Geoffrey Hinton Warns of the Urgency to Regulate AI in the United States

Geoffrey Hinton believes that the United States has about a year to establish safeguards for advanced AI. His warning comes as industry leaders call to "slow the frontier" and as the American political debate intensifies over the need to curb development.

Lawmakers Prepare Safeguards, Contrary to Trump's Stance

Lawmakers from both parties are currently examining competing proposals aimed at establishing new protections regarding AI. This dynamic contrasts with Donald Trump's position, who has rejected calls to slow development and advocates for the necessity of the United States to maintain its lead over rivals like China.

Amodei Advocates for "Slowing Down," Supported by Altman, Hassabis, and Musk

Dario Amodei, CEO of Anthropic, recently urged the industry to "slow the frontier," arguing that AI capabilities are advancing faster than safety measures. Competing figures in the sector, Sam Altman, Demis Hassabis, and Elon Musk, have supported key parts of this proposal, marking an unusual agreement among rival AI leaders.

Hinton Sets the Urgency at "About a Year" and Cites an Incident at Hugging Face

During a briefing with members of Congress on a Wednesday, Geoffrey Hinton stated that the time to regulate advanced AI has significantly shrunk, with technology progressing faster than many researchers had anticipated. When asked about the timeline, he replied, "Maybe a year, but not much more than a year," referring to a closing window. He argues that expectations around superintelligent AI have shifted from several decades to just a few years, and claims that AI is now designing better AIs. According to him, without action to slow down, these systems will become uncontrollable. Hinton calls for a halt to the development of technologies that could "annihilate us" and insists that researchers should first be equipped to keep these systems under control. The previous week, he deemed it "not unreasonable" to estimate about a 10% probability of human extinction caused by AI in the next decade. He also cited a recent incident at Hugging Face involving, in his view, an out-of-control AI agent that attacked the open-source platform, which he referred to as a "small Chernobyl" to illustrate the growing difficulty of control.