E21: VC Insights on Investing in Artificial Intelligence with Sarah Guo and Elad Gil of No Priors Podcast

Nathan Labenz and Erik Torenberg sit down with Sarah Guo and Elad Gil, notable investors and co-hosts of the AI-focused No Priors podcast. They discuss how Sarah and Elad are approaching AI investment opportunities right now, how that differs from how they've thought about investing in the past, where in the stack from hardware to applications they expect to see value accrue, what modes of human-AI interaction they are most interested in, and more. Sarah is the founder of $100M AI-focused venture fund Conviction VC, which she launched last fall. She was previously General Partner at Greylock. Elad is a serial entrepreneur and a startup investor. He has invested in over 40 companies now worth $1B or more each, and is also author of the High Growth Handbook. This episode is the first in a series centered on talking to rising voices in AI media, people who are now only working overtime to understand everything going on in AI, but also creating thought leadership and educational content meant to help others get up to speed as well. LINKS: No Priors Podcast on Spotify: https://open.spotify.com/show/0O65xhqvGVhpgdIrrdlEYk No Priors on Apple Podcast: https://podcasts.apple.com/us/podcast/no-priors-artificial-intelligence-machine-learning/id1668002688 Elad Gil's blog: https://blog.eladgil.com/ Sarah Guo's blog: https://sarahguo.com/blog -- 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. TIMESTAMPS: (00:00) Episode preview (04:43) What is software 3.0 (09:14) Disruption coming from startups or incumbents? (13:42) Sarah and Elad identify overlooked investment opportunities in AI (15:19) Sponsor: Omneky (15:46) Future of social media (22:45) AI agents & personal co-pilots (25:32) Where to invest in AI? (31:11) How our kids will interact with AI (34:50) How to gain conviction as an investor in AI (45:07) When should founders raise money and when should they bootstrap? (46:28) How should startups spend their capital now that we have AI capabilities? (48:10) Sarah & Elad’s favorite products in AI (51:39) Would Sarah & Elad get a neuralink implant? (53:41) AI hopes and fears TWITTER: @CogRev_Podcast @labenz (Nathan) @eriktorenberg (Erik) @sarahnormous (Sarah) @eladgil (Elad) SPONSORS: 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. Mention "Cog Rev" for 10% off. 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]. More show notes and reading material released in our Substack: https://cognitiverevolution.substack.com/

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

• 0:00Speaker 00

  And I started digging around. I tried to find people to build an OpenAI competitor. And I couldn't convince anybody to do it. Everybody said, well, it's not that interesting of a business. And you know, are these APIs that good and all this other stuff? And I pitched person after person and nobody was willing to try it.

• 0:15Speaker 01

  Like within consumer, character is one of the most interesting companies. Replica is one of the most interesting companies. A lot of people don't like this, even though you see like a decade or more than a decade for a lot of looking at consumer company metrics. You're like, shit, right? Like, I'm going to pay attention if people are spending hours a day on this service because it is so rare.

• 0:35Speaker 00

  You know, there's base models and there's based models with a D, right? Like, what kind of model do you want your kid to interact with? And what do you want them to learn over time? And you know, how does that, how does that get selected? And who, who adjudicates what that selection process is? Or what's the, the ethical framework based on your location around the world that should be applied or shouldn't be applied? And so I think there's lots and lots and lots of interesting questions here.

• 0:58Nathan Labenz?

  Hello and welcome to The Cognitive Revolution, where we interview visionary researchers, entrepreneurs and builders working on the frontier of artificial intelligence. Each week we'll explore their revolutionary ideas and together we'll build a picture of how AI technology will transform work, life and society in the coming years. I'm Nathan Labenz, joined by my co-host, Erik Torenberg.

• 1:21Erik Torenberg?

  Before we dive into the cognitive revolution, I want to tell you about my new interview show, Upstream. Upstream is where I go deeper with some of the world's most interesting thinkers to map the constellation of ideas that matter. On the first season of Upstream, you'll hear from Marc Andreessen, David Sacks, Balaji, Ezra Klein, Joe Lonsdale and more. Make sure to subscribe and check out the first episode with A16Z's Marc Andreessen. The link is in the description.

• 1:49Nathan Labenz?

  Hi, everyone, and welcome back to The Cognitive Revolution. After taking a deep dive into prompting, process automation and jailbreaking over several recent episodes, we're now zooming out of it and talking to some fellow AI scouts. These are people who are not only working overtime to understand everything that's going on in AI, but also creating thought leadership and educational content meant to help others get up to speed as well. Today, our guests are investors Sarah Goh, Andy Lott Gil, co-hosts of the AI-focused No Priors podcast.

• 2:24Nathan Labenz?

  Sarah was previously a partner at Greylock and is now the founder of the $100 million AI-focused venture fund Conviction VC, which she launched last fall. She blogs on her website, saragoa.com. Elad Gil is a notable angel investor with investments that include Airbnb, Coinbase, Figma, Square, Stripe, and many more, including recent AI companies such as Character AI, Harvey AI, and Perplexity AI, whose CEO, Aravind Srinivas, you may remember, was a guest on The Cognitive Revolution back in episode seven. He also blogs at blog.eladgil.com.

• 3:03Nathan Labenz?

  We spoke about how they are approaching AI investment opportunities right now, how that does or doesn't differ from how they've thought about investing in the past, where in the stack, from hardware to applications, do they expect to see the most value accrue, what modes of human-AI interaction they're most interested in developing, and plenty, plenty more. I hope you enjoyed this conversation with Sarah Goa and Elad Gil. Sarah Goa and Elad Gil, welcome to The Cognitive Revolution.

• 3:31Speaker 01

  Thanks for having us.

• 3:32Nathan Labenz?

  Yeah, very excited to have this conversation with you guys. We're in such a moment right now of just the whole world kind of turning attention to AI, and that's something that, you know, I think we're probably four or five months into since the release of ChatGPT. You guys have each been thinking about AI very seriously, and both independently and together, I think, well before that. So I thought we maybe just start by kind of revisiting a couple of things that you guys had published about six months ago, and then ask you to just kind of take us through this period of time where, you know, new models are being released, new tools, new paradigms, just, you know, attention piling in, investment, you know,

• 4:17Nathan Labenz?

  dollars, I'm sure, piling in from all directions as well. And then we'll take it from there. So Sarah, starting with you, you announced about six months ago, this $100 million conviction fund where you are investing in software 3.0. So I thought for just a start, can you set a foundation for us and tell us how you think about software 3.0 and what that means?

