20VC: Who Wins the AI Race; Startups or Incumbents & Does Having Proprietary Data Really Matter For Startups Today?

0:00I think that there's only going to be 5 or 6 foundation model companies in the world in 3 years, 5 years. So today I wanted to address one of the core questions in both the AI and the investing world, who is best positioned to win? Is it startups with speed and innovation or is it incumbents with scale, resources and distribution? Today we hear from six leading founders and investors unpacking their thoughts on not only who is best placed to win, but also how the incumbents are doing in the race so far. Let me know what you think of this show, I really want to hear the feedback on these compilation episodes, so let me know on Twitter at Harry's Debbings.

0:34But before we dive into the show's date, you've heard me talk about how Coda is the doc that brings it all together and how we can help your team run smoother and be more efficient. I know this because Coda helps me, we have many researchers in the team who bring all the content together for the show and they need a single place to work, collaborate and collect notes, date and information in any format and that's why Coda comes in. They can help your team run smoother and be more efficient. Coda allows your team to operate on the same information and collaborate in one place. By putting data in one centralized location, regardless of format and eliminating robust, they can already stop your team in their choice.

1:10And that's what slows down productivity and collaboration. It's time to get projects across the finish line faster. So help your team run smoothly, more efficiently with Coda. Get started stay for free by heading over to coder .io slash 2 -0 -V -C. that's coder .io slash 2 .0 VC and speaking of tools we cannot live without like coder, we have to talk about Brexit. The only one financial stack trusted by founders. Founders have to think globally in order to open new markets, unlock cost savings and gain access to new talent. That's why having the right financial stack is more important than ever and that's why Brexit comes in.

1:44With Brexit you get a high limit corporate card, a high yield business account with up to $6 million in FDIC protection and build pay, all build with a global first mindset. Bratis enables you to operate in more countries and currencies than any other provider, so you can pay vendors, run payroll and make international payments faster. We all know that's a must have for start -ups at any stage of growth, so are you ready to learn more about Bratis Global First Solution? Visit Bratis .com for slash 2 -0 -VC, that's B -R -E -X .com slash 2 -0 -VC, and finally Angelist is fast becoming the centre of the venture ecosystem.

2:21The startup world is just buzzing about how fast they've been shipping products that meaningfully improve the lives of both startups and fund managers and their investors. For startups, Angelist reduces the friction of camp table management, banking and fundraising, all in one place. Teams can focus on scaling and let Angelist handle the rest. Thousands of startups have moved their camp tables to Angelist in the past year. Anjelist also supports large ranch funds and their teams with an automated software first approach and the best customer service in the industry. Fun managers can focus on making great deals while Anjelist handles reporting, taxes, compliance and more.

2:56So if you're ready to scale your startup or fund with the platform at the centre of Visit angellist .com, fold slash 20VC to get started.

3:12Now we're going to kick off the showstay with E -man Moustak, Founder and CEO at Stability AI and FunFant the most downloaded 20VC of 2023. I think it's in common but there's a lot of startups that we billion dollars and even on the thin layer thing, ITA software software for 700 million and Kyaxal for 2 billion. And that was a layer on tougher by TA. We've seen many of these examples here, right? Again, we know that value and modes are not necessarily innovation first. Well, yes and no, it seems to me I had the Tom Tungers on the show. And Tom is a very famous MLAI investor and he analyzed infrastructure by the application layer.

3:44And both actually were about $2 trillion a time. The differences in the infrastructure layer, there was three companies. And in the application layer, there was 50. And so your average enterprise value was like significantly steep. I would agree with that. I think that there's only going to be five or six foundation model companies in the world in three years five years. Do you think they've all been created now? Yes, which are they? I think it's going to be us and video Google Microsoft Open AI and a meta and Apple probably are the ones that train these models. Is anthropic good? Anthropic great, but of a business model perspective, you have Claude on Google API and you have pom too.

