#151 – Dan Kokotov: Speech Recognition with AI and Humans

[0] The following is a conversation with Dan Kokodov, VP of Engineering at Rev .A .I., which is, by many metrics, the best speech -to -text AI engine in the world.

[1] Rev in general is a company that does captioning and transcription of audio by humans and by AI.

[2] I've been using their services for a couple of years now and planning to use Rev to add both captions and transcripts to some of the previous and future episodes of this podcast.

[3] to make it easier for people to read through the conversation or reference various parts of the episode, since that's something that quite a few people requested.

[4] I'll probably do a separate video on that with links on the podcast website so people can provide suggestions and improvements there.

[5] Quick mention of our sponsors.

[6] Athletic Greens, all -in -one nutrition drink, Blinkist app that summarizes books, business wars podcast, and cash app.

[7] So the choice is health, wisdom, or money.

[8] Choose wisely, my friends.

[9] And if you wish, click the sponsor links below to get a discount and to support this podcast.

[10] As a side note, let me say that I reached out to Dan and the Rev team for a conversation because I've been using and genuinely loving their service and really curious about how it works.

[11] I previously talked to the head of Adobe research for the same reason.

[12] For me, there's a bunch of products, usually it's software.

[13] that comes along and just makes my life way easier.

[14] Examples are Adobe Premiere for video editing, AzoopRX for cleaning up audio, auto hotkey on Windows for automated keyboard of mouse tasks, Emacs as an ID for everything, including the universe itself.

[15] I can keep on going, but you get the idea.

[16] I just like talking to people who create things I'm a big fan of.

[17] That said, after doing this conversation, the folks at rev .ai offered to support sponsor this podcast in the coming months.

[18] This conversation is not sponsored by the guest.

[19] It probably goes without saying, but I should say it anyway, that you cannot buy your way onto this podcast.

[20] I don't know why you would want to.

[21] I wanted to bring this up to make a specific point that no sponsor will ever influence what I do on this podcast, or to the best of my ability, influence what I think.

[22] I wasn't really thinking about this.

[23] For example, when I interviewed Jack Dorsey, who is the CEO of Square that happens to be sponsored in this podcast, but I should really make it explicit.

[24] I will never take money for bringing a guest on.

[25] Every guest on this podcast is someone I genuinely am curious to talk to or just genuinely loves something they've created.

[26] As I sometimes get criticized for, I'm just a fan of people, and that's who I talk to.

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

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

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

[30] If you enjoy this thing, subscribe on YouTube, review it on Apple Podcast, follow on Spotify, support on Patreon, a connect with me on Twitter at Lex Friedman.

[31] As usual, I'll do a few minutes of ads now and no ads in the middle.

[32] I try to make these interesting, but I give you time stamps, so if you skip, please still check out the sponsors by clicking the links in the description.

[33] It is the best way to support this podcast.

[34] This show is sponsored by, who's quickly becoming my favorite sponsor, Athletic Greens, the all -in -one daily drink to support better health and peak performance.

[35] I just, in fact, actually finished drinking it.

[36] It replaced the multivitamin for me, and went far beyond that with 75 vitamins and minerals.

[37] I do intermittent fasting of 16 to 24 hours every day and always break my fast with athletic greens.

[38] I can't say enough good things, can't stop raving about these guys.

[39] It helps me not worry whether I'm getting the nutrients I need.

[40] One of the many reasons I'm a fan is that they keep iterating on the formula, keep improving it like all good engineers and scientists always should be.

[41] Life is not about reaching perfection.

[42] It's about constantly striving for it and making sure each iteration is a positive delta.

[43] The other thing I've taken for a long time outside of Athletic Greens is fish oil.

[44] So I'm especially excited now that they're selling fish oil and are offering listeners of this very podcast.

[45] Free one -month supply of wild caught, a mega -3 fish oil.

[46] When you go to Athletic Greens .com slash Lex to claim the special offer.

[47] By the way, if the link doesn't seem to work for you for whatever reason, sometimes it doesn't if you have an ad blocker enabled.

[48] So try to turn off your ad blocker for this one particular case.

[49] But they're also trying to fix it, so they're on top of it.

[50] Click the Athletic Greens .com slash slacks link in the description to get the fish oil and the all -in -one supplement I rely on for the nutritional foundation of my physical and mental performance.

[51] This episode is supported by Blinkist, my favorite app for learning new things.

[52] Blinkist takes the key ideas from thousands of non -fixirms books and condenses them down into just 15 minutes that you can read or listen to.

[53] I'm a big believer in reading at least an hour every day.

[54] As part of that, I use Blinkist to try out a book.

[55] I may otherwise never have a chance to read.

[56] And in general, it's a great way to broaden your view of the idea landscape out there and find books that you may want to read more deeply.

[57] With Blinkist, you get unlimited access to read or listen to a massive library of condensed nonfiction books.

[58] I also use Blinkist's shortcasts.

[59] That's a lot of S's.

[60] To quickly catch up on a podcast episode I've missed.

[61] Right now, Blinkist has a special offer just for the listeners of this podcast.

[62] Let me take a sip of this drink first.

[63] If you're listening to this, I dare you to try to guess which drink I'm drinking.

[64] Go to Blinkist .com slash Lex to start your free seven -day trial and get 25 % off a Blinkist premium membership.

[65] That's Blinkist spelled B -L -I -N -K -I -S -T, Blinkist .com slash Lex, to get 25 % off and a seven -day free trial.

[66] Blinkist .com slash Lex.

[67] This episode is also brought to you by Business Wars podcast.

[68] Tech entrepreneurs are in all -out race to cash in on our collective addiction to social media.

[69] That sentence was way hard to pronounce than I thought when I first started saying it.

[70] In the newest season of Wanderies, business wars, TikTok versus Instagram, they tracked a war between two social media giants.

[71] I have spoken about possibly entering this space by helping build a new social network.

[72] This is something I struggle with quite a bit because it feels like standing on the edge of a cliff hoping to fly.

[73] I want to keep my mind and heart open, fragile, but it seems that the world can too easily destroy such a mind.

[74] And so I wonder if I'm able to face such challenges.

[75] Perhaps the choice isn't mine to make.

[76] Perhaps it's already been made.

[77] Anyway, this podcast season looks at just one heated competition in the space where the game in my view is not one that makes for a better world.

[78] Listen to the latest season of Business Wars, TikTok versus Instagram on Apple Podcast, Spotify, or listen ad -free by joining Wonderry Plus in the Wonderry app.

[79] I just in general, I highly recommend Wondery.

[80] There's a lot of good podcasts on there.

[81] Finally, the show is presented by Cash App, the number one finance app and the App Store.

[82] When you get it, use code Lex Podcast.

[83] Cash app lets you send money to friends, buy Bitcoin, and invest in the stock market with as little as $1.

[84] I'm thinking of doing some conversation with folks who work in and around the cryptocurrency space similar to artificial intelligence, There are a lot of charlatans in the space, but there's a lot of free thinkers as well, technical geniuses that are worth exploring ideas with in depth and with care.

[85] And I think it's pretty clear that cryptocurrency is here to stay.

[86] Bitcoin just hit $32 ,000.

[87] Anyway, if you get Cashab from the App Store, Google Play, and use code Lex Podcast, you get $10.

[88] And CashUp will also donate $10 to first.

[89] an organization that is helping to advance robotics and STEM education for young people around the world.

[90] And now, here's my conversation with Dan Kokodov.

[91] You mentioned science fiction on the phone, so let's go with the ridiculous first.

[92] What's the greatest sci -fi novel of all time in your view?

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

[94] The greatest sci -fi novel of all time is Dune and the second greatest is the children of Dune and the third greatest is the god emperor of Dune so i'm a huge fan of the whole series i mean it's just an incredible world that he created and i don't know if you've read the book or not no i've not it's one of my biggest regrets especially because the new movie is coming out right everyone's super excited about it i used to it's ridiculous to say and sorry to interrupt is that i used to play the video game used to be dune.

[95] I guess you would call that real -time strategy.

[96] Right, right.

[97] I think I remember that game.

[98] Yeah, it's kind of awesome, 90s or something.

[99] I think I played it actually when I was in Russia.

[100] I definitely remember it.

[101] I was not in Russia anymore.

[102] I think at the time that I used to live in Russia, I think video games were about like the suspicion of Pong.

[103] I think Pong was pretty much like the greatest game I ever got to play in Russia, which was still a privilege, right, in that age.

[104] So you didn't get color?

[105] You didn't get like...

[106] Well, so I left Russia in 1991.

[107] right 191 okay so i was wanted to feel like a kiss because my mom was a programmer so i would go to her work right i would take the the metro i've got our work and play like on i guess the equivalent of like a 286 pc you know nice with floppy disks yes so okay but back to dune what you get back to dune and by the way the new movie i'm pretty interested in but the original i'm a little skeptical i'm a little skeptical i saw the trailer uh i don't know so there's there's a david lynch movie, Dune, as you may know.

[108] I'm a huge David Lynch fan, by the way.

[109] So the movie is somewhat controversial, but it's a little confusing, but it captures kind of the mood of the book better than I would say most any adaptation.

