Bringing AI to the Masses with Adam D’Angelo
28:27 Share

Continuing on the current paradise. If you just play this forward, there's so much further that I can go. We're building a network for people and at all share knowledge together.

And sometimes the people will be getting knowledge from the A, I. And sometimes the A I will need to get knowledge from humans. There's just such a huge base of needs people have and such a huge space of different like input you can combine to try to address those needs.

Humans are always onna play some role. There's knowledge that people have in their heads that is not on the internet and is not in any book. And so no, L, M is gonna have that knowledge.

Hello everyone, this is stuff if you didn't recognize the voice before mine, that was adam dangle. Is the cofer of cora now building A I aggregator poo and so much more, including being on the board of OpenAI. We do have a very timely episode for you today, so I won't keep waiting any longer.

There's a extensive growth fund on our partner sera wang, with a problem introduction of puts on deck, will also hear her reference R A I revolution series, but you can dive into more deeply at asic scene without com flash ai. As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any asic sensei fund. Please note that a six sense e and zoho oates may also maintain investments in the companies discussed in this park cast. For more details, including a link to orm investments, please see a extended outcome slash disclosures.

Generate A I has kicked off a paradise ft. That is already transforming our world in our A I revolution. We talk to the people who are actually building the technology to understand where we are, where we're going and the big open questions in the field.

I guess this episode is adam d. Angelo. Adam has built in grown companies that have connected billions of people across the globe. He was a co of facebook until two thousand eight before found in cora. In two thousand nine, after being interested in A I for decades, adam joined the board of OpenAI in two thousand and eighty, and now he's using his experience scaling some of the biggest consumer companies in the world to build poe, a platform that brings A I to the masses.

And this conversation, adam speaks with a sixteen z general partner and my colleague, David George, about building A I infrastructure for creators, the multi model, multi model future of A I, how A I will shape knowledge sharing on the internet and much more. There's a lot to dig into here. So let's start at the beginning of Adams's journey AI. He's gone to take us back to his colleges in two thousand five.

I was very excited about A I early on in my career, remember trying to build some A I products in college actually. And IT was just very difficult. The technology just wasn't there wasn't at the point where you can be able to make something that was ready for consumers.

And meanwhile, I just watched social networking start to boom and can actually look at a lot of social networking technology as so was an alternative to A I. So instead of trying to get the computer to do everything, you could just connect people with other people over the internet who could do those things in the same way that, like globalization, can be a substitute for automation. Also not working.

You think of IT is letting people access everyone else in the world for entertainment, for fun, for communication, for whatever you want to do. I think was incredibly powerful technology. And given that A, I wasn't quite there yet, that was the main thing that there was to do to apply all the technology.

So I first got interested in social networking. And through my experience at cora, we started out with a product that was entirely human driven. So people would come and ask questions, and they would put topics on them, and other people would sign up to answer questions, and they would tell us about what they knew about by tagged themselves with these topic.

And we will try to route the questions to the people who knew about the predict topic. And IT was all manual, but we knew that at some point, we're going to go to the point where software would be able to generate answers. We want to make experiments using GPT three to generate answers and compare them to the answers that humans have written.

And cora, and a lot of the time, GPT three could not write as good of an answer as what the best human answer was that have been written, but you could an answer instantly to any question. And that can strain for corr had always been the amount of time that high quality answer writers had to answer questions. And so think that was really new about l ms.

Was the ability to extremely low cost generate answer instantly to any question. Through that experience, we realized that a chat kind of experience, where you can write a question and and get an answer instantly from A I was more likely to be the best paradigm for interacting with a ee, as opposed to this kind of look like publication paradise, of course. course. And so based on all that, we landed on building poe as a new chat or at A I product.

I think many people be familiar, but explain, just for us the product, how you find in the first place.

how do you interact with the in the same way that cora aggregates knowledge from many different people who have knowledge, want to show knowledge. We want poe to be away for people to access A I from many different companies and many different people who are building on top of ai. And so you can come to poll and use IT to talk to a very wide variety of models that are available today.

