The AI Daily Brief: Artificial Intelligence News and Analysis - A Report from the AI Frontlines with The Neuron's Pete Huang

Episode Date: March 27, 2024

In this conversation, NLW talks with The Neuron co-founder Pete Huang. They discuss Pete's recent time at Nvidia's GTC event, the state of AI media, and changing public perception of AI. Find Pete online: https://twitter.com/nonmayorpete ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI Breakdown, we are talking to Pete Huang, the founder of The Neuron. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.network for more information about our YouTube, our newsletter, and our Discord. Hello, friends, back with another travel-era interview. And today I'm really excited to have on the show, non-Mayer Pete as he goes by on Twitter, who is, of course, one of the creators of the Neuron, an extremely popular AI newsletter. Hello, friends, quick note before we get to the rest of the episode, you have probably heard me talk about the AI education beta over the past few months.

We've had a ton of you participate, which has been amazing, and now we're almost ready to announce something big and something new. If you want to be one of the first to hear about our new approach to learning AI that is hyper-practical, hands-on, immediately relevant, continuously upgrading, and anchored by community, go to B-Supert.a.i and sign up to be notified when the project goes live. We're getting there in just a few weeks, and I want all of you along for the journey. Once again, that's B-super.a-I. In this conversation, we talk about the rapid pace of development in AI, some of Pete's impressions from the recent Nvidia conference, and of course, the way that an entire new media

establishment is being built up around AI, which, as you know, is very interesting to me. It is a really fun conversation, so let's dive right in. All right, Pete, welcome to the AI breakdown. How are you doing, sir? I'm good. How are you? Thanks for having me on. Yeah, super excited. You know, having been through a couple different sets of call it emergent content creators in emerging tech fields, I can very confidently say that I think that the people who are doing AI content right now are way nicer and more sincere and more sort of like driven by the right things than as a for example, the folks who might have been creating content in crypto a couple years ago. And so it's great to have you on.

For those who don't know, you or the neuron, let's just start with a little bit of a background about what you're building and the context, I guess, that you have to sort of be paying as much attention to AI as you do. Totally. So I wrote a newsletter called The Neuron. We have about 400,000 readers. The whole angle is to teach business leaders and knowledge workers how to use and leverage AI in their work and transform their business, right?

It is very much about applied AI is not just, oh, all this like drama that's going on and everything. It's like, okay, like, how do we actually make use of this kind of stuff, right? So we ship a newsletter every Monday through Friday. We've been at this since January of last year. I think going back to your earlier point there in the introduction, I think the motivations here are the people creating content have been very different than the people who, again, are doing crypto stuff, right?

Like, crypto was all about financial gain and financial incentives, like things like that. And then here, I think, you know, at least my motivation for doing this is very much, look, like, this wave is coming. And there's going to be this very, very fast acceleration. and the more that we can bring people along and sort of share what's going on and sort of drip feed the information of, you know, again, how this is going to impact their work, the better, right? And so we saw this as an opportunity to get to do that. And of course, it's been super fun to be in people's inboxes every day and share, of course, the news and contextualize it, but then also

the tips and tricks along the way as well. Yeah, it's super interesting. So first of all, I think huge congratulations are in order. There's been a million newsletters. And we were saying this a little bit before the show, but you guys are in the sort of the very small. number that have sort of risen to the top of the heap, which means you're definitely doing something right. I guess what have you learned? Is there anything unexpected about the audience that's come here or the types of things that they're looking for that you wouldn't have imagined when you started? Oh, such a good question. I think my biggest surprise is just how much interest there is in AI today. Of course, like, I think that sounds pretty obvious

when I say it out loud, but, you know, I spent the last five or six years of my career in tech startups, right? And so scaling a lot of these software. for products. And we oftentimes encounter these quote unquote early adopter types. And in our minds, that's a very specific persona, right? These are the folks who are gadget geeks. These people are always doing research online. They sort of are very self-sufficient on the internet. They just somehow know how to use things on the internet and find a way to certain resources. And when I started, I was imagining, okay, there's going to be a wave of people interested in AI who definitely fit this profile. Certainly, you know, I'm based in San Francisco. A lot of those folks are here.

