Why are AI execs suddenly saying that superintelligence is just around the corner?
This is an edition of The Atlantic Daily, a newsletter that guides you through the biggest stories of the day, helps you discover new ideas, and recommends the best in culture. Sign up for it here.
Even for Silicon Valley, where executives have spent the past two years likening their chatbots to fire, electricity, and nuclear weapons, the past few months have been extraordinary. Some of the most important and well-respected figures in AI have recently said they are very close to building superintelligent software. So close, in fact, that they are giving themselves deadlines.
Bots that are far smarter than humans will arrive by 2030, the recent Nobel laureate Demis Hassabis, who runs Google DeepMind, said in August. Or in a “few thousand days,” according to OpenAI CEO Sam Altman in September. Or, if you listen to Dario Amodei—the chief executive of Anthropic and perhaps Altman’s biggest rival—by 2026. “Many will be literally moved to tears,” he wrote in a self-published essay this month. It will be a future, the AI leaders claim, free from disease, climate change, and poverty.
But these prophecies might not come from a place of strength, as I wrote yesterday. The energy, water, and capital demands of generative AI are astonishing, requiring perhaps trillions of dollars of spending this decade alone. Revenue hasn’t kept up, so one way for these companies to maintain the flow of investment dollars is to double down on the hype. “That omniscient computer programs will soon end all disease is worth any amount of spending today,” I wrote. This might be AI’s more important “rhetorical scaling law: bold prediction leading to lavish investment that requires a still-more-outlandish prediction, and so on.”
The AI Boom Has an Expiration Date
By Matteo Wong
Altman, Amodei, Hassabis, and other tech executives are fond of lauding the so-called AI scaling laws, referencing the belief that feeding AI programs more data, more computer chips, and more electricity will make them better. What that really entails, of course, is pumping their chatbots with more money—which means that enormous expenditures, absurd projected energy demands, and high losses might really be a badge of honor. In this tautology, the act of spending is proof that the spending is justified.
What to Read Next
P.S.
Did parallel parking trip up your first, or second, or third driver’s test? Future generations may not have to suffer in the same way. Although self-driving cars remain far off, self-parking cars are already on their way to becoming the norm, my colleague Kaitlyn Tiffany writes.
— Matteo