And then we have all these other products that people have built on top of these models. And we've got an open A P I. Where anyone can, who can. So anyone who's training their own model, so any of these research teams, anyone who's doing fine tuning, they can take the model and put IT on power. And what we allow is for them to reach a big audience quickly.

So we thought about as a company, cora, what is the role that we're onna play in this new world with A I and what are the strength that we had? And what do we learn over the past ten years building an Operate in cora? And there's actually a lot of this kind of like consumer internet know how that's important in getting a product to mass market to this. Things like building applications across I O S and android and windows and mac, localization of the interface, A B testing subscriptions, all these other kinds of small optimization that you need to make a good consumer product. We want to be away for anyone who's creating A I, whether it's one of the big labs or an independent researcher, we want to to be away for them to get that model and make IT available to mainstream users all around the world.

There's a lot that you just said that I would love to go deeply on. So one of the things you said you sort listed off all the models that you make available there is one theory, which is one model, one company is gonna provide everybody the solution that they need for everything. There's another theory, which is there's going to be tons of different models for different use cases.

The world's going to be multiple del and multiple del. The three behind poe is that the future is going to be a multiple del in multimodal. Why do you think that's the case?

I think nobody knows how the future is gna unfold, but we think that there is gonna be a lot of diversity in the kind of products people build on top of these models. And in the models themselves, think there a lot of tradeoffs involved in making one of these models. You have to decide what data are you gonna train on IT, what kind of find tuning are you gonna do, what kind of instructions does the model can expect you to give us a user.

That's what kind of expectations you want to set with your users about what to use the model for. And I think in the same way that the early internet had this huge explosion of different applications, I think we're going to see the same thing from my ye. So early on an internet, the web browser came along and made IT so that anyone who is making an internet product, they didn't need to build a special client and get distribution to people around the world.

They could just build a website. And this one web browser can visit any website. sure. And in the same way, we want people to be a single interface that can be used so that people use that to talk to any model.

We're betting on diversity just because there are so many talented people around the world who are gonna be capable of tuning these models. You can n the open source models today. There's also products from opening I and anthropic.

And I think google close to having something will you will be able to find tune all these models. And everyone has their own data sets. Everyone has their own special technology that they can add to the models. And I think through the combination of all of this, we're going to see a very wide diversity of things you can do with the I .

there are two things that i'd like to maybe go deeper on there. So one is the idea of what constitutes the product itself, what is IT today and what is that going to have to become and secondary, the idea of the long tail right bet on the long tail intent of them, give them a platform, abstract away a bunch of the infrastructure that they don't know how to build and the honors, really what they're great at, right?

So on the first, what is the product today? The AI model many people probably save is largely the product. What are the advances that you anticipate seeing that are going to have changed the way people interact with these, enable new kinds of products being built? One way of thinking about that is or the model providers themselves going .

to be the ones that built all the products. If you're a large model creator and you have tens of employees that you can allocate to building a consumer product and you have the the culture to do that, you can go direct to consumer and you can build a good product. I think most of the people who are training these models are not in that position.

If you want to take your model and bring IT to consumers around the world, you ve got to think about I S P. That's top apps. You need a web in your face.

You need to do building in all these different countries. You have to think about taxes. And there's just a lot of work. And you raise some venture funding. You can either spend some of that funding on hiring out whole team and developing all those competencies or you can spend that on making your model even Better. And I think different startups will choose different paths here. But I think for a lot of them, the right path is going to be to just set up an A P, I, or again, the poll A P, I, and use that to get to a lot of consumers very, very .

quickly. Yeah talk about the role at the sort of long tail of creators then place like how do you want to engage with them and what's the incentive for them to want to build on top of play?

yes. So we have a revenue sharing program that allows people to to get paid as a result of people using their bots on IT cost issues, amount money to provide inference for these models. And so almost no other platforms provide this kind of revenue share today.

So if you have a model that requires lot of gp s to do inference on, then this is really your best place to come. And you can have a real business. You can cover your inference costs and make more.