[110] And like, Dune is so much about kind of mood and the world, right?

[111] But back to the philosophical point.

[112] So in the fourth book, God Emperor of Dune, there's a sort of setting where Lito, one of the characters, he's become this weird, of God emperor.

[113] He's turned into a gigantic worm.

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

[115] So the worms are involved.

[116] Worms are involved.

[117] You probably saw the worms in the trailer, right?

[118] And in the video.

[119] So you kind of like merge us with this worm and becomes this tyrant of the world and he like oppresses the people for a long time, right?

[120] But he has a purpose.

[121] And the purpose is to kind of break through kind of a stagnation period in civilization, right?

[122] But people have gotten too comfortable, right?

[123] And so he kind of oppresses them so that they explode and like go on to colonize new worlds and kind of renew the forward momentum of humanity right and so to me that's kind of like fascinating right you need a little bit of pressure and suffering right to kind of like make progress not not get too comfortable maybe that's a bit of a cruel philosophy to take away but that seems to be the case unfortunately obviously I'm a huge fan of suffering so one of the reasons we're talking today is that a bunch of people requested that I do transcripts for this podcast and do captioning I used to make all kinds of YouTube videos and I would go on upwork I think and I would hire folks to do transcription and it was always a pain in the ass from being honest and then I don't really know how i discovered rev but when i did it was this feeling of like holy shit somebody figured out how to do it just really easily i i'm i'm such a fan of just when people take a problem and they just make it easy right you know like just uh there's so many it's like there's so many things in life that you might not even be aware of that are painful then rev you just like give the audio give the video you can actually give a youtube link and then it comes back like a day later or a two days later whatever the hell it is with the captions and all in a standardized format it was i don't know it was it was it was it was truly a joy so i thought i had you know just for the hell of it talk to you that one other product it just made my soul feel good one other product i've used like that is uh for people who might be familiar is called isotope rx it's for audio editing and like and that's another one where it was like you just drop it i dropped in the audio and it just cleans everything up really nicely all the stupid like the mouth sounds and Sometimes there's background sounds due to the malfunctioning of the equipment.

[124] It can clean that stuff up.

[125] It has a general voice denoising.

[126] It has like automation capabilities where you can do batch processing and you can put a bunch of effects.

[127] I mean, it just, I don't know, everything else sucked for like voice -based cleanup that I've ever used.

[128] They've used audition, Adobe Audition.

[129] I've used all kinds of other things with plugins.

[130] and you have to kind of figure it all out.

[131] You have to do it manually.

[132] Here, it just worked.

[133] So that's another one in this whole pipeline.

[134] It just brought joy to my heart.

[135] Anyway, all that to say is Rev put a smile to my face.

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

[137] And Rev .org or Rev .com?

[138] Rev .rev .com.

[139] Same thing, I guess.

[140] We do have rev .AI now as well, which we can talk about later.

[141] Like, do you have the actual domain, or is it just?

[142] The actual domain, but we also use it kind of as a sub -brand.

[143] So we use Rev .com to denote our ASR services, right?

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

[145] So it's like WordPress .com and WordPress .org.

[146] They actually have separate brands that, like, I don't know if you're familiar with what those are.

[147] Yeah, yeah, yeah.

[148] They provide almost like a separate.

[149] branch of a little bit i think with that it's like web WordPress that org is kind of their open source right and uh WordPress dot com is sort of their hosted commercial offering yes um with us the differentiation is a little bit different but maybe similar idea yeah okay so what is rev before i launch into uh what is rev i was going to say you know like you're talking about like rev was music to your years yeah your spiel was music to my ears yeah to us the founders of rev because um rev was kind of founded to improve on the model of Upwork.

[150] That was kind of the original, or part of their original impetus.

[151] Like our CEO, Jason, was an early employee of Upwork.

[152] So it's very familiar with their work to company.

[153] Upworked company.

[154] And so he was very familiar with that model.

[155] And he wanted to make the whole experience better because he knew like when you go, at that time, Upwork was primarily programmers.

[156] So the main thing, they offered us if you want to hire, you know, someone to help you code a little site, right?

[157] You could go on Upwork.

[158] You could browse through a list of freelancers, pick a programmer, have a contract with them and have them do some work.

[159] But it was kind of a difficult experience because for you, you would kind of have to browse through all these people, right?

[160] And you have to decide, okay, well, is this guy good or somebody else better?

[161] And naturally, you know, you're going to Upwork because you're not an expert, right?

[162] If you're an expert, you probably wouldn't be like getting a programmer from Upwork.

[163] So how can you really tell?

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

[165] What if I choose a bad person?

[166] They're like going to be late on the work.

[167] It's going to be a painful experience.

[168] And for the freelancer, it was also painful because, you know, half the time they spent not on actually doing the work, but kind of figuring out how can I make my profile most attractive to the buyer, right?

[169] They're not an expert on that either.

[170] So like Rob's idea was let's remove the barrier, right?

[171] Like, let's make it simple.

[172] we'll pick a few verticals that are fairly standardizable.

[173] We actually started with translation, and then we added audio transcription a bit later.

[174] And we'll just make it a website.

[175] You go, give us your files.

[176] We'll give you back the results as soon as possible.

[177] Originally, maybe it was 48 hours, then we made it shorter and shorter and shorter.

[178] Yeah, there's a rush processing too.

[179] There's a rush processing now.

[180] And we'll hide all the details from you, right?

[181] Yeah.

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

[183] You don't need to worry about the details of how the sausage is made.

[184] That's really cool.

[185] So you picked like a vertical, by vertical you mean basically a service, a service category.

[186] Why translation?

[187] Is Rev thinking of potentially going into other verticals in the future?

[188] Or is this like the focus now is translation, transcription, like language?

[189] The focus now is language or speech services generally, speech to text, language services, you can kind of group them however you want.

[190] So, but we originally, the categorization was work from home.

[191] So one of work that was done by people on a computer, you know, we weren't trying to get into, you know, task rabbit type of things.

[192] And something that could be relatively standard, not a lot of options.

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

[194] So programming wasn't like a good fit because each programming project is kind of unique.

[195] We're looking for something that transcription is, you have five hours of audio, it's five hours of audio, right?

[196] Translation is somewhat similar in that you can have a five -page document, you know, and then you just impress it by that.

[197] And then you pick the language you want, and that's mostly all it is to it.

[198] So those were a few criteria.

[199] We started with translation because we saw the need.

[200] And we picked as kind of a specialty of translation, where we would translate things like voice, because it's immigration documents, things like that.

[201] And so they were fairly even more well -defined and easy to kind of tell if we did a good job.

[202] So you can literally charge per type of document.

[203] Was that the, so what is it now?

[204] Is it pro word or something like that?

[205] Like how do you measure the effort involved in a particular thing?

[206] So now it's for audio transcription, right?

[207] It's per audio unit.

[208] Well, that, yes.

[209] For our translation, we don't really actually focus it out anymore.

[210] But back when it was still a main business of Rabbit was per page, right, or per word, depending on the kind of a...

[211] Because you can also do translation now on the audio, right?

[212] Like subtitles.

[213] So it would be both transcription and translation, that's right.

[214] I wanted to test the system to see how good it is, to see, like, how, is Russian supported?

[215] I think so, yeah.

[216] It'd be interesting to try it out.

[217] But now it's only in like the one direction, right?

[218] So you start with English and then you can have subtitles in Russian, not really the other way.

[219] Got it because I'm deeply curious about this.

[220] When COVID opens up a little bit, when the economy, when the world opens up a little bit.

[221] You want to build your brand in Russia?

[222] No, I don't.

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

[224] I'm definitely not building any brands in Russia.

[225] But I'm going to Paris to talk to the translators of Dostoevsky.

[226] Tolstoy there's this famous couple that does translation and you know I'm more and more thinking of how is it possible to have a conversation with an Russian speaker because I have just some number of famous Russian speakers that I'm interested in talking to and my Russian is not strong enough to be witty and funny I'm already an idiot in English I'm an extra level of like awkward idiot in Russian, but I can understand it, right?

[227] And I also, like, wonder how can I create a compelling English -Russian experience for an English speaker?

[228] Like, if I, there's a guy named Gregorley Perlman, who's a mathematician, who obviously doesn't speak in English.

[229] So I would probably incorporate, like, a Russian translator into the picture.

[230] And then it would be, like, a, not to use a weird term, but like a three, like a three person thing where it's like a dance of where, like I understand it one way.

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

[232] I don't know.

[233] I don't know the right way.

[234] It's complicated.

[235] It's complicated, but I feel like it's worth the effort for certain kinds of people, one of whom I'm confident of Vladimir Putin, I'm for sure talking to.

[236] I really want to make it happen because I think I could do a good job with.

[237] But the right, you know, understanding the fundamentals of translation is something I'm really interested in.

[238] So that's why I'm starting with the actual translators of like Russian literature because they understand the nuance and the beauty of the language and how it goes back and forth.

[239] But I also want to see like in speech, how can we do it in real time.

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

[241] But just to share my dad actually does translation.

[242] Not professional, he writes poetry.

[243] That was kind of always his, not a hobby, but he had a job, like a day job, but his passion was always writing poetry.