• 4:43Speaker 01

  Yeah, I think it's shorthand for just believing that there's a very unexpected new set of software businesses emerging that can be very important, right? So everybody knows, you have this exponential creation of capabilities in machine learning. I remember a lot and I actually had a debate like six plus months ago, like, hey, there've been a lot of ML-first companies that haven't worked in the past and what changes? But I think the exponential creation of new capabilities is the thing that got me really excited for the fund. And when we think about what's so different about the software, it's what it does, right? I think it attacks lots of categories of services or areas that were not

• 5:28Speaker 01

  like big software markets before, copywriting, illustration, law, right? What the companies look like. So this could be, how many people does it take to create a $1 or $10 billion company? Like now we have empirical proof, like 20, right? And that was not true in previous generations of software. And then I think one that is still unexplored is just like how the product should work from a UX and a human interaction perspective. I think we're going to get a lot more than just the single chat.

• 6:00Nathan Labenz?

  Yeah, I definitely look forward to unpacking that because I just see so many different possibilities and it feels like we're so early in kind of exploring what the modes of human AI interaction are going to be. One thing that jumped out to me about your announcement was just, there's the AI focus, but then beyond that, just like extremely broad investment thesis, right? Totally up and down the stack, all the different verticals. How has, if at all, how has your thinking evolved in terms of like which parts of the stack or which verticals are kind of most interesting over the last six months?

• 6:37Speaker 01

  Some of the areas that are really attractive from a demand, like a value for end users or for customers' perspective are like obvious in hindsight, but somewhat unexpected. Right? And so anybody who has, you know, heard of Babelfish or like interacting with somebody who speaks a different language, like I think they intuitively understand that like translation is interesting and the idea of like dubbing of the service is interesting. But I think if you zoom out more broadly, I think the question around like is synthetic voice and the ability to take one form of media and translate it to others cheaply and easily, obvious in hindsight. I think I've been like somewhat surprised by the

• 7:19Speaker 01

  demand on that side from across a range of use cases. In terms of things that we, you know, are interested in, but like the bar is just very high because the the cost to build a company and the advantages of incumbents are so high. It's like we haven't done a chip company, but we've done companies that are up and down the stack otherwise.

• 7:38Nathan Labenz?

  Cool. Well, I want to ask you guys also about, you know, some specific portfolio companies that you've invested in, that you're excited about and get a little tour of kind of some of the use cases and some of the things that will be coming at us, you know, from a consumer or business standpoint in the not too distant future. But I also want to kind of do the same thing for you a lot because about six months ago, you published this essay, AI Startup Versus Incumbent Value. And that hit me at a pretty opportune moment. I was just at the period, at the end of a period of 60 days of super intensive red teaming on GPT-4.

• 8:18Nathan Labenz?

  And I was basically not, I hadn't even really tried to synthesize what I had seen at that point. I was really just scouting, you know, all the different use cases and everything I could think of to test and try to understand what this thing could do. And right as that kind of closed, you published your essay. And so I read that and I was thinking, for me, it seemed like, boy, this GPT-4 is incredibly powerful. And, you know, the conclusion that I started to leap to is, I think that the enterprises are largely going to be able to apply this technology fast enough that they largely won't get disrupted by somebody who's starting, you know, with a language model and then thinking like, how do I build all this other stuff around that?

• 9:02Nathan Labenz?

  How would you grade my intuition, you know, from six months ago? How has your thinking evolved on that question now that you've had the advantage of seeing GPT-4 launched? And the, you know, the deal flow that you're seeing downstream of that?

• 9:14Speaker 00

  So let's see, so, you know, I got really interested in generative AI a bunch of years ago, probably four or five years ago, as all the GAN stuff was happening, simply because I thought that, you know, the GAN-based art stuff was super interesting. And before that, I'd been investing in AI and also worked on AI-related products myself directly for like 10, 15 years. You know, so when I was at Google, I worked on mobile and ads targeting and ads targeting were big ML systems. And then I sold the company to Twitter and Twitter, one of the teams that worked for me was search and that was all ML and AI. And then I invested in the area for about a decade. And for about a decade, nothing worked, right? Or I should say, a lot of things worked for incumbents, but it didn't work for startups. And so you had the Facebook news feed and you had Alexa from Amazon and you had all these really big products.

• 9:57Speaker 00

  But the startup ecosystem, in terms of companies that were started to specifically do ML, just really didn't seem to go anywhere in terms of, you know, building really massive companies. And then this generative AI wave hit. And I think things started to get really interesting around GPT-2. And then maybe as GPT-3 came out with the big step function and functionality, you realized how compelling it was. And I remember I went on a recent podcast and we talked about it specifically. And I think at the time, a lot of people were ignoring it. And I started digging around, I tried to find people to build an OpenAI competitor. And I couldn't convince anybody to do it. Everybody said, well, it's not that interesting of a business and, you know, are these APIs that good and all this other stuff? And I pitched person after person and nobody was willing to try it.

• 10:41Speaker 00

  But a lot of people who'd worked in the area before wanted to build applications. And so I started investing in companies like Character, you know, Noam Chizir is one of the main authors on the. On the Transformer paper, I helped out some of the early team that was working at Adapt, although I never got involved as an investor. I got involved with things like Perplexity and Harvey and a variety of companies that basically, I think, ended up forming, you know, some of them were interesting companies now, a year or two later, in hindsight, in terms of this wave of AI stuff. And a lot of the question in my mind is, if you look at the history of technology waves, there tends to be differential capture between incumbents and startups, and each technology wave is different. And so if you look at, you know, the first internet wave, it was like 80% startups. It was Google, and it was Amazon and all these new companies. And then people like Microsoft benefited too.

• 11:26Speaker 00

  And then you look at, you know, mobile, and that was 80% incumbent value. It was the big platforms were Apple and Google, which were already incumbents. You know, people were talking then about what is Salesforce on your phone going to be, and who's going to build it? And it turned out to be Salesforce. Built Salesforce for your iPhone, right? Or search on your phone was Google. But there was new things like Uber and Instacart and Instagram, basically anything with Insta in it, you should have just invested in. And then you had other types of platforms that emerged. You know, for crypto, it was 100% startup value. There was basically no incumbent capture of crypto, right? And so for the first decade of AI, with all the CNN and RNN and GAN-related approaches, all the value went to incumbents. That first wave of AI was an incumbent wave. And now we're seeing something really interesting, where I think it's going to be a differential

• 12:11Speaker 00

  split, and maybe it's 80/20, right? 80% incumbent, 20% startup. But 20% is a lot for what I think is probably the biggest platform shift in, you know, a decade plus, maybe two decades, maybe longer. Because to Sarah's point, you're changing a few things in a massive way, in an underlying way. You're changing the compute model and how to write code. You're changing the user interface. But you're also changing the baseline functionalities, and what this wave of computing can actually do in terms of both applications, as well as sort of other deep areas. And so, you know, there's tons and tons of places that I think is going to impact. Sarah mentioned voice and dubbing and text-to-speech. And, you know, I think those are super interesting areas, and she and I've talked about those in the past a bit. There's tons of room, I think, for social products. And I'm really interested in like, what does a generative social product

• 12:56Speaker 00

  look like? There's lots of apps on the B2B side. There's lots of tooling, like a langchain or Llama Index or other things like that. And then obviously, there's the base LLM layer. So there's just a ton, right? A bunch of that stuff will go to incumbents, you know, probably the base models are largely incumbents with maybe Anthropic and one or two others being the the counter examples, right? OpenAI, Microsoft, Google, et cetera. But there's lots and lots of room for people to build, you know, brand new De novo things that will be super exciting.