I think a ton of innovation is going to come from these companies. There are other companies that are building things on top of some of the big models, so say, from OpenAI. And in that case, they have to pay the OpenAI inference cost, which is another sort source of need for money.

And so the pole revenue show model works in the same way we'll let you. Afford your cost that you're then paying on to any other inferences provider? Yeah, absolutely.

What are some of the really fun and interesting things that creators have already built on top .

of a all of people right now excited about image models. We have uh stable 的 fusion A C X L， and then we let users go and do some prompting to customize IT to provide art of the particular style. So there's this like may style S D X L bots on poo.

Those are popular. There's this company called playground. They're making a product for people to added images, but in the process, they've treated a pretty powerful model. And they have that model available on poe. And that's tone pretty popular recently yet.

So cool that you can have a long tail of these creators make their own sort of opinionated style of these base models. But I think there's something to that where you provide this sort of infrastructure and support and then let the users of creators do .

what they do us. Yeah and it's super early days right now, but I think we're onna see over just the next year too is going to be incredible. This will go from being sort of useful to some people right now to being something that just critical to many different task .

that anyone is to accomplish. Both know very well, which is robo ks. Right early days, creators were on their building games, and they were pretty basic. The early days, and IT was a lot of kids learning how to go games. And then I sort of graduated eventually to people who were able to earn a living. So I think the ideal for you would be, you build enough scale, they can build large enough audiences to actually be sort professionals what they are doing yeah.

we're spending millions of dollars already on on inference. It's mostly going to the large model providers right now, but we want to let as much of that as possible go off to these .

independent creators. cool. I want to shift topics and get maybe a little bit more like conceptual AI your c to a facebook at the time where social was emerging and then rate the platform shift to mobile was taking place, right? So i'd love your thoughts on what are the similarities to the shift to mobile in this A I wave and what are .

some of the big differences? Yeah, you know I think it's very hard to say. I think with cora, we were a little bit slow to adopt mobile. Mobile was one of the things on our list of many priorities, and I needed to be done number one priority. And we needed to make tougher tradeoffs to prioritize IT.

We needed to do things like hire a set of different people who are gona focus on IT and really have a period where we released ed, no new features. And we were just simplifying things because the mobile U I called for a different experience. When you have such a critical change in the platform structure, you need to rethink so much that it's only gonna happen if you have this very strong kind of top down leadership.

And so you ve done IT differently this time around.

Yeah yeah ah so yeah.

Talk about some of the organisational changes and what you've done to actually refocus yourselves on the big thing that's right in front of us here.

yes. So I think the first thing is identifying this trend and and starting off doing some experimental early on just to learn. And that didn't require any kind of strong decisive leadership as much as I just require paying attention to the market.

But then from that experiment, that got us enough conviction that in our case, we said, hey, too much of the correct product has been built up around this publication model that is sort of fundamentally premised on the idea that expert time is gone to be scarce and the A I L, M time is not scarce in the same way. And so we need to rethink that. This was in, I think, August of twin two.

We got to this conclusion that chat is the right part time for this, and we need a new product. We don't want to try just trying to retrofit everything into cora. We thought we're going to move too slowly. So we so we had a small team to start working .

on poll based on that. Talk about the relationship between corn poll, how you actually envision that changing in the future? And then maybe there's even an extra position of OK corn pow. And like human experts and A I experts answering questions, do they do in the same place, is a different way of interacting.

Yeah, yeah. We d love to have all of this as integrated as possible. And think, if you think about maybe the relationship between facebook and facebook, these are two products built by the same company, but they share a lot.

I think that poe cora might evolve to a similar kind of relationship. We'd love to get more of the human aspects of cora into poe. We'd also love to get the whole core data set into the pool.

But and we're also working we've launched on this already to get some of the poll A I to generate answers that are available on cora. As these models continue to scale up, the quality is gna go higher and higher to the point where IT actually will be as good as human quality in a lot of cases. And so the core appa, i'm actually, I think, becomes more appropriate for A I as the cost of inference hired more quality.

