E4: The AI Moment with Amjad Masad, Flo Crivello, Dan Romero, and Antonio Garcia Martinez

We're hiring across the board at Turpentine and for Erik's personal team on other projects he's incubating. He's hiring a Chief of Staff, EA, Head of Special Projects, Investment Associate, and more. For a list of JDs, check out: eriktorenberg.com. (0:00) Preview of the debate (2:23) Sponsor (3:10) Flo’s take on the AI moment and AGI (5:45) Amjad’s take on the AI moment (8:03) Antonio's skepticism about transhumanism (9:10) Human tendencies color the debate about agency (11:25) The inflection point is a jump in generality (13:14) AI's effect on coding (16:57) Weakness in current models (19:18) Bounties and the future of work (21:48) Antonio's skepticism around bounties (25:00) Who will capture value, incumbents or startups? (27:26) "All knowledge work will be automated in 20 years" (30:10) Who will be the biggest beneficiaries of this shift? (31:50) What AI companies should VCs invest in? (32:26) Sponsor SPONSORS: SHOPIFY | OMNEKY Shopify is the global commerce platform that helps you sell at every stage of your business. Shopify powers 10% of ALL eCommerce in the US. And Shopify's the global force behind Allbirds, Rothy's, and Brooklinen, and 1,000,000s of other entrepreneurs across 175 countries. From their all-in-one e-commerce platform, to their in-person POS system – wherever and whatever you're selling, Shopify's got you covered. With free Shopify Magic, sell more with less effort by whipping up captivating content that converts – from blog posts to product descriptions using AI. Sign up for $1/month trial period: https://shopify.com/cognitive Thank you Omneky for sponsoring The Cognitive Revolution. Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work, customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data, to generate personalized experiences at scale. Twitter: @CogRev_Podcast @amasad (Amjad) @antoniogm (Antonio) @altimor (Flo) @dwr (Dan) @eriktorenberg (Erik) @labenz (Nathan) Subscribe to our Substack: https://cognitiverevolution.substack.com/ Websites: cognitivervolution.ai replit.com teamflowhq.com omneky.com Show Notes & references: Eliezer Yudkowsky: https://www.lesswrong.com/posts/FKNtgZrGYwgsz3nHT/bankless-podcast-159-we-re-all-gonna-die-with-eliezer Paul Christiano: https://paulfchristiano.com/ AI Feedback: https://www.anthropic.com/constitutional.pdf RECOMMENDED PODCAST: The HR industry is at a crossroads. What will it take to construct the next generation of incredible businesses – and where can people leaders have the most business impact? Hosts Nolan Church and Kelli Dragovich have been through it all, the highs and the lows – IPOs, layoffs, executive turnover, board meetings, culture changes, and more. With a lineup of industry vets and experts, Nolan and Kelli break down the nitty-gritty details, trade offs, and dynamics of constructing high performing companies. Through unfiltered conversations that can only happen between seasoned practitioners, Kelli and Nolan dive deep into the kind of leadership-level strategy that often happens behind closed doors. Check out the first episode with the architect of Netflix’s culture deck Patty McCord. https://link.chtbl.com/hrheretics This show is produced by Turpentine: a network of podcasts, newsletters, and more, covering technology, business, and culture — all from the perspective of industry insiders and experts. We’re launching new shows every week, and we’re looking for industry-leading sponsors — if you think that might be you and your company, email us at [email protected].

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Nathan Labenz?

  Today, we're sharing a special episode from the Moment of Zen podcast, which Eric co-hosts with Antonio Garcia Martinez, founder of web three growth metrics company Spindel, and Dan Romero, founder of Farcaster, a sufficiently decentralized social network. The guests for this episode were Amjad Masad, CEO of Replit, a leader in AI coding assistance with its ghostwriter and ghostwriter chat products, Blo Krevelo, founder of Teamflow, and myself. We cut the original hour and a half discussion down to the parts that focus exclusively on AI. So in the next 30 minutes, you'll hear a mix of visionary and skeptical takes on short and midterm AI impacts. From the potential rise of the 1000x developer

• 0:44Nathan Labenz?

  to the question of what makes this AI moment different from previous hype cycles, to analysis of which types of AI companies will prove enduring. This is a fast-paced, wide-ranging discussion among some very smart people, all of whom are grappling with AI developments in real time. Enjoy.

