Does AI Have a Capital Concentration Problem?

Today on the AI Breakdown, a discussion of AI's Uber problem. 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 Discord and our newsletter. Welcome back to the AI Breakdown. It is a weekend and that of course means it is a long reads episode. And today we turn to an op ed in the Information by Tim O'Reilly called AI has an Uber Problem. Tim writes, a handful of deep pocketed investors distorts the market, fueling a race for monopoly that inhibits product market fit for those of you who are unfamiliar with Tim O'Reilly, he is a long time Silicon Valley denizen. You may recognize him from O'Reilly Books, which are how a lot of people used to learn coding and other technical concepts in the eras before copilots and I think this is a really thoughtful piece. So I'm going to do what I normally do. I'll read some sections of it, excerpt others, and then have a bunch of conversation. Tim begins with a Hayek the economic problem of society is a problem of the utilization of knowledge which is not given to anyone in its totality. The piece then begins Silicon Valley venture capitalists and many entrepreneurs espouse libertarian values. In practice, they subscribe to central planning. Rather than competing to win in the marketplace, entrepreneurs compete for funding from the Silicon Valley equivalent of the Central committee. The race to the top is no longer driven by who has the best product or the best business model, but by who has the blessing of the venture capitalists with the deepest pockets, a blessing that will allow them to acquire the most customers the most quickly, often by providing services below cost. Reid Hoffman called this pattern blitzscaling, claiming in the subtitle of his book with the name that it is, quote, the lightning fast path to building massively valuable companies. I disagree. It is a dark pattern, a map to suboptimal outcomes rather than the true path to competition, innovation, and the creation of robust companies and markets. As Bill Janeway noted in his critique of the capital fueled bubbles that resulted from the ultra low interest rates of the decade following the 2007-2009 financial crisis, capital is not a strategy. Venture capitalists don't have a crystal ball. To the extent that entrepreneurial funding is more concentrated in the hands of a few, private finance can drive markets independent of consumer preferences and supply dynamics. Market discipline is significantly delayed until the initial public offering or later. And of course, today IPOs are delayed often precisely because companies can get all the capital they need for from a small number of deep pocketed investors. Founders and employees are even able to cash out some of their shares without having to face the scrutiny of public markets, much as if bettors on a horse race could take their money off the table as the horses round the first turn. Thus far from finance being an extension of the market, with lots of independent signals aggregated to ensure competition and consumer choice, capital can ignore the will of the market. So first of all, back to NLW here. Before I get into the next section of this piece, I want to tell you that as someone who lived in Silicon Valley through a big chunk of the ZIRP era following the great financial Crisis, and who has also spent a ton time on macro level analysis, it is remarkable the extent to which and for how long. Silicon Valley didn't really consider macro dynamics as part of what was going on in its own market. For at least the first half of the teens, and honestly probably longer, people talked about increasing valuations of startups and them staying private later, as though it was some phenomenon of startups rather than an issue created by the mass availability of capital that pushed many more types of actors who had previously never touched the venture capital asset class further out on the risk spectrum looking for yield. When more and more capital moved into the private equity and venture capital realm, it meant more startups were able to hoover that up. Which meant, as I'm anticipating, Tim is going to get into a lot of weirdness in how business models were constructed and what the game of winning actually looked like. Tim's next section is called How Capital Distorts the Market. The example he focuses on is the Uber Lyft example. He writes, the ride hailing business began with bold prophecies of ride hailing, replacing not just taxis but all private vehicles, and ended with a national duopoly of on demand taxis. Prices know better and often worse than those of previously over regulated local taxi markets. Tim writes that he believes in a different era, in a different time, what competition would have looked like would have been very different. In an alternate history, he suggests, entrepreneurs might have competed around pricing different rate structures for drivers, or even explored totally fundamentally different business models. But he says that didn't happen because the era in which Uber and Lyft were born was an era in which there was so much capital sloshing around the system that venture capital effectively became a subsidy for customer acquisition, a fake and ultimately temporary cost lowering mechanism that allowed people for a time to get used to these new modes of transportation at sub market prices. Because of the subsidization, he wrote that those companies have gone through a painful period of promising an eventual business model based on cost savings from self driving cars, but instead were only able to reach profitability through massive price increases. Tim points out that this was not always the case. He suggests that during the dot com bubble it didn't take the winning companies very long to become profitable. He points out that Google only raised 36 million in venture capital, that while Facebook raised billions, it did so only to fund faster growth of