Yeah, yeah, yeah. So we'll see what the exact right relationship is. But we think of this is we're building a network for people, and A I all share knowledge together.

And sometimes the people will be getting knowledge from the A I. And sometimes the A I will need to get knowledge from humans. And we'd love to be as much of a kind of IT for that as possible.

Yeah in cora or pro, depending on how they interact, is a place you get answers. And sometimes your answers is going to come from an expert. Sometimes it's gona come from a eye yeah right.

What do you think about just the internet, like you strap ate that out? People going to be engaging with this collection of bots that have different personalities in different expertise, and will also be inter sperits with real humans. Will, like real humans, be interspersing in the A. I. What do you think actually happens?

Personally, I think that humans are always gone to play some role. There is knowledge that people have in their heads that is not on the internet and is not in any book. And so no L M is gone to have that knowledge.

So ah kAndra brothy called the LLM lossy compression internet. So this is the internet. There's experts that know a lot of stuff.

It's not that. So I think there's a lot of potential in the kind of interplay between humans and the LLM going forward. Also says the problem with illustrations right now. And I think the right is going to go down as the models get Better, but it's never gonna a get to the point where it's one hundred percent perfect. And so I think there will be a lot of value placed on the idea that the source of your information, which human said or which publication originally printed IT. And I expect that, that is going to lead to some kind of products or some kind of user experience where the l line is helping you sort through your sources and quoting exact experts or exact sources as opposed to just synthetic sizing at all and giving you something where you can't exactly trust where IT came from.

Yeah and is that a new technology that gets built outside of the models themselves? Or do you think that that's incorporated inside of the model?

I can see IT going either way. I mean, if you just look at a model, the raw model doesn't have access to these other database where I can get exact words and so it'll have to be some augmentation of the model. But how tightly integrated into the model, I think we all know that.

Yeah, agree. I think that's gonna critical. It's one thing like we ve started out with these use cases of companionship and creativity and like fluctuations are a feature of that, right? Like that makes a more fun and exciting, especially when you get into business use cases or more utility types stuff.

It's obviously needed. What are the other big advances that you are excited about? Just probably a space for language model.

I am personally the most excited just about scale, just continuing on the current paradise. If you just play this forward, there's so much further that I can .

go and you think the sky laws will hold or holding.

So far, they have held my prediction to be, there are some issues that need to be overcome, but there is just this incredible industries s so many talented people right now who are trying to make this technology advance, and there's so much money behind IT. The force is there to help overcome any roadblocks. S that we had is so massive.

So I expect that it's just gona continue. I think there will be road bumps and issues that we need to be worked around. They'll be breakthrough of people needing incredible creativity.

We have many of the smartest people in the world, the most determined people in the world, the most talented people in the world, all focused on this problem. And I think we're gonna continue to see the kind of expansion, al growth progress that we've had so far. I think they gone for many years.

Do we talk about the last shift, right, like the mobile shift that you live to the room and some of the lessons that you had from IT? What do you think ultimate market structure looks like in the genre I space?

In order to train these frontier models, you need billions of dollars of capital and you need many years of investment in infrastructure. There is a very small set of people who can do that. And so that's leading to this world where there's only a small number of players that can be on the front here.

And so right now, it's open. Google, maybe anthropic, maybe meta can be there. Those who can get there.

I think it's going a good business. You'll be able to make a lot of money. You can have good profit margins, also work very hard to stand the frontier to keep up.

But it's not a commodity. I think when you go six months behind the frontier, definitely one year, it's brutal. There is just way too many people that are able to get the capital and the resources to train those models.

And so it's gonna be either fully open source or there will be too many different competitors for anyone to make a good business at that point. On the pure technology, I do think they're be very good businesses at that level that you are not using frontier models but are combining some other kind of unique thing with the model. So you might be that you're providing some tool that the model can use or you have some unique data that you're using for fine tuning.