A lot of them are located in big cities, sort of, again, very self-sufficient. What I've noticed is with AI, it is truly everyone. It is truly, truly everyone that is interested in that can benefit from the kind of resources that we're able to share and context that we're able to provide. The same folks who would not be trying sort of the next startup or the next fancy product that tech Twitter loves, they are going into chatDB or Claude or trying new prompt frameworks or trying new recipes that affect their work, right? So I've been very pleasantly surprised by just how much interest there is around AI,

the amount of attention that people give. And I think it's appropriate because I think you and I both know this, right? Like this is having dramatic impact on businesses today. It's going to continue to be that way and ramp in the future. And so I love that people are spending the time and energy to pay attention to this because that's going to pay off for them. It's going to pay off in time savings every day. It's going to impact their P&L if they are business owners.

And I'm really excited to see just how the future of business and work changes because of this. Yeah, I couldn't agree more. And I think that it's a very, there's a real mental reorientation that's required when you're sort of digging into this space to understand that on the one hand, there are similar sort of early adopter dynamics, but that the profile of who's early adopting is so much more diverse and just larger than things we've seen in the past, right? So it feels weird to say like chat GPT has 100 million weekly users or, you know, probably more at this point, but it's weird to call 100 million weekly users and the fastest growing product of all time early adopters. But if you kind of understand that the total addressable market of people using

this is probably more like 4 billion, then it feels less like crazy to call it early adopters. And I've certainly found that the people who are paying attention to this stuff now kind of get that. There's this interesting combination of urgency from the standpoint of not wanting to be left behind and, you know, sort of fear of missing out of opportunities, but also an excitement that, hey, I am seeing the future a little bit before, you know, the guy in the cubicle next to me to be a little bit reductive. And that's an opportunity for me to, you know, jump some rungs on the corporate ladder or maybe finally break out of the corporate ladder and do that thing that I wanted to do. And that's a very exciting energy, I think, to be around.

Totally, totally. I completely agree. And I don't know if you've seen these stats, But it's something like, you know, the people using chat dbt or AI at work oftentimes don't tell their bosses. I mean, it's sort of this sort of, oh, I've got this secret with me that I know how to do this thing. And it's giving me massive leverage in my day job. It's allowing me to go save an hour or two and go pick my kids, like go run errands during the day or whatever it might be if you're working a remote job or otherwise just do the things that I kind of want to be doing instead of the menial tasks every day or the boring work. Right. So I agree with you. The impact here is huge. It is definitely billions of people. There's also a geographic element, right? So I've taught some courses where the audience was primarily located in Southeast Asia and the Middle East. The awareness of the products, how many people have actually gotten hands on with those products, way less than I would say in the U.S., in English-speaking countries, in Western Europe, et cetera. So I think it's just with how fast this is moving, it's creating some disparity a little bit, both within mature markets, but then also across geographies as to who is actually getting their hands on this thing and spending the time to be able to test and experiment and figure out recipes that work for

them. Yeah, well, you know what's interesting about that too? Contrast it a little bit with the fact that almost every survey that I see now shows that attitudes towards AI in, for example, the developing world relative to the West, it's way more positive in like the attitudes towards AI in India are massively more positive than in the U.S. And I've spent a lot of time thinking about to what extent that reflects a natural sort of property that we're starting to see of generative AI to equalize opportunity from the bottom up, you know, where it's sort of like levels the playing field in a way and that being a better thing sort of for emerging economies. Or if it's just a matter of how sort of viciously negative, the media cycle around it, you know, has been for the last year based on, you know, real concerns that people could have. but it's just like, you know, a real concern about, you know, X-risk turns into just like the favorite story that you could ever have from a Click Bay perspective for media. And, you know, that's all Americans

here at this point, I think about it. Yeah, yeah. I think, I agree with that. I think the X-risk stuff is, look, that's a big issue, right? Like, I don't have the answer. I certainly can't predict the future, but it is such a lopsided thing where like, yeah, theoretically, the thing could be so bad in the end state or has some, like, devastating consequence that even carrying one percent, about it is probably the right amount, right? And once you average it all out, it's all the stuff around, you know, I know the artist communities, the gaming communities are massively anti-AI. This is a huge question around copyright and who deserves what compensation. And I generally think our current law and economic structure isn't set up properly to kind of serve those communities.