• 1:03Speaker 01

  The level of exponential improvement is so tight that like you could go to lunch and come back and the world has changed, right? So that's a singularity because you can't know what's next. And then the sort of the laser branch of that view is that is that the most likely outcome is death of humanity. And the reason that's most likely outcome is because it is impossible to align a enormous computing force that at the same time is sort of dumb. It doesn't understand sort of human preferences.

• 1:46Speaker 01

  And therefore, any goal that you give it is not going to be specific enough. And there are a lot of potential sort of attribution or explanation of that goal that that gets you in trouble.

• 2:01Speaker 02

  Yeah, I call bullshit. I think they're completely fucking full of it. Sorry. You know how like in a lot of the like the sci-fi apocalypse literature that appeals to nerds? Like somehow, when literally the fucking world ends and there's no law and order, it's Mad Max, the guys who make the computers work somehow end up running the show? This is an expression of that fantasy.

• 2:23Speaker 03

  The Cognitive Revolution podcast is supported by Omnikie. Omnikie is an omni-channel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work, customized across all platforms with the click of a button. Omnikie combines generative AI and real-time advertising data to generate personalized experiences at scale.

• 2:41Speaker 04

  I think we're going to see a great convergence of a lot of different AI advances that suddenly all come together and really make a tremendous impact on modern life. And so calling it the cognitive revolution, because I think the changes that we'll see are going to be every bit as transformative as what we saw, we weren't here to see it, but as our ancestors saw in the agricultural revolution and of course, the industrial revolution.

• 3:08Flo Crivello?

  Yeah, I totally agree that everything Nathan said. I was grabbing dinner last night with an aristotle who described what we're going through as what he calls the sealed great convergence. So he says, you know, the first one was 2012 with like the image that moment and deep learning and all of AI turned into that. The second one was transformers with attention is always need deeper in 2017. And then we're going through the third one since you're roughly 2022. You could think of those convergences that's happening once every five years with large language models. So large language models, your chat GPT is the most famous example. They're really good at manipulating languages, like really, really, really good. It just turns out that they're so good at it that the moment you model out a task as a language task, which turns out can model a lot of tasks with a language task, it becomes really good at that as well.

• 3:53Flo Crivello?

  So we're starting to use large language models for things ranging from, obviously you can ask it questions, but it can also do a little bit of math. You know, we're starting to use it as a search engine. We're starting to use it to code. We're starting to use it in robotics as a sort of reasoning engine. And so I think that like that alone makes AI dramatically undermined. And I say that knowing full well how hype it is. I think even if we stop the progress and the discoveries that we've made right now, which we're not stopping, they're going exponential. Even if we stop them right now, all of civilization, I think is going to be dramatically impacted in the next 10 years. And then I think zooming out even further. Yeah, I am more and more a believer of the AGI moment. My timelines are compressing rapidly, meaning I actually believe that AGI is going to happen sooner and sooner.

• 4:37Flo Crivello?

  And my concerns are slowly increasing. So as a reminder of what AGI is, that recursive loop of self-improvement, AGI becomes better and better. And so it ends up with the net zero one trillion. I think the steel man for it is that, you know, we, and the reason why people are so worried is that Moesla means that there won't be just one AGI. Moesla means there will be a lot of AGIs. And so in the limit, anyone with a laptop has an AGI. And now the risk that Yudkowsky talks about is that it's impossible to align a single AGI, let alone a million AGIs. So we need to align it, we need to do that impossible thing a million times. Anyway, that's not to say that I think like we should like burn all the GPUs or whatnot, which I think is what Yudkowsky may prescribe.

• 5:22Flo Crivello?

  I just think it is something that is under discussed. I think it is a real risk. And I just wish there was a lot more funding and attention brought to this issue.

• 5:30Speaker 01

  Yeah, so the steel man of the flow side of the argument, which is essentially like, I think Eliezer Yudkowsky is like the big, like the major sort of authority there. And then a lot of, now there's another guy, Paul Cristiano, who's like an offshoot of Les Ron, or the Yudkowsky. But that sort of branch of AGI is going to kill us all is what started by Eliezer. And the sort of main argument there is that if you accept the secular view that you know, humans have these meat computers, then there's no fundamental physical law

• 6:11Speaker 01

  that says you can't build this meat computers in, in, in sort of chewing machines, right? So if you accept that, then you also accept that at some point we're going to have human level AI. If you accept that, that we're going to have human level AI, you also have to accept that there's going to be an AI explosion. So an AI takeoff event is basically when the AI comes online and trains the next generation of AI. And that next generation of AI creates the next generation of AI. And that could start like a slow process over a year or two. Actually, like, you know, people at OpenAI are already using GPT-4 to train GPT-5, right? And so that's already in some way happening.