4:20How are they going to keep up with pom too? They can raise billions but Google spent $20 billion a year on AI. DeepMind salary budget is 1 .2 billion a year. So it deemly salary budget is 1 .2 billion a year? Yes. So that's in the public kind of findings. They takefully make a billion a year from their internal counter payments with Google as well. But again, Google, how much money do they have? $150 billion to win this? I saw it as the objective function of OpenAI is to build AGI and they reckon they need $10 billion to do it and they did that. Like they're building a business on product and things, but they don't care.

4:50They're not trying to build a stand -held business, they're trying to build an AGI. Why? What would I just tell me I'm just saying, AGI to build a sustainable business, because they're in the end of the day. They're building an AGI to turn the world into utopia. It's written in their path to AGI thing, so they think this can basically bring about utopia. Apple's a black box, right? So we'll see it WWDC, and so they could surprise us all, but let's face it, series crap. But they have all the ingredients in place, they had entity architecture, the secure wrong -clave, other things. Neural engine, a stable diffusion was the first model ever optimized on the neural engine etc.

5:20But let's see that one. Amazon have moved faster than I think they moved before. Amazon is interesting because they're an engineering organisation. So they have self -driving cars. They have satellite internet because once they've got it and they can take it from research to engineering it's there. One of the struggles they've had is that it's not moved from the research side yet. You're still evolving on research so they're like what do we do now? But they are inclusive Jeff Bezos said for his first hundred billion revenue he envisioned half of it being proprietary and half of it being marketplace and they're having the same approach with bedrock and things.

5:46Meta I think is the dark horse. I think Mark's probably pissed off the OpenAI bought AI .com so he couldn't change it from Meta to AI. But again, having him at the head, he can shift these things, right? Because the Metaverse obviously is a complete waste. They're fully in generative AI. Look at Lama, look at OPT Fair, which is their res揵 in, is leading in this field and they're pushing out amazing stuff. But who is best for a chatbot? Who has the most data for a chatbot? Meta. Again, let's see how they evolve. I think the biggest AI companies would be services based implementation companies for a large enterprise, 100%.

6:16That's why I said, if you're a startup, the best thing to do is you identify an enterprise that will be transformed by this and you go to them and you say, I have a solution and I'm going to start with you and I might go bigger, but I'm going to help you through this period by doing this and this and they will appreciate that and they'll be capital available for that in a way that you've never seen before. Because you know how difficult it is for small companies to sell to big, for the big companies have no idea except for their CEO and their board telling them. You look at the number of mentions and earnings call this like that.

6:42Every earnings call next quarter and then And by next year, literally every single one, they're like, what is our strategy? Again, it's what is our COVID strategy. It'll be that level of urgency within a few quarters. When this starts going in enterprise, it's going to be a fricking train because so much of enterprises about services and information flow. My term analysis is that a thousand companies have been 10 million in the next year, a hundred companies spend 100 and 10 companies spend a billion. Like, PWC just announcing they spend a billion over the next three years. And that's a accountancy firm.

7:09Where's that going to go? They don't know nobody knows. And so multiple of that will be allocated to this as the only growth theme in the entire market against the backdrop of rising rates real estate crashing etc So the amount of capacity versus the amount and whale and wall of money into something that's growing faster than anything We've ever seen is completely mismatched and what will I cause already you're seeing GitHub stars leading to $100 million funding rounds with zero traction and zero business model likes to really we actually have a business model I think this will be a big event on the impact than covid I don't know in which direction, but again, you lose BPO jobs, you can make it up an entrepreneurship, you embrace the technology.

7:45From the founder side of the table, we're moving to the investor side of the table and hearing from Vince Hanks at Thrive Careple, who led the deal for Thrive into Open AI. I have Tom Tungers on the show recently and he's had the Google have been the most disappointing of all. We've seen AWS partner with Hugging Face. Who do you think is doing well? I think it's hard to be dismissive of these companies. If you think about where the best talent in AI is right now, I think it's open AI. And then I think everyone would tell you it is Google and Facebook and Microsoft and Amazon. And maybe there are folks that have kind of dripped their way into the startup ecosystem, but by and large, the talent is so clustered in these big tech companies.