a business model that he said was very close to profitable throughout the company's entire life. And what's more, that they weren't buying users with subsidized prices, but they were buying data centers. Now, Tim points out that taking less capital isn't some moral imperative, that there are times when the nature of what companies are doing just requires a ton of capital. He points to Tesla and SpaceX as great examples, that doing serious R and D, building factories, rockets, satellites, et cetera is really expensive, and that there is a very, very long path to creating a self sustaining business. Now, this phenomenon that he's talking about was not just limited to Uber. If you were living in San Francisco in the first half of the teens, you were eating good. Every new service, and there were tons of them, was artificially cheap, at least for a time, and so your food delivery was cheaper than it would ever be again, as was your ride sharing, as was your. You name it. Of course, as we know, that party eventually came crashing. And a lot of the story of the last couple years are the markets adapting to new realities. And this brings Tim to artificial intelligence. In a section he called Expecting Big Returns, he writes, in the case of artificial intelligence, training large models is indeed expensive, requiring large capital investments. But those investments demand commensurately large returns. The investors who pile billions of dollars into a huge bet are expecting not just to be paid back, but to be paid back a hundredfold. Tim argues that this capital fueled race has already led to bad behavior. By way of example, he points to accusations that OpenAI has trained on copyrighted content from pirate sites and that in general, the way that they absorb all content will, quote, eliminate opportunities for the owners of specialized content repositories to profit from their own work. He also argues that OpenAI is trying to follow a sort of apple path by building a platform or market where other entrepreneurs can build vertical applications, but only if they give that market owner a cut. Tim also argues that it's not just venture capital that is a axis of premature market concentration. He also points to the deals between Microsoft, Amazon and Google with OpenAI, anthropic and a handful of other labs. Tim writes, the risk of these deals is again that a few centrally chosen winners will quickly emerge, meaning there's a shorter and less robust period of experimentation. Tim concludes, at least based on recent reporting by the information about Anthropic's operating margins, it may be that like Uber and Lyft, the overfunded AI market leaders may only be able to deliver on investors heated expectations by crushing all competition. That's not betting on the wisdom of the market and what Hayek called the utilization of knowledge, which is not given to anyone in its totality. That's betting on premature consolidation and the wisdom of a few large investors to choose a future everyone else will be forced to live in. Really thought provoking stuff from Tim. I'm glad that he's writing and bringing up these issues. I think that while it's tempting to draw analogies to Uber and the ZIRP era period, we are living through something that is a little bit and pretty fundamentally different. These megadeals between Microsoft and OpenAI, for example, are a totally different type of force and are something that we have not seen before. These are deals where there is literally not enough venture capital for these labs to compete at the cost of what competing looks like. The big tech companies are the literal only source of capital at the scale needed to acquire the scale of compute needed. And so we've seen this weird thing where startups pair off with big tech giants in uneasy sort of frenemy relationships that are sometimes exclusive and sometimes not and rub people as weird because they just represent something different, something new. Certainly there are big questions about these things. One of the reasons that Elon seems so upset at OpenAI is that he accuses them of basically being a profit center for Microsoft. Now the really interesting question around competition is one that we've explored here before, which is not just whether Anthropic or OpenAI will ultimately win the foundation model game, but whether the world will come to be dominated by mega large multimodal models that can do everything, or whether instead we'll see more specialized models, more verticalized models, be able to continue to compete and hold their own. I think it's an open question, and I also think that the forces pushing for innovation and technological advancement with lower levels of computing access are a powerful force in the markets right now as well, because there is simply not enough capital and compute to go around in some ways by having a very small number of companies hoovering up all of the capital, it's forcing a whole different part of the market to innovate on a very different dimension. Now, I don't want to be Pollyannish at all about the risks that Tim is talking about. I think they are there and I think in some ways, as I've talked about on the show before as well, the bigger question is not just about whether we see winners take all in the LLM market, but whether LLMs become so significant in the structure and shape of society that having single companies own them becomes problematic in its own right. Given how challenging just big tech companies are and how much more powerful LLMs and generative AI could be, I think those are questions worthy of asking. Then again, I think most questions are worthy of asking, especially when we are in such a transitional moment as we are right now. And I hope that this podcast has given you a little bit more food for thought. For now, however, that is where we will wrap this AI breakdown. Appreciate you listening as always and until next time. Peace. Sam.