• 6:57Speaker 01

  That will shorten to milliseconds at some point. And that's the Ray Kurzweil's singularity, right? Where the level of exponential improvement is so tight that like you could go to lunch and come back and the world has changed. Right? So that's the singularity because you can't know what's next. And then the sort of the laser branch of that view is that is that the most likely outcome is death of humanity. And the reason that's most likely outcome is because it is impossible to align a enormous

• 7:40Speaker 01

  computing force that at the same time is sort of dumb. It doesn't understand sort of human preferences. And therefore, any goal that you give it is not going to be specific enough. And there are a lot of potential sort of attribution or explanation of that goal that that gets you in trouble.

• 8:03Speaker 02

  Yeah, I call bullshit. I think they're completely fucking full of it. Sorry. You know how like in a lot of the like the sci fi apocalypse literature that appeals to nerds? Like somehow when literally the fucking world ends and there's no law and order and it's Mad Max, the guys who make the computers work somehow end up running the show. This is an expression of that fantasy, right? That ne- never mind like the weird Western obsession. Like the original robot was coined in a, in a, in a play in which robots took over or the Gollum legend inside Jewish folklore in which humanity creates a thing and that thing rebels. There's always been this deep latent thing that that's going to happen. But I think it's basically bullshit. And not only that, the transhumanism is also bullshit. You literally take any of little Kurzweil's little spiel's and like global reg X like replace singularity with rapture and you get an evangelical sermon. It's basically Christian eschatology expressed in scientific form. They're just not aware of it because they don't actually read any religion.

• 8:49Speaker 02

  That, that's what I think it is. And I, I don't think it's ever going to happen. I, I don't doubt that the house cat theory though, i.e. that like H G Wells time machine thing where they have the Eloi and the Morlocks. And it turns out we just become Eloi in which we're living in the service of it. And the Morlocks are like machines or something else. Maybe that happens. I think that's already happened in the sense that we get worked up about these Twitter fights over nothing in which nothing happens.

• 9:10Speaker 01

  I think fundamentally like people ascribe. Agency to to things like, you know, anyone who has kids knows that like one of the first things that kids do when they grow cognitively is they try, they give um and names and they give personalities to their toys. Or even to like simple things like boxes or whatever. Humans have this tendency to ascribe agency. Uh I I think people just extrapolate from this idea that we see some kind of glimpses of agency in these things.

• 9:47Speaker 01

  To the fact that these things can't formulate abstract goals and desires and go execute on them, which I don't think is entirely true, right? So the AI is ultimately a tool for humans to do things in the world, right? It's another, like I I think of L L M's are uh as another computer. That's how I kind of build on top of L L M's. We're doing a lot of things with L L M's at Rappler. It's my mental model for it. It's it's it's it's a computer, right? And it's it's a very powerful new type of computer. Let's have the intellectual honesty to say that we don't understand what consciousness is. And we don't understand the thing that really gives us agency.

• 10:29Speaker 01

  But I think it's just a tool as as it were right now. I think we need some kind of science of agency of consciousness before we're able to say that we can build these things to recruit enormous amount of value of the world requires a lot of planning, a lot of emotions. Well, just think about being an entrepreneur. You've done it multiple times. Uh just like how much convincing you have to do. Uh you have to have some kind of theory of mind. You have to think what other people think. And I'm not sure we can build that just yet. So no AI girlfriends, eh like you don't think that's a

• 11:05Speaker 01

  It will look like it. It might act like it. It might fool some people. Chat G P T is already fooling some people. But that doesn't make it it and that doesn't get it to the level of power that kills us all. With the transformer model broad is a kind of generality. So in a way, there's a generality jump here. I I think if you if you go back to the early era of AI, they would they would probably think that the large language models that we have today are some kind of AGI. Because, you know, it's sort of this boiling frogman phenomenon where like slowly we're increasing generality.

