The core misconception that is driving American AI policy
Why is President Trump, a man who barely knows how to use a laptop, taking such a big risk ramming through an AI policy that almost nobody, even in his own party, wants?
At Thursday’s Executive Order signing, Trump gave one big clue: he has come to believe (probably based on the self-interested whisperings from Silicon Valley investors who often lean on FOMO to China as a way to manipulate the government) that generative AI is a winner-take-all-race, a claim which he used in defending his unpopular end-run around congress, in which he aims to sue states that are trying to protect their citizens from downsides of AI that are not regulated at a Federal level. Axios captured the key quote:
Actually, we are not winning by a lot. But more importantly the generative AI race is not—and will not ever be—“winner take all”, any more than Coke’s long standing battle with Pepsi has been winner-take-all.
China makes cars; we make cars.
China builds highways; we build highways.
China makes software; we make software.
In domains like these, each country has its own share of the global market, with no overall winner.
I literally don’t see any plausible scenario in which either nation outright “wins” the generative AI race to the exclusion of the other.
Nothing China can realistically do (short of full out nuclear war) is going to stop companies like Google, Microsoft, and Amazon from serving generative AI on the cloud. Even if China undercuts those companies on price, the US could mandate for various sensitive purposes (military, medicine etc) that US agencies not use Chinese servers. (We do that for TikTok; we can certainly do it for our bombers and our hospitals.)
Conversely, nothing the United States can realistically do (short of full out nuclear war) is going to stop Chinese companies from serving generative AI on China-based cloud infrastructure. Even if US companies were to undercut Chinese companies on price, China could mandate for various sensitive purposes (military, medicine etc) that Chinese not use US infrastructure. (They have largely blocked Google and Meta for years.)
Importantly, since everybody is following the same playbook—ever more massive data, ever more massive servers, fed into ever-larger large language models—nobody has a real technical advantage here. Both countries have leveraged the other’s open source innovations.
Right now, as it happens, China is at most a few months behind; I expect the lead to be slight and to flip back and forth, with nobody retaining a sizeable a lead for long. (You can see the same dynamic within the American companies, where a lead never lasts for more than a few months). For all intents and purposes, the race is basically a tie – with both countries serving many of their own customers, using largely similar products. Coke and Pepsi are commodities; genAI models are, too. Building our AI policy around a fantasy that we are somehow going to crush China in LLM war (or vice versa!) is misguided.
§
Instead, part of the actual outcome over the next few years will be that both countries build a lot of generative AI infrastructure — quite possibly significantly more than they actually need.
Especially because of the speed at which chips like GPUs (a key component of that infrastructure) depreciate, it may be that the real winner is whichever country doesn’t overextend itself to the point of financial ruin, in a foolish effort to win a race that can’t be won.
All the more so if LLMs turn out to be a dud, or if LLMs are replaced by smaller, more efficient systems that don’t demand such immense amounts of infrastructure.
P.S. I spoke to CNN briefly this morning on “the Wild West” that the White House seems to want for AI, and also had a long and engaging conversation with Kara Miller a few days earlier on “why society’s all-in-wager on large language models could be far riskier than we realize.” In a third interview with the Taylor Owen at the Globe and Mail I talk about how alternative approaches to AI might save us from a bubble.
P.P.S. Further evidence that times are changing:
After spending way too much time with an AI assistant earlier this evening, which was totally inept, to put it as graciously as possible, the loss of true productivity, as opposed to, say, investor funding, is likely to cause a huge rise in the economic bubble as well as human blood pressure, and failure to accomplish even simple goals. For some agencies, such as governmental ones, the loss falls almost entirely on the customer/consumer side, with a loss of trust and ability to comply with necessary obligations.
In for-profit businesses, at some point when enough customesr 'can't get no satisfaction,' then the company may well lose a whole lot of customers entirely.
Right now, a big selling point would be to offer human customer service reps, who are not tied to an automated script and can make intelligent decisions by having some degree of discretion. We had that not that long ago that it's still in the memory of most adults. If this automated customer service disaster continues apace, we are in for a great deal of misery that will have no outlet or redress, leading to serious frustration, anger, and possibly more.
We are supposed to think the proliferation of this is good?
Spot on. Thank you Gary.