna write a blog post

Dan Kokotov (50:09.400)

about your conversation.

Lex Fridman (50:10.840)

So they wanna be able to pull stuff.

Lex Fridman (50:13.040)

And it's like, they're essentially doing

Lex Fridman (50:15.480)

on your conversation transcription privately.

Dan Kokotov (50:18.360)

They're doing it for themselves and then starting to pick,

Lex Fridman (50:21.800)

but it's so much easier when you can actually do it

Dan Kokotov (50:23.960)

as a reporter, just look at the transcript.

Lex Fridman (50:26.200)

Yeah, and you can like embed a little thing,

Lex Fridman (50:28.360)

like into your article, right?

Lex Fridman (50:29.600)

Here's what they said, you can go listen

Dan Kokotov (50:31.240)

to like this clip from the section.

Lex Fridman (50:33.580)

I'm actually trying to figure out,

Dan Kokotov (50:35.960)

I'll probably on the website create

Lex Fridman (50:39.320)

like a place where the transcript goes,

Dan Kokotov (50:41.440)

like as a webpage so that people can reference it,

Lex Fridman (50:44.360)

like reporters can reference it and so on.

Dan Kokotov (50:46.720)

I mean, most of the reporters probably want

Lex Fridman (50:50.880)

to write clickbait articles that are complete falsifying,

Dan Kokotov (50:54.360)

which I'm fine with.

Lex Fridman (50:55.400)

It's the way of journalism, I don't care.

Dan Kokotov (50:57.760)

Like I've had this conversation with a friend of mine,

Lex Fridman (51:01.680)

a mixed martial artist, the Ryan Hall.

Lex Fridman (51:04.920)

And we talked about, you know,

Lex Fridman (51:07.120)

as I've been reading The Rise and Fall of the Third Reich

Lex Fridman (51:09.640)

and a bunch of books on Hitler and we brought up Hitler

Lex Fridman (51:13.240)

and he made some kind of comment where like,

Dan Kokotov (51:17.280)

we should be able to forgive Hitler

Lex Fridman (51:19.480)

and, you know, like we were talking about forgiveness

Lex Fridman (51:23.640)

and we're bringing that up as like the worst case

Lex Fridman (51:25.880)

possible things, like even, you know,

Dan Kokotov (51:28.520)

for people who are Holocaust survivors,

Lex Fridman (51:32.040)

one of the ways to let go of the suffering

Dan Kokotov (51:34.660)

they've been through is to forgive.

Lex Fridman (51:38.000)

And he brought up like Hitler as somebody

Dan Kokotov (51:39.760)

that would potentially be the hardest thing

Lex Fridman (51:42.400)

to possibly forgive, but it might be a worthwhile pursuit

Dan Kokotov (51:45.720)

psychologically, so on, blah, blah, blah, it doesn't matter.

Lex Fridman (51:48.560)

It was very eloquent, very powerful words.

Dan Kokotov (51:50.860)

I think people should go back and listen to it.

Lex Fridman (51:53.160)

It's powerful.

Lex Fridman (51:54.000)

And then all these journalists,

Lex Fridman (51:55.680)

all these articles written about like MMA fight, UFC fight.

Dan Kokotov (52:00.000)

MMA fighter loves Hitler.

Lex Fridman (52:01.920)

No, like, well, no, they didn't.

Dan Kokotov (52:03.680)

They were somewhat accurate.

Lex Fridman (52:05.720)

They didn't say like loves Hitler.

Dan Kokotov (52:07.120)

They said, thinks that if Hitler came back to life,

Lex Fridman (52:13.280)

we should forgive him.

Dan Kokotov (52:14.420)

Like they kind of, it's kind of accurate ish,

Lex Fridman (52:18.520)

but the headline made it sound a lot worse

Dan Kokotov (52:23.800)

than it was, but I'm fine with it.

Lex Fridman (52:27.800)

That's the way the world, I wanna almost make it easier

Dan Kokotov (52:31.640)

for those journalists and make it easier

Lex Fridman (52:33.640)

for people who actually care about the conversation

Dan Kokotov (52:35.800)

to go and look and see.