And so I understand where they're coming from. I do think that a lot of it is the equalizer, right, both on the micro level at the individual level and then also at the state sort of nation economic level, right? On the first side, plenty of research that shows that the lowest performers without AI magically start to get to 80th percentile with AI usage, right? And that is powerful. That is very, very powerful. If you didn't have certain training, if you didn't have a certain context, if you didn't work a certain job before, you can ramp up in a matter of hours, days, weeks, right, and get to a pretty dangerous level of knowledge very, very quickly with help of AI, right? So on the micro level, at the individual level, that is game changing for an

individual at work, right? And on the economic level for an entire country, I mean, these waves are massive. And certainly a lot of it has to do with resourcing as to, in this case, you know, who has the sort of R&D talent, who has the computing resources, all this kind of stuff. But all these opportunities create room for certain countries to reorient their entire economy. economies, right? So in this case, I'm thinking of India and the Philippines, where those particular countries are huge in BPO, outsourcing, sort of offsourcing, sort of offshoring services, like these sorts of things, right? The U.S. benefits tremendously from working with those countries. A lot of that work is getting replaced by AI or theoretically could be done by AI very soon.

And so for those countries, it is a huge opportunity for them to shift their focus and reconstruct a big portion of their economy to upskill their workers, to focus on high. value add services, and that fundamentally changes kind of, for example, India's position in the world, right? And Jensen Huang spoke at this. I was at the Nvidia GTC conference. There's a very targeted question around certain countries that that will benefit. And he called that India, right? He said, like, this is a big moment for the Indian economy to look a lot more like the U.S. than it currently does today. And again, like changes position in the world in the global sector. So all that put together, look, all this is going to play out in like many, many years, decades maybe. but it is something that's very interesting for us to think about on a longer time horizon. Yeah, absolutely. Let's actually talk about the Nvidia event because you were there. So we're recording this on Friday, March 22nd. It'll be out sometime this week while I'm traveling.

But I'd love to hear what the event was like from the inside in. It seemed like this sort of crazy step change in just the energy of this community. But what was it like to be there? I'm exhausted. I will say like I woke up this morning. I was like, I am just beat. So it was in San Jose, California. Invita normally is a developer conference.

So this is just some of the contexts I picked up while I was there. It's normally a developer conference. They have not had this in person in the last few years. So the first time kind of getting back together there. And look, it was wild. It was absolutely wild. Because it normally is like this IT data center, engineer, sort of persona that goes there.

And this time it kind of just felt like everyone just wanted to be there to go meet the CEO, right? And Jensen Huang, every time he was so good about it, right? He would just show up at the expo floor. He would just be walking around casually. Every time, absolutely mobbed. Mobbed for selfies and autographs and all these different things. And he was very kind enough to sort of stop and do that.

You felt the hype for sure. And a lot of people wanted to have invidious point of view of like where, AI is going. They're obviously such a critical player. Their stock has been going crazy. They're, you know, now I think like the third largest company in the world, like something like that, right? And all the energy, I think, was just trying to figure out where all this is going. And look, like, Nvidia talked mostly about their chips, right? Like, that is the majority of their revenue. And I would bet that most of the people in that room did not know actually what goes into GPUs. they probably were thinking like me as like, oh, bigger number is better.

Like that's pretty crazy. Cool. And like more more power equals, I guess, like more powerful AI at some point. But it's not like you really knew the engineering behind it and like how big of a deal it was. So I think people are just excited to be there. And they just love anything AI related. I remember last year the first conference that had the generative AI name on it, even though it was run by a particular AI startup. up, again, people just mobbed it because they were just like, look, anything related to AI, I want to be there. And this was just, it felt very peak. And I imagine a lot of this was, you know, pretty different than the last time they ran Invita GTC in person. But certain a lot of like really, really cool discussion around all the opportunities in enterprise, in, in creative work,

in medicine, in protein and drug discovery, all these sorts of things. Right. And so it was a very, very wide sort of agenda on all the different ways that AI can be. And really, really exciting to be there for sure. In terms of like the trend lines of things going on in AI, particularly in the context of what people seem to be interested in or talking about, was there anything that was sort of, you know, common point of discussion at the event that was surprising to you versus like, you know, there's certain things that we've seen it over and over and over again are interesting to people. But anything that was maybe outside of what you would have expected. I would say the most surprising stuff was actually hearing Jensen talk about what the future, future looks like.