[244] And then we get to America, and he started also translating, first he was translating English poetry to Russia, now he also goes the other way.

[245] You kind of gain some small, fame in that world anyways, because recently this poet, like Louis Gluck, I don't know if you know, some American poet, she was awarded the Nobel Prize for Literature, and so my dad had translated one of her books of poetry into Russian, and he was like one of the few, so you kind of like, they asked him and gave an interview to Radios Wabode, if you know what that is, and he kind of talked about some of the intricacies of translating poetry.

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

[247] Because translating poetry is even more challenging than translating just, you know, it's hurtful.

[248] Do you remember any experiences and challenges to having to do the translation that stick out to, like something he's talked about?

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

[250] The way Russian is structured is, first of all, quite different than the way English is structured, right?

[251] There's inflections in Russian and gender is and they don't exist in English.

[252] One of the reasons actually why machine translation is quite difficult for English to Russian and Russian to English, because there are such different languages.

[253] But then English has like a huge number of words, many more than Russian, actually, I think.

[254] So it's often difficult to find the right word to convey the same emotional meaning.

[255] Yeah, Russian language, they play with words much more.

[256] So you were mentioning that Rev was kind of born out of trying to take a vertical on upwork and then standardize it.

[257] We're just trying to make the freelancer marketplace idea better, right?

[258] Better for both customers and better for the freelancers themselves.

[259] Is there something else to the story of a rev, finding Rev?

[260] Like what did it take to bring it to actually to life?

[261] Was there any pain points?

[262] Plenty of pain points.

[263] I mean, as often the case, it's with scaling it up, right?

[264] And in this case, you know, the scaling is kind of scaling the marketplace, so to speak, right?

[265] Reve is essentially a two -sided marketplace, right?

[266] Because, you know, there's the customers and then there's the Revers.

[267] If there's not enough Revers, the Revers are we call our freelancers.

[268] So if there's not enough reveres, then customers have a bad experience, right?

[269] You know, it takes longer to get your work done, things like that.

[270] You know, if there's too many, then the Revers have a bad experience because they might log on to see, like, what work is available, and there's not very much work.

[271] So kind of keeping that balance is a quite challenging problem.

[272] And that's like a problem we've been working on for many years.

[273] We're still like refining our methods, right?

[274] If you can kind of talk to this gig economy idea, I did a bunch of different psychology experiments on Mechanical Turk, for example.

[275] I've asked to do different kinds of very tricky computer vision annotation on Mechanical Turk, and it's connecting people in a more systematized way.

[276] I would say, you know, between task and, what would you call that, worker is what mechanical terror calls it, what do you think about this world of gig economies, of there being a service that connects customers to workers in a way that's like massively distributed, like potentially scaling to, it could be scaled to like tens of thousands of people.

[277] right is there something interesting about that world that you can speak to yeah well we don't think of it as kind of gig economy like to some degree i don't like the word gig that much right because to some degree diminishes the works being done right it sounds kind of like almost amateurish well maybe like music industry like gig is the standard term but in work it kind of sounds like oh it's it's frivolous um to us it's improving the nature of working from home on your own time and on your own terms, right?

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

[279] So, you know, many of our freelancers are maybe work from home moms, right?

[280] And, you know, they don't want the traditional nine to five job, but they want to make some income.

[281] And Ruff kind of like allows them to do that and decide like exactly how much to work and when to work.

[282] Or by the same token, maybe someone is You know, someone wants to live the mountain top, you know, life, right?

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

[284] And, like, generally, that wouldn't be compatible before, before this new world.

[285] You kind of had to choose.

[286] But, like, with Rev, like, you feel like you don't have to choose.

[287] Can you speak to, like, what's the demographics, like, distribution?

[288] Like, where do Revers live?

[289] Is it from all over the world?

[290] Like, what is it?

[291] Do you have a sense of what's out there?

[292] It's all over the world.

[293] Most of them are in the U .S., that's the majority.

[294] Because most of our work is audio transcription, and so you have to speak pretty good English.

[295] So the majority of them are from the U .S. We have people in some other of the English -speaking countries.

[296] And as far as like U .S., it's really all over the place.

[297] You know, for some of the years now, we've been doing these little meetings where the management team will go to some and we'll try to meet reveres.

[298] And pretty much wherever we go, it's pretty easy to find a large number of reveres.

[299] The most recent one we did is in Utah.

[300] But anyway, really.

[301] Are they from all walks of life at these young folks, older folks?

[302] Yeah, all walks of life really.

[303] Like I said, one category is the work from home on, students who want to make some extra income.

[304] There are some people who may be, maybe they have some social anxiety so they didn't want to be in the office, right?

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

[306] So it's really pretty wide variety.

[307] But like on the flip side, for example, one wherever we were talking to was a person who had a fairly high -powered career before and was kind of like taking a break and just wanted, she was almost doing this just to explore and learn about, you know, the gig economy, quote -unquote, right?

[308] So it really is a pretty wide variety of folks.

[309] Yeah, it's kind of interesting through the captioning process for me to learn about the the revers because um like some are clearly like weirdly knowledgeable about technical concepts like you can tell by how good they are at like capitalizing stuff like technical terms like machine learning or deep learning right like i've used the rev to annotate uh to caption um the deep learning lectures or machine learning lectures i did at mit and it's funny like a large number of them were like I don't know if they looked it up or were already knowledgeable but they do a really good job but like they invest time into these things they will like do research they will Google things you know to kind of make sure to they get it right but to some of them it's like it's actually part of the enjoyment of the work like they'll tell us you know I love doing this because I get paid to watch a documentary on something right and I learn something while I'm transcribing right pretty cool yeah so what's that captioning transcription process look like for the reverer?

[310] Can you maybe speak to that to give people a sense, like how much is automated, how much is manual, what's the actual interface look like, all that kind of stuff?

[311] Yeah, so we've invested a pretty good amount of time to give like our reverers the best tools possible.

[312] So typical day forever, they might log into their workspace, they'll see a list of audios that need to be transcribed.

[313] And we try to give them tools to pick specifically the ones they want to do, you know.

[314] So maybe some people like to do longer audios or shorter audience.

[315] People have their preferences.

[316] Some people like to do audios in a particular subject or from a particular country.

[317] So we try to give people the tools to control things like that.

[318] And then when they pick what they want to do, we'll launch a specialized editor that we've built to make transcription as efficient as possible.

[319] They'll start with a speech.

[320] track draft.

[321] So we have our machine learning model for automated speech recognition.

[322] They'll start with that.

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

[324] So it's basically a process of correction.

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

[326] If audio itself is pretty good, like probably like our podcast right now would be quite good.

[327] So they would do a fairly good job.

[328] But if you imagine someone recorded a lecture, you know, in the back of a auditorium, right, where, like, the speaker is really far away and there's maybe a lot of crosstalk and things like that, then maybe the ASR wouldn't do a good job.

[329] So the person might say, like, you know what, I'm just going to do it from scratch.

[330] Yeah.

[331] So it kind of really depends.

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

[333] Like, what's the fastest?

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

[335] as you're listening, can you write as fast as...

[336] Real time would be pretty difficult.

[337] It's actually a pretty...

[338] It's not an easy job.

[339] We actually encourage everyone at the company to try to be a transcriber for a day, descriptions for a day.

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

[341] Because people talk fast and people have accents and all this kind of stuff.

[342] So real time is pretty difficult.

[343] Is it possible?

[344] Like there's somebody...

[345] We're probably going to use Rev. to caption this.

[346] They're listening to this right now.

[347] What do you think is the fastest you can possibly get on this right now?

[348] I think on a good audio, maybe two to three X, I would say, real time.

[349] Meaning it takes two to three times longer than the actual audio of the podcast.

[350] This is so meta.

[351] I can just imagine the rubber is working on this right now.

[352] You're way wrong.

[353] You're way wrong.

[354] This takes way longer.

[355] But yeah, it definitely works.

[356] with me. I could do real time.

[357] Okay, so you mentioned ASR.

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

[359] How much, like, what is the gap between perfect human performance and perfect or pretty damn good ASR?

[360] Yeah, so ASR automatic speech recognition.

[361] It's a class of machine learning problem, right, to take, you know, speech like we're talking and transform it into a sequence of words, essentially.

[362] Audio of people talking.

[363] Audio, audio to words.

[364] And, you know, there's a variety of different approaches and techniques, which we could talk about later if you want.

[365] So, you know, we think we have pretty much the world's best ASR for this kind of speech, right?

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

[367] Like one domain might be voice assistance, right?

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

[369] because Siri, there's fairly limited vocabulary.

[370] You know, you might ask Siri to play a song or, you know, word of a pizza or whatever.

[371] And it's very good at doing that.

[372] Very different from when we're talking in a very unstructured way.

[373] And Siri will also generally, like, adapt to your voice and stuff like this.

[374] So for this kind of idea, we think we have the best.

[375] And our accuracy, right now I think it's maybe 14 % word error rate on our test.

[376] suite that we generally use to measure.

[377] So word error rate is like one way to measure accuracy for ASR, right?

[378] So what's 14 %?