• 11:47Speaker 01

  There's not going to be any point of time where we're going to say, ha, this is AGI. I I I think, you know, every jump is going to be like a significant jump in generality, but it's still going to feel somewhat slow. So this this creates a sort of a rising tide and, you know, makes everyone more productive, makes software a lot easier to create. Creating software used to mean learning all sorts of arcane knowledge. And now you just have to write English to create a piece of software. You can create a meaningful piece of software. So now you take something that was the capability of expert software engineers and you give it to everyone in the world.

• 12:33Speaker 01

  And I think the impact of that is going to be A hugely deflationary, B is going to like give people new superpowers. I think new type of entrepreneurship is on the rise. And we're seeing a lot of people that would have required companies and armies of people around them to build something useful are able to do it on their own. And I think we're going to see a new crop of winners and entrepreneurs and and sort of millionaires and billionaires coming out of this phenomenon. And and and it's also like a fundamentally new way of of working and and automating things. So would would you teach your kids to code? Yeah, I would teach my kids to code and the reason is because code is still going to be super relevant.

• 13:19Speaker 01

  We have ways to generate more accurate code. We're gonna, we're coming up with a chat G P T like thing inside Replit. And that uses a larger model and it'll give you more accurate code, but still the accuracy is not gonna be a hundred percent. So you need to learn how to debug that code. I think surprisingly, most programmers will spend most of their time reading and understanding code. Um, that's gonna put pressure on tooling for, for debugging and comprehension. And L L M's will help there. L L M's can explain code for you. But I think there's gonna be more innovation and visualizing code. You know, other ways to debug and comprehend code.

• 14:02Speaker 01

  On the front end side of things, you're gonna see super productive front end coders that are heavily powered by code generation. Like, it might be the case that they don't actually code, they're actually just like plumbing things together. They're talking to a lot of different L L M's. They're, uh, they're acting more like project managers and product managers and actual programmers. On a recent podcast, I called this the Steve Jobs black pill. In the early vision of, of, of computing, uh, we thought that everyone was gonna be a programmer, right? There wasn't this user programmer dichotomy.

• 14:42Speaker 01

  And then, like, you, Steve Jobs popularized end user, the idea of end users with user interfaces, really lovely user interfaces. And that became the dominant thing and most people are consumers of software as opposed to the creators. I think the idea of software creation will come back. I think there's gonna be a lot more people wanting to create, uh, you know, personal software and software for their business use case. And there's gonna be a lot of end user programming. Those people were not, will not have to read code. Because the, the level of accuracy needed is probably not a hundred percent. Uh, optimization needed is not like, performance needs is not that high. A lot of the code is gonna be throwaway code.

• 15:24Speaker 01

  And so if you're like a, you know, professional, you, you probably don't have to learn the code. Uh. But if you want to be sort of a low level programmer, I think that's still going to be relevant. Or if you want to be a front-end engineer, you you need to have some knowledge of code, but maybe not that deep.

• 15:43Speaker 06

  It feels like it's it's going to make the ten X engineer, a hundred X engineer, in the sense that there's still a base level of you need to be able to guide it in the right direction. It's a tool. And so for the person who already has the skill level or the talent, it's going to be a, um, you know, increasing the advantage. Whereas for maybe the mediocre engineer, it doesn't really make a huge difference. And or maybe they just lose their job completely. Like, I'd be curious how you guys think about that.

• 16:11Flo Crivello?

  Think of it as a, as a, as a rising tide lifting all boats. So I agree. I think like the ten X engineer is going to become a hundred X engineer. The one X engineer might become a ten X engineer. And I think to answer your question earlier, AGM about is that a new step change? I think folks who were not engineers before will become engineers and be able to perform simple tasks at first and more and more complicated ones.

• 16:29Speaker 06

  I don't think this is like a replacement dynamic, but it's also really worth looking into Anthropic's recent publication that they're calling Constitution AI. And it's kind of the next generation of reinforcement learning from human feedback. They're now doing reinforcement learning from AI feedback.

• 16:47Flo Crivello?

  I think it's interesting also figuring out, I'm not sure whether AI is going to replace the code. Like, we're gonna run models instead of running code or whether it's going to write the code. Um, I I do agree with Amjad that these models have a weakness right now, which is that they they suck at systematic thinking. So for example, you can give them, uh, this, uh, task of, hey, pass this date instead of having it go a day, month, year, have it go month, day, year, right? And if you give it enough examples, GPT-3 will succeed at this task something like ninety nine point eight percent of the time. But it will fail point two percent of the time, um, which shows that it is, it is succeeding enough. It kind of understands the task, but it's not thinking about the task in a systematic way. So the other weakness of these models is that they cost a fortune to run, like they're very, very expensive.