So when we talk about the short and medium term, I imagine the stuff that you're referring to is, look, AI agents. Like, you know, video is going to get even more insane. Like the images are going to get more insane. And like, you know, all of a sudden TikTok is going to get flooded with like all this AI generate content. Like, yes, like a lot of that was there. And that I think was to be expected. Another expected thing is all these enterprise leaders trying to figure out what generative AI means for them.

What are the resources I need? What are the technologies I should be looking at? What are the use cases I should be looking at? All that I think was also expected just because I think the current state of AI and the enterprise is everyone knows they need to be paying attention, but they don't actually know what the use cases are yet. But the surprising thing was to hear Jensen say things like completely generated games will appear in the market in five to ten years.

And more broadly, his opinion that right now the state of computers is that when you use a browser, for example, and you're, you know, listening to this podcast, what happens is like your computer goes and navigates the internet, sends a request and then goes to chase down that podcast content from a server somewhere and then brings it back. It's sort of retrieving all this information from a data server somewhere in the world that is hosting this podcast's content, right? And what he was saying is in the future, everything will just be generated on the fly. There is no more of this computer goes to talks to a server,

server fetches the content and kind of brings it back. Everything will be generated. Everything will be as if you had chat GPT on your computer all the time. And it was generating all this sort of stuff, everything that you see. And so the tail effects, I think we didn't really get into. But just to paint that very clearly for a vision of the future of what computing actually looks like was pretty wild. And there were some of these sessions, too, that talked about agents in particular and the impact of that, that I think was, again, very forward looking at it's, again, five to ten years from now, maybe even more. But just to get a little bit crisper, a more concrete image of what that all feels like was pretty crazy.

Yeah, this is a particularly interesting line of conversation and maybe to intersect it with some of what we were just talking about with the society level issues. So I find that the folks who are enthusiastic in the long run about AI's impact on the world and the economy and stuff tend to have a sense that 100x the capacity doesn't mean that we need one, one hundredth of the people. 100x the capacity means we'll produce 100 times as much stuff and 100 times as much content and 100 times as much entertainment. And that's certainly my sort of belief. I think that if you just look at the pattern of human history, like, we have a basically never-ending appetite for more variety, more options, more whatever, right? Now, I think a totally separate question, I think that the transition to that could be enormously painful, have incredibly problematic consequences, like things that we really need to address, even though I'm long-term optimistic, and I don't think

those things are mutually exclusive. But if you kind of zoom out to that future, when you start to think about, like, just 100x entertainment, for example, 100 times more entertainment, you really do start to get into a world where the thing that makes sense is this sort of hyper-personalization of content and games and, you know, and I think that the way that that plays out and how much sort of value that that places on things that can actually bubble up to the top and create, you know, shared experiences across people, like all of them are really interesting questions, but it does feel like, even if you just look at the difference between like what, what, you know, our parents had options to watch on TV in the 80s versus the entertainment options.

that we have from a consumption standpoint now, it does feel like you're sort of on this inevitable track towards mass personalization. So interesting that that was sort of a type of topic that was at the event. That's exactly right. And just to give you a little bit more color on what those conversations look like, right? Certainly, well, I was there with a group of content creators. Everyone was sort of talking about this question, which, you know, most immediately affects a lot of YouTubers, for example, right? They were all sort of like, all about to just, you know, do we have a limited lifespan on this thing? Like, are we, do we actually feel optimistic, right? So, I mean, the hyper-personalization thing, I think, is very real, right?

We already saw early versions of this. I believe it was like Carvana that did a campaign where they took two or three million customers. Each one had a data point, right, which is this person bought this specific car, it was this model in this location at this time. And so they translated all that into, okay, we're going to make a video of like this particular car model driving through a scene that with images that were reminiscent of, the location that they were in at the time.

They took what was going on in the world at the time, like certain holidays on that particular day or trends that were going on in social media, worked that into the script. And they generated two million unique videos in a matter of three weeks. Now, the videos themselves were low quality. You can clearly tell that it was like an early version of this AI thing. But they could just repeat that same exercise two, three years from now.