[379] So 14 % means across this test suite of a variety of different audience, it would be, it would get in some way 14 % of the words wrong.

[380] 14 % of the words wrong.

[381] Yeah.

[382] So the way you kind of calculated is you might add up insertions, deletions, and substitutes.

[383] right?

[384] So consortions is like extra words, deletions are words that we said, but weren't in the transcript, right?

[385] Substitution says you said Apple, but I said, but the ASR thought it was able, something like this.

[386] Human accuracy, most people think realistically it's like 3%, 2 % word error rate would be like the max achievable.

[387] So there's still quite a gap, right?

[388] would you say that so YouTube when I upload videos often generates automatic captions are you sort of from a company perspective from a tech perspective are you trying to beat YouTube a Google it's a hell of a Google I mean I don't know how seriously they take this task but I imagine it's quite serious and they you know Google is probably up there in terms of their teams on um on A .S. or just NLP natural language processing different technologies.

[389] So do you think you can be Google?

[390] On this kind of stuff, yeah, we think so.

[391] Google just woke up on my thing.

[392] This is hilarious.

[393] Now Google is listening, sending it back to headquarters.

[394] Who are these rough people?

[395] But that's the goal?

[396] Yeah, I mean, we measure ourselves against like Google, Amazon, Microsoft, you know, some of some smaller competitors.

[397] And we use like our internal tests.

[398] We try to compose it of a pretty representative set of ideas.

[399] Maybe it's some podcast, some videos, some interviews, some lectures, things like that, right?

[400] And we beat them in our own testing.

[401] And actually Rev offers automated, like you can actually just do the automated captioning.

[402] So I guess it's like way cheaper, whatever it is, whatever the rates are.

[403] Yeah, yeah.

[404] It's a, by the way, it used to be a dollar per minute for captioning and transcription.

[405] I think it's like a $1 .15 or something like that.

[406] $1 .25.

[407] $1 .25 now.

[408] Yeah.

[409] It's pretty cool.

[410] That was the other thing that was surprising to me. It was actually like the cheapest thing you could one of the, I mean, I don't remember it being cheaper.

[411] You could on Upwork get cheaper, but it was clear to me that this, that's going to be really shitty.

[412] Yeah.

[413] So like you're also competing on price.

[414] I think there were services that you can get.

[415] like similar to rev kind of um feel to it but it wasn't as automated like the drag and drop the entirety of the interface it's like the thing we're talking about i'm such a huge fan of like frictionless like uh amazon's single uh buy button whatever yeah yeah that one click the one click that's genius right there like that is so important for services yeah that simplicity and i mean Reve is almost there.

[416] I mean, there's like some trying to think.

[417] So I think I've stopped using this pipeline, but Rev offers it and I like it, but it was causing me some issues on my side, which is you can connect it to like Dropbox and it generates the files and Dropbox.

[418] So it closes the loop to where I don't have to go to Rev at all.

[419] and I can download it.

[420] Sorry, I don't have to go to Rev at all and to download the files.

[421] It could just like automatically copy them.

[422] Right, you're putting your Dropbox on, you know, a day later or maybe a few hours later.

[423] Yeah, it just shows up.

[424] Yeah, I was trying to do it programmatically too.

[425] Is there an API interface?

[426] You can, I was trying to through like, through Python to download stuff automatically, but then I realized this is the programmer in me. Like, dude, You don't need to automate everything, like, in life, like, flawlessly.

[427] Because I wasn't doing enough captions to justify to myself the time investment into automating everything perfectly.

[428] Yeah, I would say if you're doing so many interviews that your biggest roadblock is clicking on the rough download, but now you're talking about Elon Musk levels of business.

[429] But for sure, we have, like, yeah, a variety of ways to make it easy.

[430] You know, there's the integration.

[431] You mentioned, I think, a store company called Sapier, which kind of can come connect Dropbox to Rev and vice versa.

[432] We have an API, if you want to really customize it, you know, if you want to create the Lex Friedman, you know, CMS or whatever.

[433] For this whole thing.

[434] Okay, cool.

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

[436] Like, what does it take, like, approach -wise, machine learning -wise, how hard is this problem?

[437] How do you get to the 3 % error rate?

[438] Like, what's your vision?

[439] of all of this.

[440] Yeah, well, the 3 % error rate is definitely, that's the grand vision.

[441] We'll see what it takes to get there.

[442] But we believe, you know, in ASR, the biggest thing is the data, right?

[443] Like, that's true of, like, a lot of machine learning problems today, right?

[444] The more data you have and the high quality the data, the better label the data.

[445] That does get good results, and we at Rev have kind of like the best data.

[446] Like we have.

[447] Like you're literally, your business model is annotating the data.

[448] Our business model is being paid to annotate the data.

[449] Being paid to annotate the data.

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

[451] Yeah.

[452] And so we've kind of like ridden this flywheel to this point.

[453] And we think we're still kind of in the early stages of figuring out all the parts of the flywheel to use, you know, because we have the final transcripts.

[454] and we have the audios, and we train on that.

[455] But we, in principle, also have all the edits that the reverers make, right?

[456] Oh, that's interesting.

[457] How can you use that as data?

[458] That's something for us to figure out in the future.

[459] But, you know, we feel like we're only in the early stages, right?

[460] So the data, but the data is there.

[461] That'd be interesting, like, almost like recurrent, Yolnet for fixing, for fixing transcripts.

[462] I always remember we did segmentation annotation for driving data, so segmenting the scene, like visual data.

[463] And you can get all, so it was drawing people drawing polygons around different objects and so on.

[464] And it feels like it always felt like there was a lot of information in the clicking, the sequence of clicking that people do, the kind of fixing of the polygons that they do.

[465] Now, there's a few papers written about how to, to draw polygons with recurrent neural nets to try to learn from the human clicking.

[466] But it was just like experimental, you know, it was one of those like CVPR type papers that people do like a really tiny data set.

[467] It didn't feel like people really try to do it seriously.

[468] And I wonder, I wonder if there's information in the fixing that's high, that provides deeper set of signal than just like the raw.

[469] data.

[470] The intuition is for sure there must be, right?

[471] It must be.

[472] In all kinds of signals and how long you took to, you know, make that edit and stuff like that.

[473] It's going to be like up to us.

[474] That's why like the next couple of years is like super exciting for us, right?

[475] So that's what like the focus is now.

[476] You mentioned rev .a .ai.

[477] That's where you want to.

[478] Yeah, so rev. That AI is kind of our way of bringing this ASR to, you know, the rest of the world, right?

[479] So when we started, we were human only.

[480] Then we kind of created this Temi service.

[481] I think you might have used it, which was kind of ASR for the consumer.

[482] So if you don't want to pay $1 .25, but you want to pay, now it's $25 a minute, I think.

[483] And you get the transcript, the machine generated transcript, and you get an editor, and you can kind of fix it up yourself, right?

[484] Then we started using the ESR for our own human transcriptionists.

[485] And then the kind of RAI is the final step of the journey, which is, you know, we have this amazing engine.

[486] What can people build with it, right?

[487] What kind of new applications could be enabled if you have speed track that's that accurate?

[488] Do you have ideas for this, or is it just providing it as a service and seeing what people come up with?

[489] It's providing it as a service and seeing what people come up with and kind of learning from what people do with it.

[490] And we have ideas of our own as well, of course.

[491] But it's a little bit like, you know, when AWS provided the building blocks, right?

[492] And they saw what people built with it, and they try to make it easier to build those things, right?

[493] And we've kind of hope to do the same thing.

[494] Although AWS kind of does a shitty job of like, I'm continually surprised, like, Mechanical Turk, for example, how shitty the interface is.

[495] We're talking about, like, Rev making me feel good.

[496] Like, when I first discover Mechanical Turk, the initial idea of it was like, it made me feel like Rev does.

[497] but then the interface is like, come on.

[498] Yeah, it's horrible.

[499] Why is it so painful?

[500] Does nobody at Amazon want to, like, seriously invest in it?

[501] It felt like you could make so much money if you took this effort seriously.

[502] And it feels like they have a committee of like two people just sitting back, like a meeting.

[503] They meet once a month.

[504] Like, what are we going to do with Mechanical Turk?

[505] It's like two websites make me feel like this.

[506] and Craiglist .org, whatever the hell it is.

[507] It feels like it's designed in the 90s.

[508] Well, Craigslist basically hasn't been updated pretty much since the guy originally built.

[509] Do you seriously think there's a team, like, how big is the team working on a mechanical truck?

[510] I don't know.

[511] There's some team, right?

[512] I feel like there isn't.

[513] I'm skeptical.

[514] Well, if nothing else, they benefit from, you know, the other teams, like, moving things forward.

[515] Right, in a small way.

[516] Possibly.

[517] But no, I know, I mean, we use mechanical turf for a couple of things as well, and it's painful UI.

[518] It's painful.

[519] But, yeah, it works.

[520] I think most people, the thing is most people don't really use the UI, right?

[521] Like, so, like, we, for example, we use it through the API, right?

[522] But even the API documentation and so on, like, it's super outdated.

[523] Like, it's, I don't even know what to, I mean, the same, like, same criticism, as long as we're ranting.