• 13:21Erik Torenberg?

  Just just on that for a moment, you know, you mentioned a lot that you tried to get people to build OpenAI competitors. And they, you know, you couldn't couldn't get people to bite. What are you guys trying to get people to build today? For the talented people listening to this podcast who want to do something and what are the things that people aren't spending as much time as they should perhaps, or maybe overlooked opportunities within space?

• 13:42Speaker 00

  I would say voice applications, social, certain big B two B applications, and then certain types of infrastructure. I don't know what Sarah, what you think, but those would be kind of the four quick ones.

• 13:50Speaker 01

  Yeah, I'd add, there's this idea of tool use, right? So, you know, an L M can be a reasoning engine against knowledge that it holds in the model itself or some database or, you know, some repository of information. But it can also take action now, right? So if you think that tool former an automation in the previous generation of products in the sort of RPA category, I think that's going to get a lot more interesting when the approaches get more robust. There's a lot of workflows in every part of the enterprise, but especially the back office and some verticals like healthcare where like you have a lot of people moving data around between systems or filling out forms based on some policy.

• 14:30Speaker 01

  And like, we've been unable to do that flexibly today, but it is a, you know, basic tool use in natural language tasks. So I think that's a really interesting one. Like, there are some areas that we think are just going to get more important from a core architecture perspective over time. So the idea of retrieval and just like, how do you guarantee? Like retrieval and memory are two concepts that I think are really interesting in research that people can't figure out how to use in like actual enterprise applications. And then, you know, I'm really interested in some of the more emergent stuff. Like if you look at companies like Mid Journey, the idea of democratizing capabilities like that people didn't have before. If that's illustration, which I promise you, not a lot of VCs were focused on before.

• 15:14Speaker 01

  But generally, like, I think media creation is is really interesting.

• 15:18Erik Torenberg?

  A lot, can you can you say just a little bit more about what AI social could look like?

• 15:23Speaker 00

  You know, I actually have a lot of ideas, but they're probably really bad ideas, right? I think with social products, the key thing is you want to launch things and then quickly iterate and sort of see what gets adopted. And you know, I know people doing a range of things, but I also don't want to kind of like dox them or something. In terms of what they're doing, but you know, over the last couple of months, I think I've heard a couple of really interesting ideas on the social side. And you know, it's a variety of different formats and approaches and everything else. And I just think there's, there's a ton to do there. And I think the issue is when I look at social products today, people are basically constantly trying to rebuild Twitter for some reason. You know, every, every month there's a new, hey, we're doing Twitter, but just decentralized. We're doing Twitter, but it's whatever. Not with generative AI, right?

• 16:07Speaker 00

  Um, or people are trying to do like, you know, Facebook throwbacks or, and you're like, what's, what's, where is the technology heading? And what does that mean in terms of entirely new types of interactions? Where you're still taking advantage of core social behaviors, right? Ruloff has this seven deadly sins, right? Every social product basically is like gluttony or lust, or you know, one of the seven deadly sins. And I think if you think of that through a generative aspect, there's really interesting ideas that you can start coming up with versus saying, I'm just going to throw things back. It reminds me a lot, actually, when I left Google, it was a long time ago, either way, uh, whenever I left Google, whatever year that may have been. Um, the, uh, a lot of people were building things that they shouldn't have been building because they were building for the past.

• 16:49Speaker 00

  So they're like, oh, I'm going to build this SEOable thing and get traffic that way and blah, blah, blah, instead of saying, hey, I'm going to build a developer tool, which is a new thing, or I'm going to build a mobile company, which is a new thing. And so I think, um, a lot of the social products I see are reflections of the past versus the future. And, uh, that may work. It may actually create really big companies, but the flip side of it is there's probably some really interesting things that we can all kind of squint and imagine that's coming.

• 17:16Speaker 01

  I think that there is a lot of open mindedness required for some of the consumer stuff. In, and that if you like consume, like within consumer, character is one of the most interesting companies. Replica is one of the most interesting companies. A lot of people don't like this, even though you see like, you know, a decade or more than a decade for a lot of looking at consumer company metrics, you're like. Shit, right? Like, I'm gonna pay attention if people are spending hours a day on this service because it is so rare. I think it's very easy not to like it because it's a weird thing that people want to have these parasocial relationships and and they're because there's demand for NSFW use cases. But like, that's how a lot of things on the internet start, right?

• 17:58Nathan Labenz?

  Yeah, we had Eugenia from Replica on as a guest. It was certainly one of the more fascinating conversations that we've had to understand. First of all, just like what the user base is today and has been historically while the models have been so limited, frankly. And then to kind of extrapolate that into, you know, the present and the future where it's like, this was not honestly super compelling to me, but I see how it could easily become, you know, much more compelling. It's, there is a phase change or kind of a threshold that we've hit, I think, that is gonna kind of take Replica 1.0 and make it look pretty quaint as we hit 3.0.

• 18:36Speaker 00

  Well, I think it's gonna be deeper than a lot of people imagine.

• 18:39Speaker 01

  Seven, eight years ago, we started investing in what you'd think of as like mobile coaching, like marketplace applications for different areas. So that could be like something in health, like nutrition, or like people are doing like fitness training and such. And as you might imagine, like having an accountability partner or somebody you're building an emotional relationship with means you can affect behavior change, which is really hard for humans. And one of the most interesting things I've seen recently is bots can convince and like coax people to do things, right? Plan their days, change behaviors. And so I think that's something we're gonna see a lot of.

• 19:19Speaker 00

  Yeah, I think the applications of it are gonna be broader than anybody thinks. Like, if you look at it, you know, if you think about education, how do you revamp education? And everybody's gonna have like a bot. Is there a kid growing up that's gonna teach them things and help them with stuff? And maybe it becomes their best friend, right? And there's very positive and very negative implications of that, right? And so I do think people are dramatically underestimating the degree to which, on the one hand, there's a bunch of lonely people or people who wanna interact online more and they don't have the capacity to do it otherwise. And then on the other hand, there are these really deep, fundamental societal use cases that are coming through the generation of these agents that... Interact with you like a real person. And in some cases, I mean, every parent is going to want the

• 20:03Speaker 00

  thing that's going to educate their kid in a hyper customized way. And that's going to be both very powerful, but which company is going to control that? And what does that mean for our kids and how they're taught and raised and all the rest of it? Right? So I think there's some very deep fundamental things here that people are just barely touching the surface on. And some of it's in old sci fi literature, like the Diamond Age, right? It's the young lady's illustrated primer. But in some cases, I think people haven't really thought about it very deeply. Or there's another book called Lady of Mazes, where every time there's a sufficiently large block of people that believe something that substantiates into an AI agent that represents them in Congress. And so why even vote when you can have a perfect representative and you suddenly appear and actually, you know, fight for and adjudicate the things that you truly care about?