And it's going to look completely different, right? It's going to start to look pretty high quality. I think the broader question is, you know, as human consumers, we are going to be limited by just how much bandwidth we have in our brain to consume all this content. To give an example on the B2B side, there was a stat where it was, you know, X many years ago, maybe 10 years ago. It would only take one, two, three phone calls or emails in order for a seller of a product to get in touch with a buyer, to have some level of engagement. That number is now 10, 15, 20 touches across, so.

social media, across email, across phone calls, whatever it is, in order for you to get some response. Now, let's play out this scenario that you're talking about where there's infinite supply, 1,000x supply, right? And today, again, to paint out the impact of this, if you are a company that is selling, let's call it B2B software, and you decide to change your positioning, how you talk about your product, because it takes 10 to 15, 20 touches for someone to engage with you, on average, it takes six months from the day you start to change your positioning in order for

the market to hear your message and start to resonate with it. In that time period over the six months, you have to just keep on hammering the market with the same content over and over again. The question in my mind is, again, if you thousand X the content, how long will it take for you to make a change to the actual positioning of your product, right? does that go to two years, three years? Or somehow with hyper-personalization, and hyper-personalization also implies, by the way, new ways of consuming content.

We are no longer probably going to be viewing YouTube, the same YouTube, right? Like our YouTube experiences are going to be like Pete's YouTube versus Nathaniel's YouTube. All of these are going to be different things, right? And so the delivery mechanism is also going to change. But how does that, what do we do with the content, right? Does that mean that every piece of content only gets one view, which is the person that it was designed for?

That's a little bit different than the million sort of YouTube views that you get on a video every time, the many thousands of podcast lessons that you get. All these things change the structure of, again, when you run the supply to a thousand acts, a million acts of what it is today, it's really, really hard to predict what happens. Things get kind of crazy. Absolutely. We also have no idea what sort of counter reaction it provokes in people where. some meaningful number of people like go the opposite direction. And it's just there's so much personalization that even if I engage with that personalization in one specific area, like, I don't know, I, person X like really likes

some fantasy series. And so they just like endlessly consume fanfic for that fantasy series. But otherwise, they don't want hyper-personalization. They just want the one cultural touchstone that, you know, stands out above the rest, right? So, you know, whatever that might be at any given time. I think it's going to be really interesting to see because, you know, things never happen all in one direction. They happen in, you know, the direction and then the counterreaction all at once. Totally. I'm so glad you said that. I feel like that is, that actually surprisingly rarely comes up in my conversations.

And I always wonder why that is that people somehow tend to be over. Sorry, they underestimate society's resilience, right? Like, people will react to things and they will develop a countertrent to all this. It reminds me of Alexis O'Hanian, the co-founder of Reddit. his bet on AI is sports because of this exact argument, which is there's a possibility that people's reaction to an excess of AI generated content is to disengage from technology completely. And instead, what they're going to search for is in-person, non-tech, non-AI-related stuff. And sports, in his mind, is the biggest bet on that.

Sports as the biggest unifier, the biggest cultural touchstone of all that sort of brings people together. that's his play, right? Which is, it's just an interesting to think about like all the different ways that that society could react with this. Yeah, I mean, listen, it is a super interesting thesis. I think that it's very credible.

And, you know, the thing that's fascinating, too, is that it's not at all mutually exclusive from, like, these things both could happen at the same time, you know? The, I mean, we're also living in a world where Taylor Swift has just completely destroyed all, all records for what a tour can be. And again, it shows that the value that people place on these sort of unique in-person, you know, group experiences where you're,

you feel a part of something. And it does feel that those things, you know, what, what loses out is the stuff in the middle that are sort of sad approximations of those really great experiences, right? Again, it's just for power law distribution. But the, the stuff that really delivers against those goals can get even more valuable to people. Yeah, absolutely, absolutely. Yeah, I'll be so curious when we have the same conversation three years from now, what the world is going to look like. I'm going to be very curious what our reaction is going to be then. Yeah. So, okay, so I have a specific question about a thing that happened while you were there that was disconnected from, from Nvidia. So I think it might have been the first day. It might have been Monday or it might have been Tuesday. But the news that inflection was basically all heading on over to Microsoft hit, which I think as a, you know, as a much of dramatic impacts or ramifications that we could talk about. But, you know, what was the, what was the response, you know, among the folks that you were hanging out with at the time? The general take is, one, whoa, that is something like that happened. And then two, kind of expected,

kind of expected. I think the general take on inflection is, of course, like huge story, like the CEO and it was like, like, Reid Hoffman's involved, like all these like great backgrounds of the people starting the thing. They raised such a massive amount of money. they bought all these GPUs to build these models and they built a pretty good model, like a pretty good product. But at some point in the last few months, the narrative really has turned against them and was sort of like a, what are they even doing? Like even the product that they released, they weren't charging money for.