[524] my same criticism goes to the APIs of most of these companies like Google for example the API for the different services is just the documentation is so shitty like it's not so shitty I should I should actually be I should exhibit some gratitude okay let's practice some gratitude the the you know the documentation is pretty good like most of the the things that the API makes available is pretty good.

[525] It's just that in the sense that it's accurate, sometimes outdated, but like the degree of explanations with examples is only covering, I would say, like 50 % of what's possible.

[526] And it just feels a little bit, like there's a lot of natural questions that people would want to ask that doesn't get covered.

[527] And it feels like it's almost there.

[528] Like it's such a magical thing.

[529] like the Maps API, YouTube API, there's a bunch of stuff.

[530] I gotta imagine it's like, you know, there's probably some team at Google, right, responsible for writing this documentation.

[531] It's probably not the engineers, right?

[532] And probably this team is not, you know, where you want to be.

[533] Well, it's a weird thing.

[534] I sometimes think about this for somebody who wants to also build the company.

[535] I think about this a lot, you know, YouTube, the, you know, the service is one of the most magical, like, I'm so grateful that YouTube exists.

[536] And yet, they seem to be quite clueless on so many things, like that everybody's screaming them at.

[537] Like, it feels like whatever the mechanism that you use to listen to your quote -unquote customers, which is like the creators, is not very good.

[538] Like, there's literally people that are like screaming why like uh their new youtube studio for example there's like features that that were like begged for for a really long time like being able to upload multiple videos at the same time that wasn't missing for a really really long time now like there's probably things that i don't know which is maybe for that kind of huge infrastructure is actually very difficult to build some of these features but the fact that that wasn't communicated and it felt like you're not being heard.

[539] Like, I remember this experience for me, and it's not a pleasant experience.

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

[541] And that's something to think about.

[542] I'm not sure what that is.

[543] That might have to do with just, like, small groups working on these small features, and these specific features.

[544] And there's no overarching, like, dictator type of human that says, like, why the hell are we neglecting?

[545] Like, Steve Jobs type of character is, like, there's people.

[546] that we need to speak to the people that like want to love our product and they don't let's fix at some point you just get so fixated on the numbers right and it's like well the numbers are pretty great right like people are watching you know doesn't seem to be a problem right doesn't see you don't and you're not like the person that like build this thing right so you really care about it you know you're just there you came in as a product manager right you got hired sometime later your mandate is like increased this number like you know 10 % right right right and you just that's brilliantly put like if you this is okay if there's a lesson in this is don't reduce your company into a metric of like how much uh like you said how much how much people watching the videos and so on and and and like convince yourself that everything is working just because the numbers are going up yeah there's something you have to have a vision you have to uh you have to want people to love your stuff because love is ultimately the beginning of like a successful long -term company is that they always should love with your product you have to be like a creator and have that like creator's love for your own thing right like and you paint by you know these these comments right and probably like Apple i think did this generally like really well they're they're well known for kind of keeping teams small even when they were big right and you know he was an engineer like there's that book creative selection i don't know if you read it by an Apple engineer named Ken Kosyenda.

[547] It's kind of a great book, actually, because unlike most of these business books where it's, you know, here's how Steve Job ran the company.

[548] It's more like, here's how life was like for me, you know, an engineer.

[549] Here, the projects I worked on and hear what it was like to pitch Steve Jobs, you know, on like, you know, I think it was in charge of like the keyboard and the auto correction, right?

[550] And at Apple, like, Steve Jobs reviewed everything.

[551] And so he was like, this is what it was like to show my demos to Steve Jobs.

[552] you know, to change them because like Steve Jobs didn't like how, you know, the shape of the little key was off because the rounding of the corner was like not quite right or something like this, but he was famously a stickler for this kind of stuff.

[553] But because the teams were small, you really owned this stuff, right?

[554] So you really cared.

[555] Yeah, Elon Moss does that similar kind of thing with Tesla, which is really interesting.

[556] There's another lesson in leadership in that is to be obsessed with the details.

[557] And like, he talks to like the lowest level engineers.

[558] Okay, so we're talking.

[559] talking about ASR and so this is basically where I was saying we're going to take this like ultra seriously and then what's the mission to try to keep pushing towards a 3 % yeah and kind of try to try to build this platform where all of your you know audits all of your meetings you know they're as easily accessible as your notes right like so like imagine all the meetings the company might have right Now that I'm like no longer a programmer, right?

[560] And I'm a quote unquote manager.

[561] That's less like my days in meetings, right?

[562] Yeah.

[563] And, you know, pretty often I want to like see what was said, right?

[564] Who said it?

[565] You know, what's the context?

[566] But it's generally not really something they can easily retrieve, right?

[567] Like imagine if all of those meetings were indexed archived, you know, you could go back, you could share a clip like really easily, right?

[568] So that might change completely.

[569] It's like everything that's said converted to text.

[570] might change completely the dynamics of what we do in this world, especially now with remote work, right?

[571] Exactly, exactly.

[572] With Zoom and so on, that's fascinating to think about.

[573] I mean, for me, I care about podcasts, right?

[574] And one of the things that was, you know, I'm torn.

[575] I know a lot of the engineers at Spotify, so I love them very much because they dream big in terms of like, they want to empower creators.

[576] So one of my hopes was, was Spotify that they would use the technology like Rev or something like that to start converting everything into text and make it indexable.

[577] Like one of the things that sucks with podcast is like it's hard to find stuff.

[578] Like the model is basically subscription.

[579] Like you find, it's similar to what YouTube used to be like, which is you basically find a creator that you enjoy and you subscribe to them and like you just yeah uh you just kind of follow what they're doing but the search and discovery wasn't a big part of youtube like in the early days but and that's what currently with podcast like is the search and discovery is uh like non -existent you're basically searching for like the dumbest possible thing which is like keywords in the titles of episodes yeah but even the from searching this kind of all the time.

[580] So I listen to like a number of podcasts and there's something said and I want to like go back to that later because I was trying to remember, what do you say?

[581] Like maybe like recommended some cool product that I want to try out.

[582] And like it's basically impossible.

[583] Maybe like some people have pretty good show notes.

[584] So maybe you'll get lucky and you can find it.

[585] But I mean if everyone had transcripts and it was all searchable, it would be a game changer.

[586] It would be so much better.

[587] I mean, that's one of the things that I wanted to, I mean, one of the reasons we're talking today.

[588] is I want to take this quite seriously, the rough thing.

[589] I've just been lazy.

[590] So because I'm very fortunate that a lot of people support this podcast, that there's enough money now to do transcription and so on, it seemed clear to me, especially like CEOs and sort of like PhDs, like people write to me who are like graduate students at computer science or graduate students and whatever the heck field.

[591] it's clear that their mind, like they enjoy podcasts when they're doing laundry or whatever, but they want to revisit the conversation in a much more rigorous way.

[592] And they really want a transcript.

[593] It's clear that they want to like analyze conversations.

[594] So many people wrote to me about a transcript for Yoshibach conversation.

[595] I had just a bunch of conversations.

[596] And then on the Elon Musk side, like reporters want like, they want to write a blog post about your conversation.

[597] So they want to be able to pull stuff.

[598] And it's like they're essentially doing on your conversation transcription privately.

[599] They're doing it for themselves and then starting to pick.

[600] But it's so much easier when you can actually do it as a reporter.

[601] Just look at the transcript.

[602] Yeah.

[603] And you can like embed a little thing, you know, like into your article, right?

[604] Here's what they said.

[605] You can go listen to like this clip from the section.

[606] I'm actually trying to figure out.

[607] I'll probably on the website create like a place where the.

[608] transcript goes like as a web page so that people can reference it like reporters can reference and so on i mean most of the reporters probably i want to write clickbait articles that are complete falsifying which i'm fine with it's the way of journalism i don't care like i've had this conversation with a friend of mine a mixed martial artist the ryan hall and we we talked about you know as i've been reading the rise and fall of the third rike and a bunch of books on Hitler.

[609] And we brought up Hitler and he made some kind of comment where like we should be able to forgive Hitler.

[610] And, you know, like we were talking about forgiveness and we're bringing that up as like the worst case possible things.

[611] Like even, you know, for people who are Holocaust survivors, one of the ways to let go of the suffering they've been through is to, is to forgive.

[612] and he brought up like Hitler as somebody that would potentially be the hardest thing to possibly forgive, but it might be a worthwhile pursuit psychologically, so on, blah, blah, blah, it doesn't matter.

[613] It was very eloquent, very powerful words.

[614] I think people should go back and listen to it.

[615] It's powerful.

[616] And then all these journalists, all these articles written about like MMA fight, UFC fight or right.

[617] The United States, no, like, well, no, they were somewhat accurate.

[618] They didn't say like loves Hitler.

[619] said, thinks that if Hitler came back to life, we should forgive him.

[620] Like, they kind of, it's kind of accurate -ish, but it, the headline made it sound a lot worse than, than it was, but I'm fine with it.

[621] That's the way the world.

[622] I want to almost make it easier for those journalists and make it easier for people who actually care about the conversation to go and look and see.

