Opinion | Unintelligence on Super Intelligence

What’s behind the confusingly unspecific demands for government intervention from AI leaders? I reject the cynicism argument (it’s all about regulatory capture) in favor of cultural determinism—they’re true believers.

That said, Anthropic and OpenAI have valuations now to defend. The people who founded these companies may be deeply invested in the idea of artificial intelligence as the most important technology ever. But they’re also in a situation where the investing public, including sophisticated venture capitalists, interpret their creations as potential cash machines, worth maybe $2 trillion. However, under an equally plausible reading of the evidence, the companies are black holes of spending on rapidly commoditizing capabilities. Their specialty, the highest-end AI potential, the U.S. government won’t even let them sell to customers.

Now a documentary movie perhaps threatens to pop their bubble by proxy. The subject is Elon Musk, who director Alex Gibney says threw his support behind Donald Trump in 2024 because he feared under a different administration he’d land in jail.

My cards on the table: Tesla’s valuation never made sense to me, but the risks haven’t been hidden from investors. At the same time, if the markets turn on Mr. Musk, history shows the SEC and criminal prosecutors won’t be far behind.

This intersects with the AI debate in a way I think the world is getting wrong. David Robinson is the latest OpenAI veteran to show up on a New York Times podcast to lament his former employer’s lax safety efforts—keyword “latest.” The safety debate has obviously broken through in a big way. He’s talking to the New York Times, after all.

But AI remains unregulated, shriek those wedded to an undynamic view of reality. I don’t doubt we’ll get lots of AI legislation, but nothing in our society is unregulated. We have tort law, consumer safety statutes, etc.—and they’re working, as was obvious in April when Anthropic withheld Mythos from distribution except to trusted partners in the battle against cyber hacking.

But the obliviousness that really blows me away: Under a “voluntary” Trump process that isn’t really voluntary, the frontier labs are getting a de facto regulator. It’s the National Security Agency, repository of advanced U.S. cyber and spyware capabilities.

On OpenAI’s board even sits a former four-star NSA chief. This week Mr. Trump further put another intelligence official in charge of AI, Director of National Intelligence Jay Clayton, who will lead a task force on super intelligence. What’s more, everybody understands, because we been have told as much by both the Biden and Trump administrations, the draconian Defense Production Act of 1950 is ready be invoked as necessary in pursuit of U.S. national-security goals.

Are IPO investors paying attention? Anthropic this week allowed certain customers access to advanced models. It did so “in collaboration with the US government.”

At a San Francisco conference last weekend, an unnamed frontier lab executive was reported to have said there’s some level of machine intelligence beyond which the industry should be prepared not to go. We’ll see how the NSA feels about that.

In his new book, Kevin Roose fills out an anecdote previously relayed in this column. As far back as 2017, OpenAI leaders Sam Altman and Greg Brockman considered future government nationalization so likely, they debated how to get money in advance from China, Russia and the U.S. for the tools they expected these governments to expropriate.

The somatic metaphor has seized the punditry—the AI industry is tumbling forward uncontrollably at the hands of its private leaders. This hasn’t described reality for a while. Add it all up and the public panic of the past few weeks smells a great deal more like epiphenomena—a thing incidental to the real thing. Yes, AI will be important in the economy, but the super intelligence everyone keeps talking about? You will likely encounter it only as a rumor from the deepest bowels of the world’s top intelligence establishments.

The idea of runaway doom is one of those fun and frightening pictures that survive only in the moment between dreaming and waking. For the U.S. frontier labs, spending billions to be six months ahead of rivals, the thing to worry about is whether their spending will pay off—and whether government will let it.