• 17:32Flo Crivello?

  And here again, most of life is going to be on your side. It's always going to be more expensive to run, to to use a large language model to, uh, pass a date than it is to run a piece of code. So I I I think like in this very interesting time right now, where like nobody really knows how these things will shake out. But I think that again, these AIs will learn how to use tools. And I think that on some level, they will be smart enough to discriminate to understand, hey. Is this a task that I expect to perform a lot? Is this a task that I am performing a lot right now? And is this a task that benefits from more systematic thinking? If so, I'm going to write a piece of code to perform this task for me because it's going to be cheaper and more reliable and more systematic.

• 18:08Speaker 06

  And you think that the model will know that? Or is it the human guides it to saying, okay, this is not a good use of the model?

• 18:14Flo Crivello?

  I think there are ways to build systems that are either part of the model or upstream of the model such that the model would, in a way, know that.

• 18:21Speaker 01

  I actually think that the sort of engineering is lagging way behind of the capability. It's just, it's kind of frustrating to me because like I'm working on Rappel, working really hard to build some LLM into Rappel capability. So we have the GoShire product. But I feel like we're all scratching the surface. Like, there's so much to do. I think total usage is possible today. It's possible to build a chat GPT that has a Python interpreter on the side, that has a search engine on the other side, but nobody's really built it. So I think it is possible today to ask it a question like, book me a flight and it going to kind of Google whatever,

• 19:05Speaker 01

  and then maybe writing a program, hitting an API and booking you a flight. I think that's totally possible today. Bounties and the reason why we started to keep at the world this way is that, I'm actually like, sort of a crypto canon, but the sovereign individual had a description in it that talks about the sort of the future of work in a way. And it talks about how AI, crypto, sort of the future of the internet would support this sort of world where people are less full-time employed. They're more like freelancers. They're able to jump from work to work. They're able to construct companies on the fly and kind of dissolve them right after the work is done.

• 19:49Speaker 01

  And that's been the picture in my mind for a long time. And I think for the first time, it's really possible. All these technologies are maturing in a way that allows this new crop of entrepreneurs to be able to be hyperproductive and be able to get things done super quickly and super cheaply. You know, when we talk to younger programmers,

• 20:18Speaker 01

  Like almost without fail, their ambitions no longer to join Microsoft or Facebook or whatever. They they wanna build businesses, they wanna make money, they wanna go into freelancing, they wanna be like this free spirit. They wanna build a career that is like freedom maximizing. And I I think having the ability to use sort of an army of AI assistants in being really powered, supercharged by this technology will give people amazing opportunities in the future.

• 20:59Flo Crivello?

  Yeah, I I think that just like the PC was a tool of a sort of great individual empowerment. I think you're right that AI is going to give even more labor rates to the individual. I think like code let companies like WhatsApp sell for like 20 billion with like, I think like 40 or 50 people. I think you're right. I think we're going to see 1 billion, 10 billion, maybe 1 trillion dollar company in, you know, a decade or so. With perhaps one or two people. I totally agree with that. Man, if you pay software engineers by the ticket, I think you would see a lot fewer rest investors and you would actually see 10x engineers making 10x the salary. I think if you really believe in the 10x engineer, which I do, I think you should see software engineers make 10 million dollars a year at at big companies and you don't see that.

• 21:38Flo Crivello?

  And conversely, you see people whom I call professional coffee bar builders, who just rest invest and who make 500K a year package. So I'm very excited by the potential here.

• 21:48Speaker 02

  I think that's totally not gonna work, by the way. By the way, the bounty system is one of the parts of Web 3 that totally doesn't fucking work. There's entire companies that are based around bounties and those if for companies that are based around paying someone a bounty to do a thing are always like the ones that you have to like route around and somehow use the product without it. I mean, think about it, do you pay anybody in your company in bounties? No. Would you, would a 10x engineer who actually wants to make like generational wealth? We're talking 100 million dollar exits, but they sit there and do basically mechanical turks for coding all day? Or were they actually, I mean, you're you're creating an arbitrary binary duality between rest investors and people working like coding chipmunks on on bounties, right? But the reality is that most coders who make lots of wealth, it's neither one nor the other, right? It's people who actually do work their asses off, but it's with some sort of committed product in which they have an overarching design ethos.