[623] Right, they can see it for themselves.

[624] For themselves.

[625] There's the headline, but that you can go.

[626] There's something about podcasts, like the audio that makes it difficult to go, to jump to a spot and to look for that particular information.

[627] I think some of it, you know, I'm interested in creating, like myself experimenting with stuff.

[628] So like taking Rev and creating a transcript and then people can go to it.

[629] I do dream that, like, I'm not in the loop anymore, that, like, you know, Spotify does it, right?

[630] Like, automatically for everybody.

[631] Because ultimately that one -click purchase needs to be there.

[632] Like, you kind of want support from the entire ecosystem.

[633] Exactly.

[634] From the tool makers and the podcast creators, even clients, right?

[635] I mean, imagine if, like, most podcast apps, you know, if they're, if they, if it was a standard, right?

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

[637] Podcast is just an RSS feed, ultimately.

[638] And actually, just yesterday I saw this company called Buzz Sprout, I think, they're called.

[639] So they're trying to do this.

[640] They propose the spec, an extension to their RSS format to reference podcasts, reference transcripts in a standard way.

[641] And they're talking about, like, there's one client dimension that will support it.

[642] But imagine, like, more clients support it, right?

[643] So any podcast you could go and see the transcripts right in your like normal podcast app.

[644] Yeah.

[645] I mean somebody, so I have somebody who works with me, works with helps with advertising.

[646] Matt is an awesome guy.

[647] He mentioned bus brought to me, but he says it's really annoying because they want exclusive.

[648] They want to host the podcast.

[649] Right.

[650] This is the problem with Spotify too.

[651] This is where I'd like to say like F Spotify.

[652] there's a magic to RSS with podcasts.

[653] It can be made available to everyone.

[654] And then there's this ecosystem of different podcast players that emerge and they compete freely.

[655] And that's a beautiful thing.

[656] That's why I go on exclusive.

[657] Like Joe Rogan went exclusive.

[658] I'm not sure if you're familiar with.

[659] He went to just Spotify.

[660] As a huge fan of Joe Rogan, I've been kind of nervous about the whole thing.

[661] But let's see.

[662] I hope that Spotify, steps up they've added video which is very surprising that they were so so explicit meaning you can't subscribe to his RSS feed anymore it's only in Spotify for now you can until December 1st and December 1st it's all everything disappears in it's Spotify only I uh you know and Spotify gave him a hundred million dollars for that so it's it's uh it's an interesting deal but I I you know I did some soul searching and I'm glad he's doing it but if Spotify came to me with $100 million, I wouldn't do it.

[663] I wouldn't do, well, I have a very different relationship with money.

[664] I hate money, but I just think, I believe in the pirate radio aspect of podcasting, the freedom.

[665] And that there's something about the open source spirit.

[666] The open source spirit, it just doesn't seem right, doesn't feel right.

[667] That said, you know, because so many people care about Joe Rogan's program, they're going to hold Spotify's feet to the fire.

[668] Like, one of the cool things, what Joe told me is the reason he likes working with Spotify is that they're like ride or die together, right?

[669] So they want him to succeed.

[670] So that's why they're not actually telling him what to do, despite what people think.

[671] They don't give him any notes on anything.

[672] They want him to succeed.

[673] And that's the cool thing about exclusivity with the platform is like, you kind of want each other to succeed, and that process can actually be very fruitful.

[674] Like YouTube, it goes back to my criticism, YouTube generally, no matter how big the creator, maybe for PewDiePy, something like that, they want you to succeed.

[675] But for the most part, from all the big creators I've spoken with, Veritasim, all those folks, you know, they get some basic assistance, but it's not like YouTube doesn't care if you succeed or not.

[676] They have so many creators.

[677] They have like a hundred other.

[678] They don't care.

[679] And especially with somebody like Joe Rogan who YouTube sees Joe Rogan not as a person who might revolutionize the nature of news and idea space and nuanced conversations.

[680] They see him as a potential person who has racist guests on.

[681] Or like, you know, they see him as like, a headache, potentially.

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

[683] It's a hard place to be for YouTube, actually, is figuring out with the search and discovery process of how do you filter out conspiracy theories and which conspiracy theories represent dangerous untruths and which conspiracy theories are, like, vanilla untruths.

[684] And then even when you start having meetings and discussions about what is true or not, it starts getting weird it starts getting weird yeah it's difficult these days right I worry more about the other side right of too much you know too much censorship well maybe censorship is the right word I mean censorship is usually government censorship but still yeah putting yourself in a position of arbiter for these kinds of things it's very difficult and people think it's so easy right like well you know like no Nazis right what a simple principle but you know yes I mean no one likes Nazis but there's like many shades of gray like very soon after that yeah and then you know of course everybody you know there's some people that call our current president a Nazi and then there's like you start getting a Sam Harris I don't know if you know that is wasted in my opinion his conversation with Jack Dorsey and I'll also I spoke with Jack before on this podcast and we'll talk again but Sam brought up Sam Harris does not like Donald Trump I do listen to to his podcast.

[685] I'm familiar with his views on the matter.

[686] And he asked Jack Dorsey's like, how can you not ban Donald Trump from Twitter?

[687] So, you know, there's a set, you have that conversation.

[688] You have a conversation where some number or some significant number of people think that the current president of the United States should not be on your platform.

[689] And it's like, okay, so if that's even on the table as a conversation, then everything is on the table for conversation.

[690] And yeah, it's tough.

[691] I'm not sure where I land on it.

[692] I'm with you.

[693] I think that censorship is bad, but I also think the show.

[694] Ultimately, I just also think, you know, if you're the kind of person that's going to be convinced, you know, by some YouTube video, you know, that, I don't know, our government's been taken over by aliens.

[695] It's unlikely that, like, you know, you'll be returned to sanity simply because, you know, that video is not available on YouTube, right?

[696] Yeah, I'm with.

[697] I tend to believe in the intelligence of people and we should trust them.

[698] But I also do think it's the responsibility of platforms to encourage more love in the world, more kindness to each other.

[699] And I don't always think that they're great at doing that particular thing.

[700] So that there's a nice balance there.

[701] And I think philosophically, I think about that a lot.

[702] Where's the balance between free speech?

[703] and, like, encouraging people, even though they have the freedom of speech to not be an asshole.

[704] Yeah, right.

[705] That's not a constitutional, like, so you have the right for free speech, but, like, just don't be an asshole.

[706] Like, you can't really put that in the Constitution.

[707] The Supreme Court can't be like, just don't be a dick.

[708] But I feel like platforms have a role to be like, just be nicer.

[709] Maybe do the carrot, like encourage people to be nicer, as opposed to the stake of censorship.

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

[711] Just be nicer.

[712] Yeah, machine learning for niceness.

[713] It is.

[714] Responsible AI.

[715] I mean, it is a thing for sure.

[716] Jack Dorsey kind of talks about it as a vision for Twitter is how do we increase the health of conversations.

[717] I don't know how seriously they're actually trying to do that, though.

[718] which is one of the reasons that in part considering entering that space a little bit.

[719] It's difficult for them, right?

[720] Because, you know, it's kind of like well -known that, you know, people are kind of driven by, you know, rage and, you know, outrage maybe is a better word, right?

[721] Outrage drives engagement and, well, these companies are judged by engagement, right?

[722] In the short term, but this goes to the metrics thing that we were talking about earlier.

[723] I do believe, I have a fundamental belief that if you have a metric of long -term happiness of your users, like not short -term engagement, but long -term happiness and growth and both like intellectual, emotional health of your users, you're going to make a lot more money.

[724] You're going to have long -term, like, you should be able to optimize for that.

[725] You don't need to necessarily optimize for engagement.

[726] Yeah.

[727] And that'll be good for society too.

[728] Yeah, no, I mean, I generally agree with you, but it really is.

[729] acquires a patient person with, you know, trust from Wall Street to be able to carry out such strategy.

[730] This is what I believe is Steve Jobs character and Elon Musk's character is like, you basically have to be so good at your job.

[731] Right.

[732] You got a pass for anything.

[733] That you can hold the board and all the investors hostage by saying, like, either we do it my way or I leave.

[734] And everyone is too afraid of you to leave.

[735] right because they believe in your vision so that but that requires being really good at really good at what you do requires being Steve Jobs and Elon Musk and there's kind of a reason why like a third name doesn't come immediately to mind right like there's maybe a handful of other people but it's not that many it's not many I mean people say like why are you like people say that I'm like a fan of Elon Musk I'm not I'm a fan of anybody who's like Steve Jobs and Elon Musk and there's just not many of those folks it's a guy that made us believe that like we can get to Mars, you know, in 10 years, right?

[736] I mean, that's kind of awesome.

[737] And it's kind of making it happen, which is like, it's great.

[738] It's kind of gone like that kind of like spirit, right?

[739] Like from a lot of our society, right?

[740] You know, like we can get to the moon in 10 years and like we did it, right?

[741] Yeah, especially in this time of so much kind of existential dread that people are going through because of COVID.

[742] Like having rockets that just keep going out there.

[743] now with humans, I don't know.

[744] Just like you said, I mean, it gives you a reason to wake up in the morning and dream.