• 20:47Speaker 00

  And so I think there's all sorts of crazy things that are coming.

• 20:50Erik Torenberg?

  On a more quote twitting day to day level, I want an agent that just helps me maximize my productivity. Uh, that's watching me at all times, watching all my interactions with people and, you know, tells me when I'm acting out of line or says that, no, say this, this would be better thing to say. Kind of like a personal trainer for all things life, coach, you know, that's watching me at all times.

• 21:09Speaker 00

  Yeah, I just want a sicko fan take AI. Let's be like, I'm so good. That was such a good joke. Ha, ha, ha, you know, just pretending. Yeah, that'd be amazing.

• 21:21Nathan Labenz?

  There's a couple of really interesting questions here that I think these examples get at. It's funny you mentioned the training one. Going back to my GPT four experimentation, one of the things that I tried to do is just see like, how many sort of chat, you know, specialized chat agents could this thing sort of play? I did the physical, you know, kind of exercise coach one. Uh, and I also did one simulating tech support for my 90 year old grandmother, uh, which was even maybe more eye opening to me. Like really spoke to a pain point that we have in my family. But how do you guys think about that as investors, right? Cause I'm, I'm sitting there using base model GPT four and it's basically working. And then I'm kind of like, this feels like it sort of hits the threshold.

• 22:06Nathan Labenz?

  I can certainly, you know, wrap this up into an app or somebody can. But at that level of like, I used to go hire a human to do this. Now I can maybe slot in an AI to kind of play that role. Are there businesses there or is that all kind of, is all that value accrue to OpenAI in your minds? Or foundation model providers in general.

• 22:27Speaker 00

  Yeah, I mean, I think there's tons of room for standalone applications, and I think a lot of them will be building workflow against it. Or some form of like storage or history or memory or something else that sort of associates with the chain of stuff that you did relative to that. You know, it's kind of funny. I'm going to give an extreme example, which doesn't quite apply. But in the 90s, everybody thought that everybody was going to set up their own email servers. Right? Oh, email is a protocol and everybody's going to use it. It's so easy to use. And then obviously, everything just centralized to like Gmail and Yahoo Mail and whatever your corporate server was. And I think the same thing happens with a lot of these things where you know, there may be interfaces like ChatGPT or the like and things like memory and some of the other things that, you know, some sort of recursive interaction across a

• 23:09Speaker 00

  language model will come into play and chaining and all sorts of things. It'll be more complicated, but I think fundamentally people will need very specific workflow for very specific applications in many cases. And in some cases, you'll have a general purpose tool where ChatGPT will be good enough to just do a bunch of stuff for you, right? Or whatever, whatever version of an agent you're using in the future. So I do think we're going to end up in a multi-agent world, but there may be like specific things that are dominant for specific use cases, just like everything else that exists today. You know, I feel like the best indicator of how things will evolve is kind of like, how do, how do market structures evolve in the past? And I think it's going to be kind of the same thing.

• 23:46Nathan Labenz?

  So how about like use cases or not use cases exactly, but modes of interaction? This is something I've really been trying to organize my thinking around. Eric's got this vision for the, you know, AI that kind of rides shotgun all the time and like, you know, helps him maintain his social graces. And you can kind of, you know, that's kind of the Reid Hoffman vision, I would say, is like the co-pilot for every profession, co-pilot for every phase of life.

• 24:14Nathan Labenz?

  And then you're speaking to also on the other end, like people are going to need Specific workflows, I kind of think of that and Sarah, you mentioned like RPA, like there's this sort of process automation context where it's like, I'm a big corporation, you know, I have like these cost centers, which are humans that have to like do these tasks. I've never had any way to even think about automating these tasks in the past, but now I sort of have that. And then there's kind of this third way that's emerging that's like the agent model, which I kind of think is bridging, I think of as bridging those two because I can like, talk to it in a sort of ad hoc, real time way, but I can also kind of send it off and say like, you go figure out the plumbing and like how things connect together, you know, so even get me out of the business of having to design, you know, architect the workflow. I guess, you know, that's enough for me.

• 24:58Nathan Labenz?

  How does that framework resonate for you? Do you have like your, a different one that you kind of bump company, you know, deal flow up against? And what modes of interaction do you think are ultimately going to predominate? And is that the same as those that give you the best return on your investment?

• 25:11Speaker 01

  I think when you start to actually like look at the tools that have succeeded at scale, there's like a whole range of ways that users want to interact with the stuff depending on the task. So like prompting is not an easy thing for like 99.99% of humans today. So just because you enjoy like Nathan messing around with GPT-4 and lots of users of your podcast might, it's hard to ask a good question. And I think like one of the things that I've seen in companies that I think will just become more common is like multimodal input, passively using context, right? I think there are a lot of companies that figured out like giving end users in a particular category,

• 25:54Speaker 01

  20 prompt templates that made sense for their use case and an easy button. So they don't have to figure out how to engineer a good output. Like that's a company right now, right? And so it's not clear to me that we're going to have, you know, generic interfaces for all the different use cases. One of the things that I think will happen to the point of like, do we? Like, I think search is going to, right, and I'd love to get, you know, a lot's point of view on this, since he actually worked on search. But I, having been invested in search companies prior and having friends starting them now, still, search has many use cases. It is weird to me that, like, getting information from the internet has fallen into one box at Google.

• 26:33Speaker 01

  And I imagine that many of the use cases, like the stereotypical one being like travel planning or buy something. Like, that's something I think an agent should be able to much better do for you in the future. So I think there's like certain things that will fragment from a market perspective. And every slice of that market is like plenty valuable to go after for a new company. So one of the basic frameworks we use. I don't know if I've got like the sort of overarching unification right now. The ground is too unstable.

• 27:02Speaker 00

  Yeah, I kind of have two answers to it. I think there's almost like a two by two matrix of like, is a person busy or do they have a lot of free time? And then as a context, you know, the context kind of, maybe it's not two by two, it's like three by two or four by two or something, right? It's like busy versus free time. And then one of a series of contexts around B two B use cases, commerce slash action based use cases, et cetera. And then based on that, you're going to have a different modality. And so I think you can almost come up with like a map on it. And it reminds me a little bit of like, if you work at a big company, the way that you interact with like the CEO is different from how the CEO interacts with you. Right? You'll write this long email of like multiple paragraphs and the CEO will send you, yep, as like a single word or whatever. Or, you know, execs tend to leave a lot of voicemails or they used to, right?