8:22And so I get it that they're tripping over themselves, trying to figure out how to navigate these giant organizations they've created. But let's not also kid ourselves. We've talked about this maybe outside of the context of open AI, the up level to AI. We work with lots of startups that compete on things like presentations or content creation What's scary to me right now is if you're a startup large companies are shipping product the kind of canonical examples were that Oh the big incumbent can't move and they're slow -footed and you got years to execute before they do something Microsoft's 200 ,000 person company.

8:54They've shipped AI and Bing AI and PowerPoint they're rolling out of the other products Adobe's got the content creation in their product already, even the big startups like Notion. Those folks I've been so impressed with how much they've shipped in their product so quickly. And so if you're a startup, I get that you have the advantage of speed, but you need a multi -year execution window to be able to take advantage of all that product you can ship quickly. If these big companies are able to ship this quickly, they've got so much distribution, so much talent. I just think it's going to be very hard to compete with them.

9:24If you're going toe -to -toe right now with an incumbent on their home turf and their shipping, I think it's a high -speed. hard bet to take the opposite side of, you know, in our business. But if you're creating something that's totally a different user experience that no incumbent has today, I'd bet on the startup 10 times at a 10 in that case. Well, why didn't we hear from one of those startups? And now we're going to hear from Clam Delaying co -founder and CEO hugging face. Maybe. And that's why there's a huge opportunity for new companies to disrupt the incumbents because they're going to go for like the easy solution, whereas like other companies that are more like AI native are going to be going for the more disruptive approaches.

10:00We see that we would have startups that are using PuggingFace if you look at models closer to RunWayML or StabilityAI or like PhotoRum in Paris. You see these AI -native startups that are actually building training their own models and how they can in my opinion build much better things than the ones that just use APIs. What you're describing is a great opportunity for AI -Native startups to disrupt the incumbent queer solutions, even if it's not sustainable in the long run. We are hearing some very different opinions in this episode, and I want to switch back to the investor side of the table now, and here from the man himself, Tom Tungers, at theory ventures.

10:41When we look at the two announcements spectrum for the next generation of AI and LLM, which is when existing incumbents integrate it well enough into their distribution channels to be highly effective and continue their dominance, or actually a start -ups with agility, flexible code bases, much better place to win in this next generation. How do you think about that kind of start -up versus incumbent megal wall? My thinking's evolved here. In the beginning, I thought the incumbents were going to win the whole thing. And I'd like that because the incumbents have far greater distribution, Microsoft has an incredible channel, Microsoft has a special relationship with OpenAI, the pace with which Microsoft is injecting its products with LLMS is astounding.

11:21But Dowie, it doesn't have the regression it deserves when it comes to using generative. I think about the applications in Photoshop, they launch a product called Firefly. I think they're right there. And so startups are in this unusual position where they have negative time to launch. They're actually behind the market, which is unusual. I think about mobile apps and the launch of the Apple Store startups were the first ones to understand how to write mobile apps with objective C. But I think any time we talk about machine learning, there's always this question around what is the mode? And I have this this reaction, which is like to data mode, to data mode.

11:48And I think the answer is the one that it's always been, which is better execution is the mode. If you can build a better CRM and get it into market, you can win. You take a look at what notion has done with documents or what snowflake did with databases facing too big incumbents. There are these stories. They're all over. They create this beautiful constellation within startup land of the David versus colliah's story. I think if you're a venture capitalist or if you're a startup founder, you have to believe. I think it's in your fabric that no matter or how big the incumbent is or the advantages that they have that if you have really great execution, you can still win and you can win big.

12:21I think at the foundational model layer, that's a big boys game or big girls game. It's because of the capital intensity required both for training and the GPU access and all those kinds of things. So maybe there's a startup or two that's able to raise a couple billion dollars in order to compete. Now Tom said the word that defensibility. He said the word mode. I wanna understand now how proprietary all data models and data sets and if a startup does not have them, Designing that they're inherently disadvantaged, this segment with MAD is fantastic on this. The reality is no models that are out today will be used in a year.