Lex Fridman (52:37.280)

Right, they can see it for themselves.

Dan Kokotov (52:38.520)

For themselves.

Lex Fridman (52:39.360)

There's the headline, but now you can go.

Dan Kokotov (52:41.640)

There's something about podcasts,

Lex Fridman (52:42.960)

like the audio that makes it difficult

Dan Kokotov (52:44.840)

to jump to a spot and to look

Lex Fridman (52:49.600)

for that particular information.

Dan Kokotov (52:53.240)

I think some of it, I'm interested in creating,

Lex Fridman (52:58.520)

like myself experimenting with stuff.

Lex Fridman (53:00.360)

So like taking rev and creating a transcript

Lex Fridman (53:03.440)

and then people can go to it.

Dan Kokotov (53:05.200)

I do dream that like, I'm not in the loop anymore,

Lex Fridman (53:09.320)

that like, Spotify does it, right?

Dan Kokotov (53:13.000)

Like automatically for everybody,

Lex Fridman (53:16.340)

because ultimately that one click purchase

Dan Kokotov (53:19.600)

needs to be there, like, you know.

Lex Fridman (53:21.800)

Like you kind of want support from the entire ecosystem.

Dan Kokotov (53:24.120)

Exactly.

Lex Fridman (53:24.960)

Like from the tool makers and the podcast creators,

Lex Fridman (53:27.960)

even clients, right?

Lex Fridman (53:28.800)

I mean, imagine if like most podcast apps,

Lex Fridman (53:33.900)

you know, if it was a standard, right?

Lex Fridman (53:35.840)

Here's how you include a transcript into a podcast, right?

Dan Kokotov (53:38.560)

Like it's just an RSS feed ultimately.

Lex Fridman (53:40.920)

And actually just yesterday I saw this company

Dan Kokotov (53:43.520)

called Buzzsprout, I think they're called.

Lex Fridman (53:46.880)

So they're trying to do this.

Dan Kokotov (53:48.320)

They proposed a spec, an extension to their RSS format

Lex Fridman (53:53.420)

to reference transcripts in a standard way.

Lex Fridman (53:58.240)

And they're talking about like,

Lex Fridman (53:59.080)

there's one client dimension that will support it,

Lex Fridman (54:02.140)

but imagine like more clients support it, right?

Lex Fridman (54:04.040)

So any podcast, you could go and see the transcripts

Dan Kokotov (54:08.280)

right in your like normal podcast app.

Lex Fridman (54:10.480)

Yeah.

Dan Kokotov (54:11.320)

I mean, somebody, so I have somebody who works with me,

Lex Fridman (54:15.720)

works with helps with advertising, Matt, this awesome guy.

Dan Kokotov (54:20.200)

He mentioned Buzzsprout to me, but he says,

Lex Fridman (54:22.280)

it's really annoying because they want exclusive,

Dan Kokotov (54:24.880)

they want to host the podcast.

Lex Fridman (54:26.280)

Right.

Dan Kokotov (54:27.120)

This is the problem with Spotify too.

Lex Fridman (54:29.240)

This is where I'd like to say, like F Spotify,

Dan Kokotov (54:33.920)

there's a magic to RSS with podcasts.

Lex Fridman (54:38.600)

It can be made available to everyone.

Lex Fridman (54:40.320)

And then there's all, there's this ecosystem

Lex Fridman (54:42.720)

of different podcast players that emerge

Lex Fridman (54:45.000)

and they compete freely.

Lex Fridman (54:47.080)

And that's a beautiful thing,

Dan Kokotov (54:48.920)

that that's why I go on exclusive,

Lex Fridman (54:50.400)

like Joe Rogan went exclusive.

Dan Kokotov (54:53.320)

I'm not sure if you're familiar with,

Lex Fridman (54:54.760)

he went to Spotify as a huge fan of Joe Rogan.

Dan Kokotov (54:59.560)

I've been kind of nervous about the whole thing,

Lex Fridman (55:01.380)

but let's see, I hope that Spotify steps up.