• 27:45Speaker 00

  Because it was a more performant way to communicate. And so that's kind of busy versus not and all the rest. And so I think there'll be all these modalities. I think the other answer is, in some sense, it kind of doesn't matter. It's funny, I met with this hedge fund guy who's really sharp, like, you know, amazing investor. And we were talking about AI and, you know, a lot of his questions tended to center on this kind of stuff. It's like, well, will you just talk into your phone? And I was like, who cares? It doesn't matter. That's missing the main point, which is what is this technology fundamentally enable? And we'll figure out the interface and we'll iterate on it and it'll be one of a series of interfaces that we use today, right? I mean fundamentally, we have like N-senses and we'll have different modalities that matches different senses depending on the use case. But it's clear that, you know, say that you use Alexa, right? People with kids

• 28:30Speaker 00

  love using Alexa because it's voice-based and the kid can yell at the thing and have a reply. But you're not going to have a lengthy information extraction conversation with it unless you're like a three-year-old, right? And so I think it kind of maps to the, you know, what are you trying to accomplish? And I think the most interesting aspect of all this stuff is just like, what are the fundamentally new capabilities that all this enables? And what does that mean in terms of the applications that can be built, in terms of how it reshapes our lives? You know, like, I think there's really, I'll give you an example. You know, obviously, these models now perform better than many doctors on standardized medical exams or other types of tests. And you can imagine a world where you start having models that are basically available to anybody in the world, which allows you to upload an

• 29:12Speaker 00

  image and, you know, describe some symptoms and then you end up with medical care that's, you know, in some senses on par with what you get at Stanford or whatever top, you know, medical association. And you can do that anywhere in the world as long as you have like a phone that has your own characteristics. That's really, really, really powerful. And to some extent, the interface is secondary to that impact. And so I'm not trying to denigrate the interface question. I think it's really important stuff. I just think like fundamentally, the capabilities are so rich that it's almost like, okay, where do the capabilities take us? And then based on that, what happens as the output? There might be multiple types of outputs.

• 29:49Nathan Labenz?

  Interface is definitely interesting. There's all, you know, I'd be interested to hear if you guys have seen any really creative ones that, you know, you would recommend that people check out. But I'm also kind of thinking even a little bit more like big picture than that, like, just how do we relate to these damn things in the first place? You know, are the, is the copilot feels kind of like a peer, you know, or something like a real-time collaborator? And that could be an audio interface or a text or, you know, UI or whatever. But then there's like, you know, the agent, you're kind of delegating to it. And then there's the sort of, you know, supervision mode perhaps where like, you largely trust it, but you kind of, you know, maybe trust but verify and hopefully you like actually do the verification and don't just, you know, start rubber stamping

• 30:32Nathan Labenz?

  everything. What about on that level? Do you have a sense for kind of?

• 30:37Nathan Labenz?

  Where this is going, like how, another way maybe to ask this question is like, how weird do you think things are going to get as these tools?

• 30:45Speaker 00

  I think eventually we're gonna call all these things your highness as they sort of take over the world.

• 30:51Speaker 00

  I love my boss, the AI. Yeah, no, I think the world is gonna get really interesting and weird. And I think that's back to the education point. For example, like if you have a bot helping raise your kids, what does that mean? Right? If one of the primary sources of information no longer becomes YouTube, it becomes some agent that's not only working them through a math and history and other curriculum, but potentially is choosing which form of history gets presented to the kid. You talked about base model and I can't remember if it's biology or something else. It sounds like something you'd say, but I don't know if you said it. You know, there's base models and there's based models with a D, right? Like what, what kind of model do you want your kid to interact with? And what do you want them to learn over time? And, you know, how does that, how does that get selected? And who, who adjudicates what that selection process is?

• 31:36Speaker 00

  And so I do think there's a lot of these really interesting things that are coming because you're RLHF-ing something. And who, who's that cohort of people who's training the thing? That are providing the human feedback? And how do you select who those people are? Or what's the, the ethical framework based on your location around the world that should be applied or shouldn't be applied, right? Should the Western viewpoint be applied to somebody in another country that may have very different values and mores relative to the model and its output?

• 32:08Speaker 00

  And so I think there's lots and lots and lots of interesting questions here.

• 32:11Speaker 01

  You know, I, I think it's like useful to like try to imagine the interactions we have with agents in a few different ways, right? And so like a lot of saying, like your kid or you, like today we are at the mercy of a bunch of algorithms that control our information flow. Right, and we can lightly curate them by like swiping correctly on TikTok or Twitter or whatever, but if you can instruct your your bot more directly for yourself or your kids or whatever, like that's quite interesting. I think like a fun interaction to imagine is in like think about the enterprise side, you have different influences in different teams because there are incentives for for

• 32:54Speaker 01

  example, like compliance or security in an R and D team. Um, I think there's a a simple version of that, that's a bot, that's like an early warning system. Like imagine in a fintech company like, oh, you're like hitting these, you know, these merchant network rules, like the thing that you're trying to do is like a definite no go. Um, but you can also imagine like a debate with that bot or um, a a fight with it as an extension of like that security champion or whatever. I think another really powerful one um, that I like, just looking at some of the examples of how people use like ChatGPT and Copilot is code generation. Right, I like, we're not that far off from a like, e- especially in areas where there's just so much content online, like web development, junior web developer, like junior Python developer

• 33:39Speaker 01

  available to everyone, right? Not like complete my code, which requires a bunch of previous knowledge, but write to a file, run code, like deploy to the cloud, use APIs, like that is really powerful. And so I think um, I think there's like, you know, I think of it as like, there's new capabilities, right, that are like thought of as human capabilities. There's like interactions where I actually have to negotiate. And then there's like the things that I can control that are like personal. So I I think we're gonna have a lot of really weird interactions with agents.

• 34:08Nathan Labenz?

  How does that sort of expectation of weirdness change how you guys are thinking about your role as investors or the investment decisions that you're making? Um, I would imagine that like, it would shift you more toward team relative to like current product, for example, but I'm wondering kind of what shifts you're finding compared to previous, you know, cohorts of companies.

• 34:32Speaker 00

  You know, it's interesting. I think it was Chris Dixon, he said that the next great company starts off looking like a toy.

• 34:38Speaker 00

  And I think that's true both on consumer, that's true with crypto, that was true with that with certain types of enterprise. And so I don't think it changes that much. I think the really weird stuff is often the most interesting stuff. And then there's going to be the standard stuff that you just know is going to work, right? And I think it's going to be that same mix. I think social products in general tend to be weirder in terms of the things that actually work, or at least the behaviors tend to break with other affordances. Kind of generationally, you know, Snap, we're going to make every image disappear. And obviously that product morphed quite a bit over time, but at the time, everybody was like, what are they doing? You know, that's so weird. My sense is just like Evan taking selfies of himself and then it would disappear. And that was the whole network for a while, you know? So I think that um, I think behavior will always start off seeming strange.