• 22:33Speaker 01

  Google needs one hundred and twenty thousand employees and most of them are just like making lattes every day. They're really good at making lattes, but at a drop of the hat, they would need them and they would find them there. And there's very low friction on using their labor and their talents, right? So it's a very rational behavior to sort of hoard talents. It's sort of like the billionaire that has all these assistants. You're like, oh, they're just sitting around doing nothing. But at the drop of the hat, when something important happens, he needs all that labor and he can't really hire it from the market at large. So anytime technology reduces the cost of going to the market, you see us going to the market more.

• 23:15Speaker 01

  Again, people had a lot more drivers and personal drivers before Uber came around. And now everyone has access to the market at a very low transaction cost. And that's the same across the board, like anything that you use today that you find, you know, very useful, like DoorDash. People used to have people that worked with them for them, like people had servants that were around all day. Just waiting for that one order of the day to go get that. And now you're able to go to the market and get that labor on demand on the fly, right? So I think bounty type systems, and I think crypto as well, could allow us to work on some coordination problems

• 23:59Speaker 01

  in order to solve the transaction cost. So you can have like, you know, hundreds of coders at any given point working for you so that you can focus on building like a billion dollar idea, right? So some way to, you know, pay people for their contribution, I think is the ethical thing to do. Like the idea that, you know, being able to pay people for the data that they create is not a bad idea. Like, you know, if we can trace the Wikipedia contributions on a character by or token by token level that gets fed into GPT and a hypothetical company in the future that wants to do the right thing

• 24:43Speaker 01

  would assign some kind of hypothetical value per token. And they can just like pay out some revenue share to the authors of the of the data that they were trained on. Or maybe the data that gets used in production.

• 24:59Speaker 02

  People listening to this who want to join AI companies or want to invest in AI companies, what are the kinds of AI companies that are going to be enduring versus the kinds of companies that are going to be commodities and not capture a lot of value?

• 25:11Flo Crivello?

  I view three categories of companies that are created right now, right? There's like what they call like big model, right? So it's like ads for big open AI that creates these giant models. Then there's the application AI on top of those and the application AI is going to be made up. I'll just arbitrarily slice it. There's like horizontal applications and vertical applications. The mode for the large language models is going to be economies of scale, meaning it costs a fortune to build a large language models. It's costing more and more money. You know, I think GPT four is going to cost of the order of call it a hundred mil to train between the resource class and the compute. It is very, very expensive, which is why OpenAI is raising all that money. Plus the label to train it using RL itself, as I'm said described.

• 25:55Flo Crivello?

  So it's just very expensive to train, but that's a one type cost. And then the inference cost, meaning the cost that it costs to run the AI once you've traded is many orders of banking to dollar. So again, if you think of a hundred million dollars to train the AI, running it is more on the order of one cent.

• 26:13Speaker 02

  I mean, you're right, Flo. We've already saw that with Google, for example, right? I mean, the reason why Google is as good as it is and not because their AI is so amazing, although no doubt it is, because they have literally the entire world typing what they want, right? It was always the data side. It was never the actual algorithm. Algorithms aren't particularly defensible, right? At the end of the day, AI isn't a product anymore than literal regression is, right? You actually have to apply it to something to create something that someone will pay you for, right? And that's where I kind of don't quite get a little bit of the AI hype, right? It's like, like, where is the actual product? If Google can't actually find a way to turn that into actual cash, and a lot of these companies are actually basically thinly veiled, just prompt your eyes on chat GPT, what is the actual product?

• 26:50Flo Crivello?

  The product is not expensive. Production of text and production of music and production of images has not been where the value lies. You can hire folks on five year to do this thing. You're not five, but like you can hire folks to do these things just fine and you'll still not gonna go anywhere, right? So the value is in the distribution. So I think that if generally AI is going to change anything to the structure of the industry is that anytime you make anything cheaper, its complements become more valuable. So here we're making content cheaper, not that it was really necessary because it was all pretty cheap. And so the complement to content, which is distribution, is going to become more valuable. So I actually think that this is going to be great for TikTok and YouTube and all of that. All knowledge work will be automated in the next 20 years. And so I think, again, code is the first one that was being automated.

• 27:35Flo Crivello?