Or there might be some unique product you build around the model and then that ends up being the source of competitive strength. So I think there's going to be this kind of choice where you're either competing on scale by being on the frontier or you're competing on some kind of like feature differentiation and you don't need a frontier model. And in some cases, you've both so know you I P able to use the opening I A P I and combine that with some unique tool that you're providing that to be a good business as well.

Yet once you get beyond the foundation models, you get take more traditional forms of business, different cities, competitive differences like computer advantage, source of modes and things like that, which I think .

totally makes them yeah. I think what was interesting about this is that is evolving. So things are moving so quickly. And so every six months, the frontier moves forward. And so the frontier players, they have to invest more capital, but then they have much more powerful models that open up even bigger markets. But then the open source one year back frontier that also ah the markets that can address are getting bigger and bigger. So I think every year that goes by, we're onna have this much larger market that can be addressed by the technology and and all the products that are built on top of IT.

So yeah, that sort brings me to another topic, which is related to market structure, which in commence versus startups and see the wear. And we hope the start up always win. But in the last cycle and maybe just from A B to b lands here, like a last I sas and cloud, there were a bunch of things that made IT really difficult for the incomes to actually innovate.

There is a business model, innovation and new talent technology required, which opened the door prey widely to startups. There's a take out there now, and A, I, which is this time is different and the comments are the real winners, right? Because the technology is available by simple API.

You can ploy IT right in and they have distribution so they should be the winners. And if you just sum up microsoft and google's business apps and all these things, it's probably somewhere between ten and twenty billion dollars of revenue over the next one to two years and years. If you have to take on that, if that's consistently that are you see IT or if you see IT different?

Yeah, I think it's going to a very so definitely and comments. They are going to access to the technology and they are going to have distribution. And so that's a big advantage that they have. I think the opportunities for new players in this wave are more in the cases where the kind of product you want to build around this technology is somehow fundamentally different. What was built before.

And so as an example of the hallucination problem, that in some ways, a good thing for startups because a lot of the existing products out there have zero tolerance for anything that's going to have a risk of producing something wrong. And so you can see this with, I think, with perplexity getting share from google right now. Yeah, google can't just go and put something on all the search results where IT has a few percent chance of being wrong.

That would be a huge problem for them perplexity. That can just be the expectation when you're using that product. That is almost always right, even though there is a small chance that is wrong. I think that same thing is actually gna play in a lot of other cases where the promise you build around this, they need some kind of fault tolerance, and there needs to be a user expectation that everything .

is not perfect and the cost advantage can be so great for this, right? If you take a highly paid person like a lawyer, and you run IT through an analyst on which I could cause sense versus a thousand box an hour, maybe you just should have a really high fault tolerance and you just have to do. And that's just a different engaging.

yes. And so you have this companies that maybe have a very strong brand of never making a mistake or never messing up, always being reliable and a start up can just come in and say, okay, well, this is gonna cost a ten or a hundred of the Price, but it's onna have a small chance of getting things wrong. And a lot of people would prefer that, but it's a real problem for the incoming .

about A I guess, just to close IT out, i'm sure a big part of the audience here, his founders were building and probably earlier stage in you, what what voice do you have for people building .

in A I think what I would do if I was starting a new company right now to spend a ton of time playing with the models and playing with integrating them with different things. There's so many different inputs you can give to the models. You can make scrapers that in just data from anywhere, you can get data from the users local screen, you can get date from voice.

And there is such a huge space of needs people have and such a huge space of different like input you can combine to try to address those needs. I think it's very hard to just think top down about where there is demand in the market. I think experiment is really the way to go to to generate ideas and to set up and start up going be able to build something really valuable.

Yeah have a place in for sure.

If you like this episode, if you meet IT this far, help us go the show, share with a friend, or if you're feeling really ambitious, you can leave us a review at rate this podcast 点 com slash basic C Y。 You know, candidly, producing a pocket can sometimes feel like you're just talking into a void. And if you did like this episode, if you like any of our episodes, please let us know. I'll see you next time.