• 35:21Speaker 00

  I tend to be, and you know, Sarah and I don't have any like formal business relationship, right? We just like collaborating on stuff and we have a podcast together and stuff. So I'm speaking for myself only, but like, I tend to be very much like a market driven investor, not a team driven investor. I should say the team is incredibly important, right? I've started two companies myself. So if I didn't think teams were important, you know, I never would have started a company. But I think the markets are more important or the product market is the most important thing. And so often what I look for is like, what are early signs of product market? And then do I think that's in, do I think it's in a big town? Do I think the team is great? Do I think there's defensibility is or a why now statement? You know, there's all these other things around it, but you know, fundamentally, I've always looked at it as product market. And the question is if something's really weird, how can you tell if the product market is there?

• 36:07Speaker 00

  And I think almost every great startup has to be non-obvious because if it was obvious, everybody, everybody would already be doing it. Right? So definitionally, these things have to be somehow off or there has to be some hurdle to overcome. Otherwise, it's not defensible.

• 36:22Speaker 01

  I think it's kind of funny that um, some number of months ago, a lot couldn't convince anybody to start an OpenAI competitor. And now he probably can't convince anybody because they're like, oh, they're too far ahead. It's too big of an incumbent. So it is like a really interesting pacing.

• 36:39Speaker 01

  Uh, related to the, I think one, one big mistake investors make is they are, they're just kind of blind based on like their current view of the world, right? Like, it's very easy to project your existing view of where the value is, especially if you were like a very focused on the more sophisticated customer. Um, and you take that point of view and then you don't see like the actual demand, which might start with people who like don't have access to something, or um, where like their use case is less sophisticated. And so I think it's, it's like really easy to see this in the media space. So like if you look at something like illustration, right? There's like, there's this view of like, oh, like, you know, it's never gonna be good enough to like make picture books or like do Coke ads.

• 37:24Speaker 01

  And like, pretty sure directionally, we are going to get there like sooner rather than later. But if you, if you take that point of view, like, oh, it's not gonna work for the AAA games, right? Or it's not gonna work for like people who need Super Bowl quality video. Or um, yeah, but I wouldn't like, you know, like use that in my production code base. Like, you're just gonna miss a lot of like, well, what are people actually using it for? And directionally, where are we going?

• 37:49Speaker 00

  Yeah, I think people also over-index on defensibility related to that. And so everybody at the beginning of a company asks too many questions around, how does this become defensible? How is this defensible now? And can't people just build it? Because two people built it in six months. And that's sort of like every SaaS company.

• 38:05Speaker 00

  Right? What was defensible about Retool in the early days? Or Notion in the early days? Or sort of Choose Your Startup in the early days? And in the very early days, nothing was defensible, right? It took two people, three months to build the thing, you know? Um, and so I think that's kind of similar here, where there's a lot of questions around, okay, what's a defensible business model? And you wanna have defensibility over time, absolutely.

• 38:26Speaker 00

  And if you look at it traditionally, there's all sorts of ways to do that, either in terms of platform effects, or um, certain aspects of sales, or certain aspects of, you know, integrations, or other things that you do over time. But fundamentally, I think, or network effects, right? There's all sorts of forms of defensibility or moats. But I think, I think people um, really early uh, tend to ask almost too much of the thing. And the real question is, does anybody care? And is anybody using it? And then I think later, it becomes, okay, now that people are using it, how, how does it become defensible and not a commodity? And how does it scale? And how much does it scale? You know, is this an N of one company or product?

• 38:59Nathan Labenz?

  So what trends are you guys seeing in like usage data? I feel like right now, everything has a waitlist. The waitlists, some of them are moving, some of them are not moving. This feels like a moment probably where like, everybody can post good signup numbers, but I would guess retention probably varies widely in the deal flow that you're seeing. I'd love to kind of get a sense for the trends that you're seeing there.

• 39:25Speaker 00

  My favorite waitlist example of all times was this early AI company. I'm not going to name which one it was. It was like 10 years ago. And the founders of the company claimed it was an AI company, but in reality, they had a bunch of ops people like answering queries in the background. And the co-founder of a very well-known, large tech company went on to it. And anytime we'd go on, they'd ping all the ops people and they'd all jump on and answer all of the queries really fast for that one person. And they end up getting bought by this like major tech company. And it was completely false. Like it really wasn't working the way that they were claiming it was. And they always were in private alpha. And they'd say, oh, we just, you know, there's so much demand and look at this giant waitlist and all this other stuff. And they never actually like really launched and they got bought for a bunch of money.

• 40:08Speaker 00

  And so I feel like often these infinitely closed waitlists are kind of a negative sign that the traction may not be real. Now, sometimes it's a real sign of demand and there's some scalability issue in the background or they want to test it and all the rest of it. But I think if you have such raw, organic adoption, you know, usually you just open the thing up because you know more and more people will join unless again, there's some constraint that prevents you from doing it. I think in general, the last decade has taught people some wrong lessons on how long they should take before launching a product. And people point to Figma or they point to Notion or they point to other companies where there is a longer period of time for development. And if you talk to the CEOs of those companies and say, I wish it was faster. We should have done certain things faster.

• 40:53Speaker 00

  Hey, it didn't work the first time, so we changed it and it worked the second time. But we tried, you know, we tried actually to get people fast. And so I think there's this whole like bespoke artisanal movement. Or similarly, one company I'm involved with was doing hand onboarding of every company superhuman style. And eventually one of their customers said, why are you getting in the way of your customers? I just want to sign up and use the thing. Why are you onboarding me? And so they stopped doing it and they had a spike in usage.

• 41:19Speaker 00

  And so in general, I think you want to get out of the way of your users. And that means, you know, you don't necessarily need a waitlist unless there's very specific reasons behind it, or you really need to test certain things. But after you've tested things enough, if you keep having a waitlist two years later, it means the thing isn't working.

• 41:32Nathan Labenz?

  Of the companies that you guys are taking a look at, how many of them seem to have sustaining usage versus kind of that surge of interest that, you know, if you give it a month, it already starts to look like it's tailing off?

• 41:46Speaker 01

  It's some tiny percentage, right? But that's kind of like the point of ventures. There's like, it's all about the tail of companies. It is not like your, your average company in the market. There's a big hype cycle in the market right now, right? And I think it's really, it's like very, um, it's very easy to feel smart by being cynical and dismissive in venture and in investing in general. And it's like, this is totally useless. I, you want to be intellectually honest, but like if, if that's your orientation, like, don't be an investor in startups that are generally weird ideas. Like what I'm looking for is there, is there any real data at all? Because you're trying to invest in the outliers and the outliers right now are insane.

