Who’s Winning the AI War?
Who’s Winning the AI War?
All of us, except the AI startups and VCs—unless a real war breaks out
“Chip War” by Chris Miller was an excellent book that seized many policymakers and technologists’ attention since it came out late 2022. If you haven’t read it, I highly suggest it.
It’s a history of the semiconductor industry but also has a central thesis that the new commodity that dictates control of the world economy—and world—is semiconductors. Like the old saying, “X is the new oil,” in this case, chips are the new oil.
So, is there an “AI War”?
Yes. If you accept the central thesis that semiconductors are a kind of Cold War right now—and it really seems like the US and China think so—AI is one of the central reasons for it. While semiconductors are in almost everything now, your car, smart water bottle, and even guided missiles, generally do not require leading-edge chips, which everyone is rushing to get an edge in.
AI is the “killer app” for lower power and faster computational capabilities, which is what each generation of semiconductor advancement provides. And, as we hit limits in what we can do because we’re literally down to atomic size being a problem, semiconductor fabs, design, and everything else gets exponentially more expensive.
Finally, as that all happens, AI leadership itself becomes a more and more important axis, not just semiconductors. So yes, I’d say there’s a war, and I don’t even think it’s that controversial that it’s being fought, technologically and literally, between countries.
However, there’s also a war in the private sector and given how AI is likely to change everything we do going forward—similar to how semiconductors did in transforming literally every aspect of modern life—everyone is caught up in it and its implications.
Who’s in the game?
Let’s go over some of the players here.
Big Tech Companies
I’ve talked about this ad nauseam now. The best-poised players to take over more general AI applications are the ones who mastered compute costs and scale in the internet era. That’s all the names you expect: Google, Microsoft, Amazon, and the like—which happen to also be the companies that have significant control and arguably own many of the best-known models and companies that make them. US tech in particular dominates this space.
AI Startups
I happen to think that there are going to be some incredible AI startups in the coming years. I also think most current AI startups are going to die, especially those pursuing “foundational models” which is usually shorthand for “we have no defensibility.”
So, what should AI startups do? Well, one interesting fact is after a major war, say, WWI or WWII, the winners tend to do much better economically (in GDP and their stock markets) than the losers who are devastated. Yes, duh. Do you know who does even better (relative to their size/scale/development)?
The countries that were never at war.
In a similar vein, AI startups who avoid the freight train of competing in the same lanes as the big tech companies and everyone else will have a far better chance.
Venture Capitalists
A bunch of VCs are pouring money into AI. At last count, nearly a third of VC money in 2023 (which had fallen significantly since 2022) went to AI.
Nonetheless, this is an industry that has gotten used to and trained up partners who mostly operate in the SaaS (software-as-a-service) environment, which looked very little like semiconductors… which means it also doesn’t look much like AI.
A SaaS company can cheaply get a product to market, iterate, pivot, and see what the market says, with steadily growing ARR (annual recurring revenue) and concrete churn/customer acquisition cost/etc numbers. You don’t really need to make a long-term bet, since it’s constantly way-finding and has a lot of wayfinders (serial entrepreneurs) who’ve done it a lot of times.
An AI company might need thousands or millions of dollars to collect data, and then thousands or millions of dollars to train the model on it… before serving a single customer for real. That looks a lot more like a (fabless) semiconductor company than it does a software company.
Anyway, VC is an industry that itself is in transition and is throwing money around, in my opinion, somewhat indiscriminately in the sector—similar to what it did with blockchain/crypto. At least in this case, it’s towards something that will truly generate a lot of value (which is slightly unfair to blockchain, but it was so tainted by cryptocurrency that it never attracted enough of the right people or took off on its own).
State Actors and States
I mentioned briefly governments above, but they are (and probably rightly) thinking about AI as a huge frontier in war. As such, we’ve seen some—in my opinion—craziness-due-to-the-current-political-environment proposals about AI regulation by people who don’t know anything about it and think command-and-control of an industry will allow it to thrive and be controlled.
You can ask China how well that experiment in going, in terms of punching far below their weight right now for where they should be in AI.
Or Europe, which, despite ever more strident declarations of technological leadership over the past five years, has mostly become a technological (especially in tech industries and advancements around AI) basket case that it’s become such a popular topic to write about that even McKinsey was willing to write a public report on how bad it is.
So far, the US has largely been restrained, with actions that don’t mean too much (like I covered on the AI Executive Order).