[745] For us engineers, too.

[746] It is inspiring as hell, man. Well, let me ask you this, the worst possible question, which is, so you're like, at the core, you're a programmer, you're an engineer, but now you made the unfortunate choice.

[747] Or maybe that's the way life goes of basically moving away from the low -level work and becoming a manager, becoming an executive, having meetings.

[748] What's that transition been like?

[749] It's been interesting.

[750] It's been a journey.

[751] Maybe a couple of things to say about that.

[752] I got into this, right?

[753] Because as a kid, I just remember this incredible amazement at being able to write a program, right?

[754] and something comes to life that kind of didn't exist before.

[755] I don't think you have that in like many other fields.

[756] You have that with some other kinds of engineering, but you're maybe a little bit more limited with what you can do like, right?

[757] But with a computer, you can literally imagine any kind of program, right?

[758] So it's a little bit godlike what you do, like when you create it.

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

[760] Do you remember, like, first program you wrote, or maybe the first program that, like, made you fall in love with computer science?

[761] I don't know what was the first program.

[762] It's probably like trying to write one of those games in basic, you know, like emulate the snake game or whatever.

[763] I don't remember, to be honest.

[764] But I enjoyed, like, that's why I always loved about, you know, being a program.

[765] It's just the creation process.

[766] And it's a little bit different when you're not the one doing the creating.

[767] And, you know, another aspect to it, I would say, is, you know, when you're a program or when you're a individual contributor, it's kind of very easy to know when you're doing a good job.

[768] when you're not doing a good job, when you're being productive, when you're not being productive, right?

[769] You can kind of see, like, you're trying to make something and it's like slowly coming together, right?

[770] And when you're a manager, you know, it's more diffuse, right?

[771] Like, well, you hope, you know, you're motivating your team and making them more productive and inspiring them, right?

[772] But it's not like you get some kind of like dopamine signal because you like completed X lines of code, you know, today.

[773] So kind of like you missed that dopamine rush a little bit when you first become.

[774] But then, you know, slowly you kind of see, yes, your teams are doing amazing work, right?

[775] And you can take pride in that.

[776] You can get like a, what is it, like a ripple effect of somebody else's dopamine.

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

[778] So is there pain points and challenges yet to overcome from becoming, from going to a programmer to becoming a programmer of humans?

[779] Programmer of humans.

[780] I don't know.

[781] Humans are difficult to understand, you know?

[782] It's like one of those things, like trying to understand other people's motivations and what really drives them.

[783] It's difficult.

[784] Maybe, like, never really know, right?

[785] Do you find that people are different?

[786] Yeah.

[787] Like, I, one of the things, like, I had a group at MIT that, you know, I found that like some people I could like scream at and criticize like hard and that made them do like much better work and really push them to their limit and there's some people that I had to nonstop compliment because like they're so already self -critical like about everything they do that I have to be constantly like I cannot criticize them at all because they're already criticizing themselves and you have to kind of encourage and like celebrate their little victories.

[788] And it's kind of fascinating how that, the complete difference in people.

[789] Definitely people respond to different motivations and different modes of feedback and you kind of have to figure it out.

[790] It's like a pretty good book, which some reason now the name escapes me, about management.

[791] First break all the rules.

[792] First break all the rules.

[793] First break all the rules.

[794] It's a book that we generally ask a lot of, like, first time manages to read it up.

[795] And, like, one of the kind of philosophies is managed by exception, right?

[796] Which is, you know, don't, like, have some standard template.

[797] Like, you know, here's how I, you know, tell this person to do this or the other thing.

[798] Here's how I get feedback.

[799] Like, managed by exception, right?

[800] Every person is a little bit different.

[801] You have to try to understand what drives them and tailor it to them.

[802] Since you mentioned books, I don't know if you can answer this question.

[803] but people love it when I ask it, which is, are there books, technical, fiction, or philosophical that you enjoyed or had an impact on your life as you would recommend?

[804] You already mentioned Dune, like all of the Dune.

[805] All of the Dune.

[806] The second one was probably the weakest, but anyway, so, yeah, all of the Dune is good.

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

[808] How many Dune books are there?

[809] Like, do you recommend people start with the first one if that was...

[810] Yeah, you've got to have to read.

[811] them all i mean it is a complete story right so um you start with the first one you got to read all of them um so it's not like a tree like that like a creation of like the universe that you should go in sequence yeah it's uh it's kind of a chronological storyline um there's six books in all uh then there's like many um kind of off books that were written by um Frank Herbert's son uh but those are not as good so you don't have to bother with those shots fired but the main sequence is good so what are some other books maybe there's a few so i don't know that like i would say there's a book that kind of i don't know turn my life around or anything like that but here's a couple that i really love uh so one is brave new world by aldous huxley um and it's kind of incredible how prescient he was about like what this what what a brave new world might be like right you know you kind of see genetic sorting in this book right where there's like these alphas and epilons and uh from like the earliest time of society like they're sort like you can kind of see in a slightly similar way today where well one of the problems with society is people are kind of genetically sorting a little bit right like there's much less like most most marriages right between people of similar kind of intellectual level or socioeconomic status, more so these days than in the past.

[812] And you kind of see some effects of it in stratifying society.

[813] And kind of he illustrated what that could be like in the extreme.

[814] Different versions of it on social media as well.

[815] It's not just like marriages and so on.

[816] Like it's genetic sorting in terms of what Dawkins called memes, his ideas being put into these bins of these little echo chambers and so on yeah and as that's the book that's think a worthwhile read for everyone in 1984 is good of course as well like if you're talking about you know dystopian novels of the future yes it's a slightly different view of the future right but i kind of like identify with brave the world a bit more uh yeah you know speaking of uh not a book but my favorite kind of uh dystopian science fiction is a movie called brazil which i don't know if you've heard of and i know i need to watch it but yeah because it's in Is it in English or no?

[817] It's an English movie, yeah.

[818] And it's a sort of like dystopian movie of authoritarian incompetence, right?

[819] It's like nothing really works very well, you know.

[820] The system is creaky, you know, but no one is kind of like willing to challenge it, you know, just things kind of amble along.

[821] It kind of strikes me as like a very plausible future of like, you know, of what authoritarian is it might look like.

[822] It's not like this, you know, super efficient evil dictatorship of 1984.

[823] It's just kind of like this badly functioning, you know, but it's status quo, so it just goes on.

[824] Yeah, that's one funny thing that stands out to me is in, what are this, authoritarian, dystopian stuff, or just basic, like, you know, if you look at the movie Contagion, it seems that in the movies, government is almost always exceptionally competent.

[825] like it's like used as a storytelling tool of like extreme competence like you know you use it whether it's good or evil but it's competent it's very interesting to think about where much more realistically is it's incompetence and that incompetence is itself has consequences that are difficult to predict like bureaucracy as a very boring way of being evil of just you know if you look at the show HBO show at Chernobyl it's a really good story of how bureaucracy you know uh leads to catastrophic events but not through any kind of evil in any one particular place but more just like the it's just the system kind of system distorting information as a that travels up the chain, that people unwilling to take responsibility for things, and it's just kind of like this laziness resulting in evil.

[826] There's a comedic version of this.

[827] I don't know if you've seen this movie.

[828] It's called The Death of Stalin.

[829] Yeah.

[830] I like that.

[831] I wish it wasn't so there's a movie called Inglorious Bastards about, you know, Hitler and World War, you know, so on.

[832] For some reason, those movies piss me off.

[833] I know a lot of people love them, but like I just feel like there's not enough good movies even about Hitler.

[834] There's good movies about the Holocaust but even Hitler there's a movie called Dawnfall that people should watch I think it's the last few days of Hitler that's a good movie turned into a meme but it's good but on Stalin I feel like I may be wrong on this but at least in the English -speaking world there's not good movies about the evil of Stalin.

[835] That's true.

[836] Let's try to see that.

[837] But I actually, so I agree with the Unanglorious Paster, I didn't love the movie because I felt like kind of the stylizing of it, right?

[838] The whole like Tarantino kind of Tarantinoism, if you will, kind of detracted from it and made it seem like unsirious a little bit.

[839] But Death of Stalin, I felt differently.

[840] Maybe it's because it's a comedy to begin with.

[841] This is not like I'm expecting, you know, seriousness.

[842] but it kind of depicted the absurdity of the whole situation in a way, right?

[843] I mean, it was funny, so maybe it doesn't make light of it, but it sounds like goes probably like this, right?

[844] Like a bunch of kind of people, they're like, oh, shit, right?

[845] You're right, but like the thing is it was so close to like what probably was reality.

[846] There was caricaturing reality to, where I think an observer might think that this is not like they might think it's a comedy in well in reality that this is that's the absurdity of how people act with dictators I mean that's I guess it was too close to reality for me yeah the kind of banality of like what were eventually like fairly evil acts right but like yeah they're they're just a bunch of people trying to survive and like i think there's a good i haven't watched you yet the good movie on uh the movie on churchill um with uh gary oldman i think is gary oldman i may be making that up but i think he won like he was nominated for an oscar something so i like i love these movies about these humans and Stalin like chernobo made me realize the the HBO show that there's not enough movies about Russia that capture that spirit.