It wasn't very clear like business use case. It was very personal and sort of meant for consumers. And so it was kind of like this thing of just like, what is this? Is this just a front somehow to kind of give the CEO? and to give Reed Hoffman something to talk about and kind of put on the resume. So I think it came down to that. I think it was on Wednesday, probably a day after it happened. I was talking to people about this.

And I'm like, yeah, I mean, it's like, it's not like we lost anything. It's not like chat chibouti went away. You know what I mean? Like their product pie was good, but it wasn't like dramatically better or significantly different or anything like that. So I think it was just a more of, okay, well, it's very clear now that just because because you had a good background and raised a ton of money, it doesn't necessarily mean you're going to survive, right?

And I think inflection was maybe the worst offender of that, perhaps, just given how much they raised, it was something like $1.5 billion or something like that, right? And no real product, nothing, nothing they were charging significant money for. The only question that that people had was, okay, who's next? Like, this AI hype is so real. Like, are there other companies that are sort of heading down that path that we aren't talking about?

And it sort of hints at this sort of washing away thing. that might happen. Maybe it is overhyped and there's a little bit of froth in the market today. Yeah. Well, so, you know, this is, so I've actually had this, some version of this conversation with almost everyone this week that I've been doing these, these interviews with. And, you know, the, there's certainly, it's very clear that the scale of resources it takes to compete to be an actual sort of frontier model, you know, creator is enormous and even more than people anticipated, right? And so, in, Infliction, in some ways, I think, to some folks feels like just right where that line is of competition, like even that 1.3 billion really wasn't enough to compete relative to everything else. Now, a million other decisions that they made, you know, like they were tempting something new that that didn't necessarily lend itself as well to, you know, the sort of like instant turn-on business model that, that, you know, a chat GPT had or something like that.

On the flip side, you have all of these companies that, like, just there's no chance that they were going to have the sort of GPU. and compute access. And so they've all sort of designed around that, right? They're trying to do things that aren't that. And then in the middle, there's this very, it's a fairly small number of companies, but they're discreet that are like just below that, you know, sort of Microsoft funded level tier and, and but still kind of like leaning towards that direction versus, you know, being sort of more the, the smaller types. And that's the, that's the segment of the market that feels most sort of ripe for calling, let's say, at the moment from where I'm sitting. Totally. Yeah, I'll be curious which one sort of end up having the same result. I think

when I step back, inflection to me was very much and it was an execution problem. Like the resourcing was, I think, generally there. They ended up building a model that was competitive, even though it wasn't the best. It was very clearly in that first batch, as you were saying, that first tier of models that would look to match Open AI's current release and sort of be in that we're all sort of one to two years behind Open AI sort of category. Look, the problem is like you have a company like Mistral right next to inflection that completely out executed from the model building front, right? Mistral, these guys are like they're located in Paris. They just like show up like they raise a whole bunch of money and they're literally tweeting out the full

download to their entire model, everything, and the model's very, very good, right? Like, every time they release something, it is higher in the leaderboard. They're starting to compete with Facebook, with meta, very effectively. And that's really impressive, right? And so there's a little bit of like, I think when people were looking at mistral, they were kind of also looking at inflection and being like, okay, well, why can't they do the thing? You know what I mean?

They're like right in the heart of everything. They have all the right talent and leadership and researchers and all the kind of stuff. And these guys are just showing up and tweeting stuff and they're like building really competitive models, right? Like, why can't inflection do that? So I think there is that. There is this general question of like, look, if you're trying to chase open AI and build these general purpose models, it might be too late. That window might be closed for the amount of resourcing, the time it takes to attract the talent, like all that kind of stuff that that you need. The other sort of category I'm curious about is building these vertical specific foundational models.

