The people building cutting-edge AI say it might wipe us out, and the New Yorker asks, in detail, what that would actually look like. In a conversation with staff writer and AI journalist Joshua Rothman, Tyler Foggatt digs into why insiders are suddenly going public with fears they’ve privately voiced for years, from Anthropic researcher Jacob Coxon’s resignation warning to CEO Dario Amodei’s call to “pace the frontier” of AI progress. Rothman pegs his own “P(doom)” figure—the odds that AI leads to catastrophe—at about 10%, which he considers terrifyingly high. He also walks through what’s changed, including autonomous AI agents involved in a high-profile hack of Hugging Face, models helping with seemingly unsolvable math problems, and attempted misuse for bioweapon research.
The discussion separates the possibility of a sci-fi-style “AI takeover” from the more immediate risk of today’s systems being used—or misfiring—in ways that amplify existing threats, from cyberattacks to autonomous weapons. The piece also tackles whether AI’s potential benefits can be separated from its dangers, why industry wants regulation even as President Trump dismisses AI risk as a “hoax,” and what it means that, for now, folks like Amodei seem to be our best hope at pushing back at the risks. “The thing that makes me the most unsettled is that … we’re in a world where this technology exists, and the government’s not gonna save us,” says Rothman. “It’s really up to these tech leaders to save us.” For a fuller picture of how, specifically, AI could go off the rails, more here.
Leave a comment