
Illustration: Aïda Amer/Axios
The AI hype cycle is in full swing — which, like all hype cycles, means a large uptick in noise and nonsense.
Why it matters: The temptation is to try to separate the signal from the noise. (Good luck with that.) Trying to pick winners is even harder — although one Caribbean island is clearly among them.
The big picture: The AI hype cycle rests on firmer foundations than, say, the web3 hype cycle of a year or two ago — although telling the two apart isn't always easy.
- What's real: OpenAI's GPT-4 is an undeniably and impressively powerful machine, and serious scientists are genuinely worried about AI — although it's not clear that the worries are related to GPT-4 or even to large language models broadly. Politicians, in another echo of the crypto boom, are all talk and no action.
- What's hype: When a company starts talking loudly about its AI abilities, the first question should always be: "Why is this company talking loudly about its AI abilities?"
Between the lines: AI is good for drumming up buzz, but most of that buzz is either vague or implausible.
- Credible claims for AI tend to be obvious and intuitive. They generally involve computers doing what they've always done, which is taking over tasks that a human would find repetitive or boring, and doing them much more quickly. (Sometimes, when you look behind the curtain of AI claims, you find that it's actually outsourced humans doing those repetitive and boring things.)
- Non-credible claims for AI involve any attempt to paint it as a competitive moat. Even if a company's AI researchers have been working on a problem for years, AI technology is moving so quickly that tomorrow anybody else might be able to do the same thing in just a few weeks.
Where it stands: A common rhetorical move is to invoke national security. "The challenge will be to ensure that such technologies remain subservient to our collective will," writes Palantir CEO Alex Karp, for example. "But we will not stand still while our adversaries move ahead."
- Karp is not-so-subtly signaling that he intends to lease his AI to the U.S. government for defense-contractor sums of money. That's how the U.S. microchip industry got started, after all.
- There's clear utility in a CEO playing up the existential fears surrounding AI — doing so makes the technology feel more real, more magical, and therefore more valuable.
The bottom line: For the time being, one of the few certainties is that there's real money in selling picks and shovels to the folks getting into AI. In practice, that means computer chips from Nvidia and Qualcomm — as well as .ai domain names from the government of Anguilla.