• 42:28Speaker 01

  Right? Um, like I know a founder in the, you know, many, many millions of revenue. That was like, ah, you know, like I'm, I'm dealing with some sort of cash flow issue, like GPUs or SVB or something. And he like removed some part of the free tier of his product and revenue went up a multiple. Right? And I'm like, that is some pretty impressive demand. Like, that's not fake when there are consumers who are paying that much for the product. And so I think the most exciting thing right now is that there are capabilities that consumers and enterprises like get that much value from. Where, you know, one could argue that this, this company that is doing this much revenue

• 43:09Speaker 01

  hasn't played any of the like good growth games or good product management games that like companies like they play act at, right? Like, oh, I'm going to do hand onboarding because the good company did that. Or I'm going to have this like waitlist with the sexy brand companies like in first, because like that's what the best companies did. But the real thing you're trying to figure out is like, can I, can I make something that just creates so much of its own demand? And like, nothing else is important, right? And so like, yeah, like the hype cycle is useful for fundraising. But if your investors know what the data should look like. Then they're not gonna buy it. I think it's especially easy to generate like that sort of waitlist activity right now because

• 43:52Speaker 01

  um a lot mentioned like the seven deadly sins, but like greed, like social signaling, like fun, fear, like people feel all of these things about AI right now. Um and it's very novel, but anybody who's worked on these products understands like the distance between demo and product is very large. And so there are a huge number of companies that have these massive waitlists because it like looked really cool as an idea and then it doesn't actually work, right? Or, you know, people thought they wanted it, but they don't. And so like, that's the whole game of trying to understand like, you know, when I talk to a company that's actually got a reasonable waitlist, like the first question I ask is just like, well, like, what would happen if you gave everyone access? What is blocking real usage growth?

• 44:36Speaker 01

  And often the answer is like, oh, we know that those are like garbage users anyway, right? So like, why, you know, don't, don't project that. Um but the thing I get excited about is like, you know, in the tail, there is usage that is unlike anything I've seen in the last decade plus.

• 44:50Speaker 00

  I think there's also like a founder perspective on this, which is when should you raise money versus just bootstrap? And I think all too often people get on the venture train when they could just bootstrap. Um, you know, if the thing is just growing organically, really rapidly and it's spinning off tons of cash, just go for it, you know, in some cases. Um, and then secondly, in some cases, you know that you're gonna have something that goes viral, a bunch of people pay for it and then it dies off. And that happened in prior social waves, for example, right? Like all the social gaming companies and people made real mistakes by raising tons of money and then having to play the venture game on those things and blowing up their companies instead of saying, I'm just gonna dividend out cash and it'll last for a year and I'll make a bunch of money off of it and I'll move on to the next thing.

• 45:33Speaker 00

  And in that way, people got stuck for four or five years working on something that was never gonna work because they had that initial burst. And you see that with some of the like um social mobile apps that are using like uh stable diffusion was like an underlying thing and some of them have ramped to, you know, $200 million of revenue in nine months and then they die off really fast. And some of those companies probably would have been better served just running it off of cash and then distributing cash to themselves instead of like raising money. So I think as a founder, you should also think through like. Is the right thing for you to raise external capital? And in many cases, the answer may be no, right? If you can avoid it, why would you do it?

• 46:10Nathan Labenz?

  How are you seeing the use of funds kind of shifting? Because it's kind of a trope at this point to say like, you know, you can probably get by without your like social media manager, or you know, you can like semi automate recruiting in ways you couldn't in the past. And so you don't need as much headcount as you used to. Arvin from perplexity said that on this show, in fact. But then on the flip side, like there are like foundation model costs. I feel like I see some companies that are like basically taking the investor checks and like turning around and spending them on OpenAI. So how are you seeing? I mean, and some of the checks are pretty big, right? They definitely are. Not everybody is following that like bootstrap if you can advice. People are grabbing the grabbing the eight figures.

• 46:52Speaker 00

  Yeah, most people aren't. And I'm not saying most people should. Yeah, I'm just saying like, it's another path. And people always forget that it's another path.

• 47:00Speaker 00

  You know, there's a lot of companies, I think, that have raised like $15 million to build a model when they should have just gone on GPT four. And they would have had the same rough outcome. And, you know, they could have done some prompt engineering or something. And so I do think there's a lot of people who kind of went down the wrong path. And there's a subset of use cases where maybe the model really, really matters. And that could either be specific vertical use cases, maybe certain types of healthcare, where you have certain unique proprietary data sets you want. It could be if you're really training for specific interaction modalities or applications. But and then, you know, obviously there's a big price differential. If you're training a diffusion model versus if you're training a transformer based model, right? You're talking about something that may cost hundreds of thousands or millions of dollars instead of tens of millions of dollars, depending on what you're doing.

• 47:45Speaker 00

  So I think it also depends on you doing things in image and video, or are you doing things in text?

• 47:50Nathan Labenz?

  Because you guys are seeing so many things, I think you're, you know, I'd be particularly curious about your answer on. Cool, what are the coolest products, the most useful products, the most, you know, interesting, you know, glimpses into future paradigms or interfaces that you would say, you know, and especially if they don't have a waitlist that you think people should go check out?

• 48:10Speaker 01

  Um, for your, I mean, for your average consumer, like, I think the, the list starts by being very simple, right? A lot of the incumbents have, um, like, very nice magical features now. And incumbent is a broad word, but like, I, I do think that the things that Canva, Adobe, Figma have shipped are like really useful and cool. Everybody should try ChatGPT. Uh, I don't know that one of these services is public yet, but, you know, just because it's the bane of my existence, like, rich contextual email completion is going to be here very, very quickly. And I think it's like in, uh, in waitlist beta. And it's like, you know, there's this meme about like, oh, like, somebody's gonna like, make your email longer and another agent's gonna make it shorter. And it's like, there's gonna be this adversarial interaction between the email

• 48:54Speaker 01

  agents. But I do think that, like, being able to do drafted responses to everything in your inbox is one way to resolve this stupid issue. I, I think those are, those are some of the fun ones. And on the, on the consumer stuff, like, character and, um, the other, like, uh, sort of bot parasocial interactions, I think are great. I think the agent stuff is probably too immature for your average consumer to have a good interaction with it. But I think we're less than six months away from, like, bots that actually do work for you.

• 49:25Speaker 00

  You know, I think with all sorts of, um, cool things happening right now, I think some of the dubbing things feel really magical. Like, if you play around with something like Easy Dub's and you just drop in a video and it, it, uh, tries to capture like the tonality of the voice of the person as it translates them into Spanish and stuff like that. Or from Spanish into English, whatever languages you want to do. I'm still waiting for them to add Hebrew, which is why I'm not a power user yet. That's why I translate all my content into Hebrew. I'm just kidding about that small audience. Um, but I think, um, I think there's lots of magical things to come, or things that feel magical. And I think it's kind of interesting. I feel like, for example, AutoGPT, you know, got a ton of attention. But for anybody who was in the AI community, it was kind of the obvious thing that was going to happen.