Dan Kokotov (55:05.000)

They've added video, which was very surprising

Lex Fridman (55:07.020)

that they were able to put on.

Dan Kokotov (55:07.860)

Exclusive meaning you can't subscribe

Lex Fridman (55:10.360)

to the RSS feed anymore.

Dan Kokotov (55:11.680)

It's only in Spotify.

Lex Fridman (55:12.680)

For now you can until December 1st.

Lex Fridman (55:15.720)

And December 1st, it's all, everything disappears

Lex Fridman (55:18.800)

and it's Spotify only.

Dan Kokotov (55:21.480)

I, you know, and Spotify gave him a hundred million dollars

Lex Fridman (55:25.260)

for that.

Lex Fridman (55:26.100)

So it's an interesting deal, but I, you know,

Lex Fridman (55:29.960)

I did some soul searching and I'm glad he's doing it.

Lex Fridman (55:34.480)

But if Spotify came to me with a hundred million dollars,

Lex Fridman (55:38.640)

I wouldn't do it.

Dan Kokotov (55:40.080)

I wouldn't do, well, I have a very different relationship

Lex Fridman (55:42.160)

with money.

Dan Kokotov (55:43.000)

I hate money, but I just think I believe

Lex Fridman (55:46.640)

in the pirate radio aspect of podcasting, the freedom.

Lex Fridman (55:50.000)

And that there's something about.

Lex Fridman (55:50.840)

The open source spirit.

Dan Kokotov (55:52.280)

The open source spirit, it just doesn't seem right.

Lex Fridman (55:54.680)

It doesn't feel right.

Dan Kokotov (55:55.780)

That said, you know, because so many people care

Lex Fridman (55:58.480)

about Joe Rogan's program,

Dan Kokotov (56:00.400)

they're gonna hold Spotify's feet to the fire.

Lex Fridman (56:02.960)

Like one of the cool things with what Joe told me

Dan Kokotov (56:06.400)

is the reason he likes working with Spotify

Lex Fridman (56:10.440)

is that they, they're like ride or die together, right?

Lex Fridman (56:16.200)

So they, they want him to succeed.

Lex Fridman (56:19.160)

So that's why they're not actually telling him what to do

Dan Kokotov (56:22.080)

despite what people think.

Lex Fridman (56:23.800)

They, they don't tell them,

Dan Kokotov (56:24.980)

they don't give them any notes on anything.

Lex Fridman (56:26.940)

They want him to succeed.

Lex Fridman (56:28.560)

And that's the cool thing about exclusivity with a platform

Lex Fridman (56:31.920)

is like, you're kind of wanting each other to succeed.

Lex Fridman (56:36.800)

And that process can actually be very fruitful.

Lex Fridman (56:39.720)

Like YouTube, it goes back to my criticism.

Dan Kokotov (56:44.600)

YouTube generally, no matter how big the creator,

Lex Fridman (56:47.800)

maybe for PewDiePie, something like that,

Dan Kokotov (56:50.160)

they want you to succeed.

Lex Fridman (56:51.680)

But for the most part, from all the big creators

Dan Kokotov (56:53.920)

I've spoken with, Veritasium, all of those folks,

Lex Fridman (56:57.040)

you know, they get some basic assistance,

Lex Fridman (56:59.000)

but it's not like, YouTube doesn't care

Lex Fridman (57:02.800)

if you succeed or not.

Dan Kokotov (57:03.840)

They have so many creators.

Lex Fridman (57:04.680)

Yeah, like a hundred other.

Dan Kokotov (57:06.500)

They don't care.

Lex Fridman (57:07.600)

So, and especially with, with somebody like Joe Rogan,

Dan Kokotov (57:12.840)

who YouTube sees Joe Rogan,

Lex Fridman (57:15.080)

not as a person who might revolutionize the nature of news

Lex Fridman (57:19.760)

and idea space and nuanced conversations.

Lex Fridman (57:23.900)

They see him as a potential person

Dan Kokotov (57:26.280)

who has racist guests on,

Lex Fridman (57:30.240)

or like, you know, they see him as like a headache,

Dan Kokotov (57:33.520)

potentially.