12:52Palm last year was 540 billion parameters, then Chinchilla 67 and now 14. 540 to 14 is a big step. You see the quality of GPT -3 versus GPT -4. Is there any extent to how low it can go? We had no idea. You already said this is impossible. Two years ago, like, no way. You have a single file that's maybe a few hundred gigabytes that can pass every exam apart from Englishlet, we need to feed these models better data and other stuff and that's why we're moving so hard as stability. There should be no more web script data near. There should be national data sets, the good quality to feed these three range organic models and national and proprietary models and others.

13:23There is no such thing as an unbiased model. Dali too, when OpenAI had that, they introduced a bias filter. Any non -gendered word they'd rather random gender and a random ethnicity. So you type in Sumer Eslone, you get Indian female Sumer Eslone. That was a good feature. I got to save somewhere. This is where you need national data sets, you need cultural data sets, you need personal data sets. They can interact with these base models and customize to you and your stories because you and I both have our stories that make up our psyche. And that's why when the research I find that letter, I think there's a six -month pause to get all of our shit together.

13:55Before things go completely insane and next year this is everywhere and everyone's investing in everything and it's just absolute chaos. You will have national champions and others, I think it's incredibly difficult to compete in proprietary. I think in open is a bit different because of standardization element there. But again, my play is to be the benchmark across every modality because there's no other company apart from you know, opening out that does every modality. There's no company that's as aggressive as me in emerging markets. And so they have to say, what is my edge because you can have an edge like you can be the open AI for government or defense or for health care and really get in and understand those and then you can be sticky.

14:30Like what of the scale is now going fully into defense. They've announced the integrations with the Air Force and all sorts of other things. What is your edge? What is a Kenya moat? What is your business model? And what are you liant upon to deliver that value that can increase? We mentioned the quality of data sets and data models that I don't want to move away from the investor and the founder side now, really to the research side. And there's no one better than this than the Anilakun. I've interviewed many leading AI experts and they say the value will accrue to the incumbents. Is that right?

15:00Will the value accrue to the incumbents? Or do you believe that given what you just said about size and not being everything in terms of models, it could be startups as well. So this scenario, I think will happen, and I'm certainly rooting for, is this scenario I described earlier, where you have some sort of open platform for base LLMs. So base LLMs basically would be seen as a basic infrastructure, TCP IP, Linux Apache, completely open, and then there would be an ecosystem of companies building stuff on top of it, which for vertical applications for specific things, to specialize those systems for particular applications, to offer support, to make it customized for enterprise applications, for personal things, there'll be a whole economy around this, which will create jobs, by the way, not make them disappear.

15:39So this is the scenario that I believe will happen. And the reason I think it will happen is because there is essentially a need to use, essentially millions of contributions for making those systems factual and correct, etc. So Wikipedia style. So I think the proprietary approaches will actually fall behind. So that's one point, okay. The second point is you can ask yourself the question, How is it that the companies that were best positioned to produce something that Chad GPT, namely Google and Meta, didn't? Why is it open AI? A small outfit with 400 people. And the answer is, it's not because Google or Meta did not have the competence or the technology.

16:14It's just that they didn't have the pressure to produce completely new products that had a lot of risk attached to them. And we know where the risks are because a few weeks before Chagin PT, my colleagues at Fair produced a large -range model called Galactica, which was an experimental system. So Galactica was a large -range model trained to train on the entirety of the scientific literature. And it was basically designed to help scientists write papers. So you would start writing a paragraph or something like that to describe the top -tiered paragraph. And then Galactica would basically complete the paragraph.

16:47And it wouldn't be factually correct. it would have to fix it, but it would like you would ask it to build a table of results and it would just put the light take commands to build the thing and populate it with the known results on the literature about the topic that you're working on or you would type a chemical formula for something it would turn it into an actual name for it, very useful for scientists. As soon as the demo was put out it was murdered by the social network Twitter sphere. People said oh this is going to destroy scientific publication because now any random person can write a relatively sounding scientific paper that is nonsense.