So I'm thinking about, for example, a company that announced a fundraise this week in partnership with NVIDIA, which is called Hippocratic AI. They're building a foundational model for healthcare. And they're starting with nursing, right? Like that is sort of their application. On one hand, I generally believe that, you know, that this approach is, you know, probably fruitful. You also have folks at Microsoft, for example. And I've talked to them and they say, look, like, maybe these general purpose model, maybe like the GPT 6-7, whatever it is, just ends up being so good that you don't need vertical-specific. You don't need fine-tune models. You don't just use the one big model that can do everything, right? And the data point there that I think about is Bloomberg tried this. Like Bloomberg wanted to train their own Bloomberg specific finance model for financial literature mid-last year. And then they did a retro where they were like, okay, well, we've

built this model. We trained on all the stuff that Bloomberg has access to, a world's biggest financial media company. And they compared it. They're like, okay, well, what about this model compared to GPT4 with no sort of specific training around finance? And then it turned out that GPT4 out of the box did better than the model that they built. Right. And so this general question of just like, look, like, shouldn't we all just use open AI stuff? Like maybe they just get so good that we don't need any task specific or industry specific stuff, that is a valid question. And I do think that it's worth trying all these things. Like, I don't believe at all that these startups should not exist.

But I'm curious to see what the landscape looks like. Like at the end of the day, is AI really just kind of like cloud, right? Is it just going to be three big players? It's like sort of like an oligopoly. One or two players sort of stand up from the rest. But it's kind of like three that are like making the most money and surviving and have the competitive advantage. And if it is that, that's interesting, right? And if it doesn't enter, actually is a sort of like wide base of models in a lot of

competitors, that is also interesting. Yeah, you know, so I'll be curious to see in, again, three to five years time how it all shakes out. One of the things, it's a super fascinating conversation. And as you were talking, one of the things that I was thinking about is part of the challenge for inflection is, this is in retrospect. This is a hindsight 2020 thing. So I don't at all blame them.

but they're actually quite uncomfortably in between chat GPT on the one hand and like the character AI down to the like even less serious sort of, you know, candy and all sort of, you know, AI boyfriend, girlfriend, girlfriend thing where like they might have been right in their instinct that people were going to want sort of a personalized interaction with AI, but they didn't want the sort of like personal chat GPT. they just wanted characters. They just wanted, you know, like, I mean, if you look at how many people are using these AI boyfriend girlfriend sites, like it's crazy the numbers. The numbers around character AI are unbelievable. So it might have been that Pai's thesis was correct and it just was the wrong form factor. You know, they were too, too, too like close to like legitimate, at least for that early adopter class.

I think that's such a good comment. I love that you brought up character AI because I also don't think they're making money as far as I understand it. And it is these very, rather niche behaviors, at least non-mainstream behaviors that are driving character AI's usage. I have no, I would love to see their financials and how much cash that are burning. I think when I step back, I think the general commentary by these founders is probably correct, which is everyone has this image of that movie, Her, where it's like this thing that, again, it's a personal thing and it's very consumer grade. And is your therapist, partner, sort of friends, you know, teddy bear all at once, right?

That product probably should exist, frankly. Like, there's probably going to be some version of that. We see, I think that there's a lot of demographic and sort of earlier trends, sort of behaviorally that point to that being a thing. But what we wrote in the neuron about this is a good product does not mean a good business. Will people pay enough to sustain this business? We don't know.

Like currently, currently no, right? is my best guess. Considering today that running these AI models is still very, very expensive, and you would need to charge the same amount that chatypte or Claude or whoever charges for their products, which is they all charge $20 a month, right? And we have no idea if they're continuing to burn money, if that's just sort of a subsidy or if that's actually enough to sustain them. But it's going to be really hard to convince the average person to pay $20 a month

for this AI sort of, again, combination, therapist, friend, partner, teddy bear type of thing, I wouldn't. You know what I mean? And I like playing around stuff and trying new things. But that's that's $250 a year out of my pocket. Like, I don't know that I care enough about that thing to pay for that experience. At least it hasn't shown me that that is like significantly better at this point. Well, and you know it's funny too, just further playing this out. Let's say that the assertion that their underlying thesis was correct, but the form factor was wrong for inflection could also be the case for character AI. And that's, the actual sort of first beachhead for this are people who are, you know, that it's, that it's sex,

like the same way that normal, you know, internet stuff starts, right? That what gets people to whip out their pocketbooks is like a very different type of experience. It's not just like, oh, cool, I'm talking to Elon online, but it's like a whole different thing, you know, and that's where money changes hands in these early days, you know? Totally, totally. It was funny because I remember as character AI was blowing up, and we were seeing those insane usage numbers, right? Like I think they were reporting like two to three hour average session times. I had the same reaction. I was just like, what is going on here? Like there has to be something that is kind of like that driving everything.