  That's, I think, huge. I think we're going to see a lot more come out in the next two years, which are going to see action-oriented knowledge work that actually does stuff. I think it's going to be automated pretty soon. I expect support, I think, is going to be pretty big one. When you think about it, all the knowledge work is, is it's a function that sits in between a keyboard and a monitor. And I think AI is going to be really good at approximating that function.

• 27:59Speaker 02

  But just to be clear, you're saying that all knowledge will be replaced in 20 years? We've been hearing that for 20 years. And in fact, if there's any bit of skepticism around AI, like, you know, AI being at the, like, the very pivot of, like, changing everything, I at least have been hearing that since I've been in tech. And if you go back even further back, Marvin Minsky in the 60s and 70s has been saying the same thing, which is why I was sort of trying to quantify what is the change that we're seeing? Because, again, there's clearly a trend line, like, no doubt, obviously, AI and automation have changed lots of things. I'm not saying it hasn't been a big deal. I'm just trying to understand if we actually are at some sort of clear, like, real inflection point, or are we continuing along the same trend line? Which already, by the way, I think is a big deal, right? Like, I mean, what I cited, right? Going from the Kernigan and Richies of the world to Replit, there's been, there's been a big change, right? Like, I'm not trying to underwhelm it, but it's, it's, it's not quite AGI, Terminator, end of the world levels of change.

• 28:44Speaker 04

  When something is not able to do a task, it just can't do the task. But we're now hitting that moment where, across very broad sets of tasks, the best AIs are outperforming the average human. They're not yet at the level of the expert human at any specific domain. They can outperform employed college grads on very wide distributions of tasks to a big bench. Big bench is a big benchmark that, that they compared poem to this, like, pool of, you know, kind of... Software QA testers essentially and AI won.

• 29:24Speaker 02

  I don't know, I like this is the same discord I heard about when Big Blue defeated, you know, Kasparov at chess. I mean, it's the exact same feeling. I mean, in times like this, right? I think a little, I think a lot of this is actually intellectual narcissism, the part of humans who are assigning a certain anthropic value to computers thinking. And I think of the famous Dykstra quote, right? Again, we've been debating these questions for decades before any of us were alive, right? He was, and he commented that asking whether a computer can think is like asking whether a submarine can swim, right? It doesn't matter. The point is that it does 40 knots and a human does one knot, right? And so along that particular dimension, which is moving through the water, the computer actually does do better than the human. And when it comes to ranking ads or, you know, filtering through stack overflow and coming up with a net result, clearly the computer does better, but that's very different than

• 30:07Speaker 02

  saying that it's more intelligent and this causes a crisis in the knowledge economy.

• 30:11Speaker 01

  The generality of LLMs can't be understated. Like for the first time, they can learn your intent from just one example. I can build Google Translate, Antonio, in three seconds. I'll just say, English, hello, French, salut. And then I can say, English, home, and that'll continue the French word. It's sort of like, that's pretty freaking amazing. That hasn't happened before. In any time we have a jump in generality, like the last time was the Turing Machine, is sort of an enormous thing. I think every app will get better. And I think actually

• 30:56Speaker 01

  the biggest beneficiaries of this shift is going to be sort of growth stage startups. And the reason is because you've got distribution, but Flo talked about distribution. So Notion adding AI is much more interesting than a new sort of knowledge management that's like based on, that's like AI first. So it's called bearish on AI first companies. Usually bullish on AI infrastructure companies. So OpenAI, or anyone really building DevTools around prompts, prompt IDEs, building fine tuning sort of technology, anyone who's, who wants, who are making it

• 31:41Speaker 01

  easier to build with transformer models, that's going to be something that every company will buy. Because again, I think it's a diffuse technology similar to cloud. Like every company will integrate this kind of technology into their. So you know, the financial sort of answer to your question is that if you want to invest in it, probably do do another Fang type strategy. Cause I think there's another set of S curve there. Plus maybe invest in infrastructure like Nvidia, whatever. Then on the startup side, DevTools, AI intelligence layer like OpenAI. Those are sort of the beneficiaries. And maybe the growth stage startups, early growth stage like us or or Notion or things like that will benefit all off of this.

• 32:24Speaker 03

  The Cognitive Revolution podcast is supported by Omnikie. Omnikie is an omni-channel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work, customized across all platforms with the click of a button. Omnikie combines generative AI and real-time advertising data to generate personalized experiences at scale.