[847] I'm sure it might be in Russian there is, but the fact that some British dude that like did comedy, I feel like he did like hangover or some shit like that.

[848] I don't know if you're familiar with the person who created Chernobyl, but he was just like some guy that doesn't know anything about Russia and he just went in and just studied it, like did a good job of creating it and then got it so accurate, like poetically and the facts that you need to get accurate he got accurate just the spirit of it down to like the bowls that pets use just the whole feel of it it was nice it's good yeah i saw this here i said yeah it's it's incredible it made me made me wish that somebody did a good like um 1930s um like starvation at stalin like leading up to world war two and in world war two itself like stalingrad and so on And like, I feel like that story needs to be told.

[849] Millions of people died.

[850] And to me, it's so much more fascinating than Hitler, because Hitler is like a caricature of evil almost, that it's so, especially with the Holocaust, it's so difficult to imagine that something like that is possible ever again.

[851] Stalin, to me, represents something that is possible.

[852] like the so interesting like the bureaucracy of it is so fascinating that it potentially might be happening in the world now like they were not aware of like with north korea another one that like there should be a good film on and like the possible things that could be happening in china with overreach of government i don't know there there's a lot of possibilities there i suppose yeah i i wonder how much you know i guess the archives should be maybe more open nowadays right i mean for a long time they just we didn't know right like anyways no one in the west knew for sure well there's a i don't know if you know there's a guy named stephen codkin he is a historian of Stalin that i spoke to on this podcast i'll speak to him again the guy knows his shit on stalin he like read everything and it's it's so fascinating to to talk to somebody like he knows Stalin better than stalin himself it's crazy like you have so he's i think he's at princeton he is basically his whole life is stalin studying stalin yeah it's great and in that context he also talks about and writes about putin a little bit i've also read at this point i think every biography of putin english of english biography of putin i need to read some russians obviously i'm mentally preparing for a possible conversation with putin so what what is your first question to putin when you have them on your on the podcast i it's interesting you bring that up the first of all i wouldn't tell you but i can't give it away now uh but i actually haven't even thought about that so my current approach and i do this with interviews often but obviously that's a special one but i try not to think about questions until last minute i'm trying to sort of get into the mindset that what and so that's why i'm soaking in a lot of stuff not thinking about questions just learning about the man but in terms of like human to human it's like i would say it's i don't know if you're a fan of mob movies but like the mafia which i am like good felts and so on he's much closer like mob morality which is like mob morality maybe i could see that but i like your approach anyways of this, um, the extreme empathy, right?

[853] It's a, a little bit like, you know, Hannibal, right?

[854] Like, if you ever watch the show Hannibal, right, they had that guy, um, um, um, you know, Hannibal, of course, like, uh, yeah, sounds and a lot, but there's a TV show as well, and they focused on this guy, Will Durant, uh, who's a character, like, extreme empath, right?

[855] So in the way he, like, catches all these killers, so he pretty much, uh, he can empathize with them, right?

[856] Like, you can understand why they're doing the things they're doing, right?

[857] It's a pretty excruciating thing, right?

[858] Because you're pretty much like spending half your time in the head of evil people, right?

[859] Yeah, but...

[860] I mean, I definitely try to do that with other, so you should do that in moderation.

[861] But I think it's a pretty safe place, safe place to be.

[862] One of the cool things with this podcast, and I don't know you didn't sign up to hear me listen to this bullshit, but...

[863] That was interesting.

[864] I in the what's his name Chris Latner who's a Google oh he's not Google anymore sci -fi he's one of the most legit engineers I talk with I talk with them again on this podcast and one of the he gives me private advice a lot and he said for this podcast I should like interview like I should widen the range of people because that gives you much more freedom to do stuff like so his idea which I think I agree with Chris is that you go to the extremes.

[865] You just cover every extreme base and then it gives you freedom to then go to the more nuanced conversations.

[866] It's kind of, I think there's a safe place for that.

[867] There's certainly a hunger for that nuanced conversation, I think, amongst people where, like, on social media, you get canceled for anything slightly tense, that there's a hunger to go full.

[868] Right, you go so far to the opposite side.

[869] And it's like to mystifies it a little bit, right?

[870] Yeah, that's...

[871] there is a person behind all of these things and that's the cool thing about podcasting like three four hour conversations that that it's very different than the clickbait journalism is like the opposite that there's a hunger for that there's a willingness for that yeah especially now i mean how many people do you even see face to face anymore right like this you know it's like not that many people like in my day to day aside from my own family that like i said across it's sad but it's also beautiful like i've gotten a chance to like like our conversation now there's somebody i guarantee you there's somebody in russia listening to this now like jogging there's somebody who is just smoke some weeds sit back on a couch and just like enjoying like i guarantee you that we'll write in the comments right now that yes i'm in st petersburg i'm in moscow i'm whatever and and we're in their head and they have a friendship with us and I'm the same way I'm a huge fan of podcasting it's a beautiful thing I mean it's a weird one -way human connection like yeah before I went on Joe Rogan and still I'm just a huge fan of his so it was like surreal we had I've been friends with Joe Rogan for 10 years but one way yeah from this way from the same Petersburg way yeah the San Francisco way and it's a real friendship I mean now it's like two way but it's still surreal.

[872] And that's a magic of podcasting.

[873] I'm not sure what to make of it.

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

[875] It's the audio that's magical.

[876] I don't know what to do with it, but it's people listen to three, four hours.

[877] Yeah, we evolved over millions of years, right, to be very fine -tuned to things like that, right?

[878] Well, expressions as well, of course, right?

[879] But back in the day on the, you know, on the savannah you have to be very attuned to whether you had a good relationship with the rest of your tribe or a very bad relationship right because you know if you had a very bad relationship you're probably going to be left behind and eaten by the lions yeah but it's weird that the tribe is different now like you could have a connection one -way connection with joe rogan as opposed to the tribe of your physical vicinity but that's why like you know it works with the podcasting but it's the opposite of what happens on Twitter right because all those nuances are removed right you're not connecting with the person yeah because you don't hear the voice you're connecting with like an abstraction right it's like some some stream of tweets right and it's very easy to assign to them any kind of like evil intent you know or dehumanize them which it's much harder to do when it's a real voice right because like you realize it's a real person behind the voice let me uh try this out on you i sometimes ask about the meaning of life.

[880] Do you, your father now, an engineer, you're building up a company, do you ever zoom out and think, like, what the hell is this whole thing for?

[881] Like, why are we descended to vapes even on this planet?

[882] What's the meaning of it all?

[883] That's a pretty big question.

[884] I think I don't allow myself to think about it too often, or maybe like life doesn't allow me to think about it too often.

[885] But in some ways, I guess, the meaning of life is kind of contributing to this kind of weird thing we call humanity, right?

[886] Like, it's, in a way, I can think of humanity as like a living and evolving organism, right?

[887] That, like, we all contributing in a way, but just by existing, by having our own unique set of desires and drives, right?

[888] And maybe that means, like, creating something great.

[889] And it's bringing up kids who, you know, are unique and different and seeing, like, you know, taking joy in what they do.

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

[891] I mean, if you're not a religious person, right, which I guess I'm not, that's the meaning of life.

[892] It's in the living and the creation.

[893] Yeah, there's something magical about that engine of creation.

[894] Like you said, programming.

[895] I would say, I mean, it's even just actually what you said with even just programs.

[896] I don't care if it's like some JavaScript thing on a button on the, the website it's like magical that you brought that to life i don't know what that is in there but that seems that's probably some version of recreation of uh like reproduction and sex whatever that's in evolution but like creating that html button has echoes of that feeling and it's magical uh right but i mean you if you're religious person maybe you could even say right like we were we were created in god's image right i like we were created in god's image right i like I mean, I guess part of that is the drive to create something ourselves, right?

[897] I mean, that's part of it.

[898] Yeah, that HTML button is the creation in God's...

[899] Yeah, so maybe hopefully it'll be something a little more...

[900] It's a dynamic, maybe bigger, some JavaScript.

[901] Yeah, maybe some JavaScript, some React and so on.

[902] But, no, I mean, I think that's what differentiates us from, you know, the apes, so to speak.

[903] Yeah, we did a pretty good job.

[904] It was an honor to talk to you.

[905] Thank you so much for being part of creating one of my favorite services and products.

[906] This is actually a little bit of an experiment.

[907] Allow me to sort of fanboy with some of the things I love.

[908] So thanks for wasting your time with me today.

[909] That was awesome.

[910] Thanks for having me on and giving me a chance to try this out.

[911] Awesome.

[912] Thanks for listening to this conversation with Dan Kokodov.

[913] And thank you to our sponsors.

[914] Athletic Greens, only one nutrition drink, Blinkist app that summarizes books, Business Wars podcast, and Cash App.

[915] So the choice is health, wisdom, or money.

[916] Choose wisely, my friends.

[917] And if you wish, click the sponsor links below to get a discount and to support this podcast.

[918] And now let me leave you with some words from Ludwig Wittgenstein.

[919] The limits of my language means the limits of my world.

[920] Thank you for listening and hope to see you next time.