And of course, when you went to the wiki on their subreddit and you were looking at all these like, oh, how to use character AI? What is it? All that kind of stuff. They just had this one section, even as early as back then, which is how to trick the character AI bot into talking dirty to you. You know what I mean? And I'm just like, okay, like this is non sort of insignificant portion of this usage, probably. I don't know if it's all of it, right? From what I observe, it's a lot of, like, anime community and like gaming community and all this kind of stuff,

which are big markets. They're quite big segments. I don't believe that's all sort of sex related. But like, there's a little bit of that motivation, right? It's at least a little bit enough to make it onto the subreddit Wiki, right? That enough people were interested in this. Yeah, it's fascinating. I think that the best reflection on all of this is that it's very hard to know these things before they actually happen, which is why I think, you know, from my standpoint, I, like the inevitable wave of failures and consolidation that's going to happen with AI companies.

I have a strong suspicion, and I think we have some evidence of this, that it will be presented by lots of outlets as evidence of sort of a move from a peak of inflated X-Pest. to a trough of disappointment because it's such an easy, you know, hero's journey kind of narrative path. And I just don't think that it's actually going to represent that. I think it's going to represent a very natural process of consolidation. Like, basically it's going to represent what capitalization and venture capital was supposed to do, which is fewer companies make Series A than got seed and fewer companies make Series B than, you know, like, and but there's going to be, you know, over the next six months, just hundreds of companies

that fold or join or, you know, it. know, because there just, there has to be. There's too many, there's 15 versions of everything that's, that's a potential business model right now. Yeah. I think you nailed it. I think you nailed it perfectly. It is mathematically, that is just why venture capital exists as an asset class, right? It is meant to fund these big moonshot ideas and the vast majority are not going to work out. The vast majority are even going to look like this, right? Where we have, again, a good product, not enough to sustain the business, right? All that's going to be true. I remember the, A friend of mine had a conversation with a GP of a pretty prominent venture firm.

And they were saying how a lot of the AI SERPs that they funded are now struggling because it's so competitive, right? Like to your point, 20 teams in every vertical doing the exact same thing. The AI phone receptionist thing is a perfect use case where there's enough so many of them doing, oh, and AI will pick up the phone for you and sort of scheduled meetings that even now, they've only been around for like six months. months, nine months, whatever it is, they all have to specialize and position themselves as the AI receptionist for plumbers, the AI receptionist for dental practices. There's just so many of them that they have to like choose that particular focus, right? And I think you're right. I think the media is going to paint it as this ultimate failure of like, oh, here we go again. Like the VC bros have

funded like yet another wave. Like you're going to try to make it look very similar to crypto probably. But I think it's fundamentally different. And it's just a matter of time. Like at the same, at the same time as I see, as I hear a lot of startups that are sort of struggling to figure out where the product actually is, where actually is the opportunity given the competition, I also am seeing a lot of companies like identify very, very clear needs. And they are backed by very specific workflow or industry knowledge that informs like they just know that there's like this particular thing that is exceptionally painful in financial services, in manufacturing and wherever. And there there's maybe only two or three teams doing it, right?

And they're sort of automating these, like, things that you and I have never heard about, these processes that we don't even know exist. Those are real businesses, right? Those are going to be going to have a real impact here. So I think, like, on the consumer front, to your point, random as hell, like consumer stuff is completely unpredictable. It's kind of like, why does the social media trend blow up?

I don't know. It just like kind of just happens, right? So I think we'll see a lot of that sort of washing out. And I think you called it perfectly. It's going to be this sort of tag versus media thing yet again. And then, you know, in a couple of years, we're going to get out of the trough. And then it's going to be at a properly level set in, not overhyped level of adoption. And that's going to be when gold hits.

Great note to end on. I could chat about this stuff all day, but really appreciate you hanging out today, Pete. For those who have not subscribed yet, check out the neuron. You will not be disappointed.