Lex Fridman (57:34.440)

So, you know, a lot of people talk about this.

Dan Kokotov (57:37.980)

It's a hard place to be for YouTube, actually,

Lex Fridman (57:40.600)

is figuring out with the search and discovery process

Dan Kokotov (57:46.840)

of how do you filter out conspiracy theories

Lex Fridman (57:49.040)

and which conspiracy theories represent dangerous untruths

Lex Fridman (57:53.360)

and which conspiracy theories are like vanilla untruths.

Lex Fridman (57:58.080)

And then even when you start having meetings

Lex Fridman (58:00.640)

and discussions about what is true or not,

Lex Fridman (58:03.400)

it starts getting weird.

Lex Fridman (58:05.080)

Yeah, it's difficult these days, right?

Lex Fridman (58:07.800)

I worry more about the other side, right?

Dan Kokotov (58:09.720)

Of too much, you know, too much censorship.

Lex Fridman (58:13.240)

Well, maybe censorship is the right word.

Dan Kokotov (58:14.660)

I mean, censorship is usually government censorship,

Lex Fridman (58:17.960)

but still, yeah, putting yourself in the position

Dan Kokotov (58:21.360)

of arbiter for these kinds of things.

Lex Fridman (58:22.920)

It's very difficult and people think it's so easy, right?

Lex Fridman (58:25.360)

Like, cause like, well, you know, like no Nazis, right?

Lex Fridman (58:27.800)

What a simple principle.

Lex Fridman (58:30.000)

But you know, yes, I mean, no one likes Nazis,

Lex Fridman (58:32.960)

but there's like many shades of gray,

Dan Kokotov (58:35.360)

like very soon after that.

Lex Fridman (58:37.320)

Yeah, and then, you know, of course everybody, you know,

Dan Kokotov (58:39.640)

there's some people that call our current president a Nazi

Lex Fridman (58:42.200)

and then there's like, so you start getting a Sam Harris.

Dan Kokotov (58:45.680)

I don't know if you know that is wasted, in my opinion,

Lex Fridman (58:49.640)

his conversation with Jack Dorsey.

Dan Kokotov (58:51.640)

Now I'll also, I spoke with Jack before in this podcast

Lex Fridman (58:54.200)

and we'll talk again, but Sam brought up,

Dan Kokotov (58:58.240)

Sam Harris does not like Donald Trump.

Lex Fridman (59:01.400)

I do listen to his podcast.

Dan Kokotov (59:03.720)

I'm familiar with his views on the matter.

Lex Fridman (59:06.480)

And he asked Jack Dorsey, he's like,

Lex Fridman (59:08.960)

how can you not ban Donald Trump from Twitter?

Lex Fridman (59:12.280)

And so, you know, there's a set, you have that conversation.

Dan Kokotov (59:15.960)

You have a conversation where some number,

Lex Fridman (59:18.240)

some significant number of people think

Dan Kokotov (59:20.760)

that the current president of the United States

Lex Fridman (59:22.880)

should not be on your platform.

Lex Fridman (59:24.880)

And it's like, okay.

Lex Fridman (59:26.240)

So if that's even on the table as a conversation,

Dan Kokotov (59:29.360)

then everything's on the table for conversation.

Lex Fridman (59:32.720)

And yeah, it's tough.

Dan Kokotov (59:34.640)

I'm not sure where I land on it.

Lex Fridman (59:37.000)

I'm with you, I think that censorship is bad,

Lex Fridman (59:39.440)

but I also think the show...

Lex Fridman (59:41.840)

Ultimately, I just also think, you know,

Dan Kokotov (59:43.960)

if you're the kind of person that's gonna be convinced,

Lex Fridman (59:46.520)

you know, by some YouTube video, you know,

Dan Kokotov (59:49.420)

that, I don't know, our government's been taken over

Lex Fridman (59:53.120)

by aliens, it's unlikely that like, you know,

Dan Kokotov (59:56.080)

you'll be returned to sanity simply because, you know,

Lex Fridman (59:58.960)

that video is not available on YouTube, right?