• 50:10Speaker 00

  Or some form of it was going to happen. And so I feel like there's a lot of things that are coming. It's back to the old saying that the future is here, it's just not equally distributed. And I think there's a lot of things that a lot of people have been thinking about or realizing is going to be really cool. Or, you know, they built demos of that are going to start hitting broader audiences reasonably soon. And I think some of those things are just going to be really magical, you know, it's just going to be kind of amazing how these things work.

• 50:36Nathan Labenz?

  So another kind of bit of the future that's starting to take shape is the Neuralink implants, which they've, you know, taken as far as the great apes. Yeah, I have two of them. Well, then you've answered my question already. I was just going to frame a hypothetical for you and say, let's imagine a near term future where, say, a million people have one of these things. And it's like broadly, you know, seen to be safe, like they're walking around, you know, doing okay.

• 51:08Speaker 01

  We're a couple thousand dead monkeys away from that, man, but I'm excited to imagine it.

• 51:13Nathan Labenz?

  Yeah, we're not quite there. So anyway, would you guys be interested in getting one and being able to interface directly with the computing world with your thoughts?

• 51:21Speaker 01

  Yeah, of course. Right. I think like that's a, that's an easy yes, but I have a house full of Gen one. You know, broken consumer hardware and like here is one we're probably not going to be an alpha tester. But if you just think of it, like actually looked, I've looked at a series of companies around this. Like, let's say like, it's a, it's a nick for your, for your brain, right? And like, there's a whole bunch of things that are still immature from a technology perspective. But I think it's very difficult to not imagine that we'd want communication bandwidth to be higher with like all of our devices and everyone else.

• 52:04Speaker 01

  And it's also hard for me to imagine that if like as a as a like as an input mechanism for you can do like knowledge capture in this way and it's an advantage for people, which it will be, that it won't become very popular if it works.

• 52:18Speaker 00

  Yeah, I'm very skeptical on timeframe for this kind of stuff, so we'll see. I think I think our understanding of the brain is so de minimis that, you know, with the exception of a handful of systems that are easy to interrogate, like the, you know, visual system and things like that, it's actually we are, I think, the depth by which we understand how most of that stuff works is really shallow and most of the deep brain stimulation stuff, which Neuralink is based on, has really been for treatments of things like depression or a few other diseases. Um, so I'm quite skeptical about it, but we'll see. At least anytime soon, and by soon, I mean five, 10 years even.

• 52:53Nathan Labenz?

  That might be the most bearish take we've had on the Neuralink question.

• 52:57Speaker 00

  Should talk to neuroscientists.

• 53:01Nathan Labenz?

  So last one's a classic, big, uh, big picture zoom out. You know, we're, and I feel like we're just seeing this wave starting to build. It's coming right for us. What are your biggest hopes for and fears for society at large as this, you know, AI wave washes over us?

• 53:21Speaker 01

  Um, uh, let's see. So on hopes, I feel like we actually talked about a lot of this stuff, right? I think if you're, your open mind is like, it's just a really amazing wealth of capabilities. And there's actually, I forget the name of the paper, maybe we can look it up and put in the show notes. But what's interesting is, um, like the, uh, a lot of the enabling technologies today, they help people with a lower skill base more than with a higher skill base, which makes sense, right? Like, if you think about the, the training set of like, if it is code or writing or, um, music generation, any like, if you're, if you're training off the entire set, you're going to help people with the like minimum of these skills or no skills more than like the highest set of skills.

• 54:05Speaker 01

  And so I think, um, I think it's just quite interesting from a democratization perspective.

• 54:09Speaker 00

  Sounds technology is going to help me, but not Sarah.

• 54:11Speaker 01

  It's very sweet, a lot. I think we're both screwed. Um, maybe I'm going to get speared for saying this, but I do think like, Not understanding that alignment and safety research is deeply tied to capability research is challenging, right? If, if like, people don't understand that, if policymakers don't understand that. Um, and so I certainly think that should be like a broad democratic conversation. But like, I think that's, uh, there's a version of the world that, um, you know, we halt a lot of this progress, or there's regulatory capture of a lot of the technologies, like, before we really figure out, like, what they can do, which I think is gonna be pretty problematic. I think there's obviously gonna be nefarious, like, actors that use all these technologies for different things, but we build defensive technologies against that, just like we have

• 54:56Speaker 01

  in the past. You know, I'm a little bit nervous about, everybody wants, like, the best opportunity for their kids and for them to, like, grow up, like, well-adjusted and, like, able to go, like, you know, be useful to the world and feel good about themselves, right? I actually don't know what to do about that in, like, this current environment with unstable ground. Like, what do I teach them? How should they interact with technology, right? Do I want them to be, like, really good prompt engineers, or do I want them to, like, go to Waldorf and, like, not interact with tech? I don't know. Um, uh-huh. And, and I think, like, very smart people have different takes on this, um, but that's one thing that concerns me personally.

• 55:35Speaker 00

  Yeah, I guess on my side, like, I'm, uh, short-term, and by short-term, I mean, next five, 10 years, very optimistic about what all this means globally in terms of, you know, you have that chart of all the things that go up in price over time, which is like education and healthcare and all these things, and all the things that go down. And I think this is one of the few technologies that may actually help address those things that have become incredibly expensive in part due to regulatory capture. So in the short run, I'm incredibly optimistic about the global implications of this technology to health, education, and other areas. And in the long run, and by long run, I mean, a few decades, I'm a huge doomer in terms of eventual species competition with AI or AGI. Um, I think the biggest short-term risk

• 56:14Speaker 00

  To the area in some sense actually is regulation. I think there's a very one-sided call right now to regulate these things in part because incumbents have an incentive to say that because they want to do regulatory capture on the models and prevent new entrants from coming in some cases and in some cases they have specific concerns, but I feel like there is some chance, maybe it's a one in five chance or something, that with this next election, we're basically going into the first AI, the first AI-driven election. We're going to see ad copy and targeting campaigns and robo-dialing with real sounding voices and, you know, all this stuff going into the presidential election and I think a lot of, there's a lot of potential for that to turn into a giant regulatory storm, depending on who wins or loses.

• 56:58Speaker 00

  Just like, you know, when Trump won, there was a giant backlash against social companies who were blamed for that, that win. I think similarly with this presidential election, AI may be blamed for all sorts of things that it may not have really impacted that much, but that may be the moment that it starts to get regulated. And if you look at, and I think there are certain types of regulation that make sense, like I don't think we should, we should have like export controls on advanced chips and, you know, stuff like that. Maybe there's some NIST-style, you know, approach, but I think most of the regulation tends to distort markets in really bad ways and tends to really kill innovation and tends to lock in incumbents in bad ways. And I think it's way too early to regulate the vast majority of the things in this area and I think all the calls I've heard have been very one-sided to regulate it and I think

• 57:43Speaker 00

  those are, it's the wrong thing to do right now.

• 57:46Nathan Labenz?

  Cool. Well, thank you, guys. This has been a lot of fun.

• 57:48Speaker 01

  Thank you, guys. Thanks for having us. Good to see you.