17:22And there was so much vitriol thrown at the system that the people at Meta who built it couldn't take it. They took down the demo because they said we can't sleep at night. So here is an example of a very useful system, a system that could have been extremely useful, particularly for writers or scientific papers who are non -native English speakers, that basically was destroyed by AI donors. People who just did not think about the risk -benefit analysis. The risk of flooding the literature with nonsense is ridiculous because scientific publications are vetted and things like that. There was not a significant danger.

17:53And then Chagypti came two weeks later and was welcome as the second coming of the Messiah. When those I tell you, and then, a few more days later, Google came out with Bard and in the demo Bard made a tiny minor factual mistake about some astronomical fact and Google's stock went down by 8%. Now what I tell you is that when something is produced by a large company that has a reputation, particularly a reputation to defend. They can't put out things that's few nonsense, but it's okay for a small company. That's the landscape of what happens now, which is why I think there's a bit of a paradox which is that the companies that have the best technology basically can't have difficulties putting it out because of those legal issues and public image.

18:33People are kind of extrapolating if we let those systems do whatever we connect them to internet and they can do whatever they want. They're going to do crazy things and stupid things and perhaps dangerous things and we're not going to be able to control them and they're going to escape or control and they're going to become intelligent just because they're bigger. And that's nonsense. There is this idea somehow the desire to and the ability to dominate is linked with intelligence. A statement that a lot of people are making, including Jeff and Tim recently, that somehow as soon as a machine becomes intelligent, it becomes uncontrollable because it's being smarter than us.

19:06It can influence us in ways that we can't even imagine. Now I think this is a gigantic fantasy, even within the human species, it is not the smartest among us that want to dominate the others. To dominate other entities, you don't necessarily need to be smarter than them, but you need to want to dominate them. This is not something that every intelligent entity is going to do spontaneously. We do it as humans. The desire to influence others was built into us by evolution because we are a social species. Same as baboons and chimpanzees and wolves and dogs and etc. It's not the case for wrong with tongues.

19:40Long tongues don't have the desire to dominate anybody because they are non -social animals. They are solitary animals. They are territorial, in fact. We need to separate those two concepts, the will, the desire and the ability to dominate on one hand and intelligence on the other hand. The fact that we're going to have super intelligent machines at our disposal means that every one of us is going to be like a business leader, politician or an academic with a staff of people working for them that are more intelligent than themselves. It's great. If you feel threatened by being the boss of other people who work, work with you, but are smarter than you, you're not being a good leader.

20:13And I want to bring it back now to the core question of startups or incumbents who are best place to win, and we're going to hear from Sarah Groh at Conviction Capital now. How do you think about the startup versus incumbent? Who's best place to more challenges to each face? Yeah, this is maybe a very discouraging answer, but I believe in like intellectual like classically, the only real advantage startups have is speed. Speed actually might matter more than ever when the environment seems to be moving at warp speed. Right? What's the quote? Some decades, nothing happens, and some years, a decade happens.

20:45Right? I feel like that is happening right now. It's hard to make a large organization move at that speed. On the incumbent advantage side, much ado has been made about this idea of a data moat. But honestly, there's a lot of data out there, and entrepreneurs are incredibly creative about collecting it, and increasingly about generating it. And I don't think it's, ah, the incumbents are going to warn this one or the startups are going to win this one. Now, as I said at the beginning, I want your thoughts and feedback on shows like this. I really like it when we unpack a topic and go deep, bring together thoughts of experts that we've previously interviewed.

21:19Let me know what you think, as I said, love to your thoughts. But before we leave you today, you've heard me talk about how Coda is the docs that brings it all together and how we can help your team run smoother and be more efficient. I know this because Coda helps me. We have many researchers in the team who bring all the content together for the show and they need a single place to work, collaborate and collect notes. Data and information in any format and that's why Coda comes in. They can help your team run smoother and be more efficient. Coda allows your team to operate on the same information and collaborate in one place.