States and state actors have a huge challenge in AI models being generally open-source and evolving too rapidly for non-specialists to pin down without killing the golden goose of their AI leadership—which might also kill the country, literally.
I’ve personally experienced seeing, talking to, and having state actors actively poking around in the AI industry. However, we should expect them to be “customers” of AI for technological, informational, and kinetic warfare, more so than active participants.
Society
Society, people… we are also all caught up in this. To explain this requires a somewhat long detour, and a long explanation, because “regular people” who are bystanders deserve the space and sympathy.
I’m a big fan of the late Iain Banks. He was a science fiction author who not only was a good writer—with great characters and extremely imaginative settings that were both familiar but completely foreign—but also wrote about the “future” (it’s not clear if it’s the future or, as with Star Wars, a galaxy far, far away) in his Culture series that is post-material with AI that are fundamentally partners and friends of humanity. The collective AI/humanity society was called the Culture, and fairly utopian.
The Terminator Series, Mission Impossible Dead Reckoning, and innumerable Hollywood movies and books are all about AI going wrong. They’re fundamentally tech phobia and fear of change. While I can understand it, the theme not only becomes somewhat stale after the umpteenth tired rehashing of essentially the same story. (As also a fan of Frank Herbert and Dune, I also feel sorry for him and his story that he didn’t get to take the story there while he was alive as he planned for the “big bad” in the series while it was still fresher).
Nonetheless, aside from a suggestion to read Iain Banks’ books, I found it refreshing to have a positive and hopeful viewpoint on what technology and AI would bring to us. Giving away my views, I think AI will ultimately be closer to that than Terminator. Despite the pronouncements of the AI Doomers, you have to anthropomorphize AI as us to follow the chain of logic that usually results in doom (aside from somewhat absurdist thought experiments like the Paperclip Problem). It’s not really a flattering picture of what we think about us, but AI also isn’t us.
All of this is to say: AI is a technology. Like a lot of fundamental technological revolutions, like the Industrial Revolution in late 1800/early 1900 that fundamentally changed life and allowed for mass-scale leisure and specialization (that wasn’t just reserved to royalty, priestly elite, etc), or like the Green Revolution in the 1950s that allowed us to feed the world, or the Semiconductor Revolution that formed the basis for the only world that anyone born after 1970 knew… AI’s revolution is going to bring a lot of good.
It doesn’t mean that the technology can’t be misused. The Industrial Revolution also brought mechanized warfare that we saw in WWI and WWII. The Green Revolution gave us industrial agriculture, monocultures, and its attendant environmental impacts. And, of course, the semiconductor revolution has also created a world of non-stop information, constant connectedness, and smart everything. It’s easy to focus on the bad and let that drown out the good, but it doesn’t mean that tech can’t be misused. It’s just a tool. Like a shovel, it can be used to dig a hole to plant something, or knock someone on the head and dig their grave.
Anyway, all of this being said, society will overall benefit. But there’s definitely going to be a lot of unintended consequences that we’ll have to strap in, learn from, and hopefully fix.
So, who wins?
Believe it or not, I think society.
Most of the benefits of this technology are going to be hard to capture for any one company (even the massive tech companies), and most of it will be what is known in economics as “surplus” that accrues to everyone. Basically, in a market economy, when profits are competed away, they don’t evaporate. They become high-quality whatever-it-is for as low a price as can be provided and still produce it. This becomes societal “profit,” which is why competition is good.
This isn’t going to be perfect competition, but most AI startups are not going to capture the profits, and the big tech companies are going to capture some, but not the lion’s share.
As such, society most wins, big tech companies win next most, and some AI startups make it out, and a few VCs get kind of lucky.
Of course, I haven’t talked to about state actors yet.
In this case, the US is leading the pack right now, just in terms of most of the “winning” players inside of its borders. However, the one wrinkle in this picture, similar to the Industrial Revolution creating machine guns, is what ends up happening with AI in the state context, especially with it being seen as so key to the next World War—or, at the very least, conflict between the US and China.
The US isn’t guaranteed to hold that position, especially if it missteps significantly in regulation and if China fixes its problems. Additionally, society may not be such a winner anymore if AI does get taken to be more Terminator than the Culture.
But, if that happens, it’s not going to be because AIs went rogue and we have a modern Frankenstein story. It’s going to be because human beings take a tool, and turn it into a weapon—and take our figurative war, and turn it into a real one.