Zuckerberg Rejects an AI Pause as Rivals Push for a Slowdown

Mark Zuckerberg says AI labs should manage safety themselves, rejecting a coordinated slowdown backed by rivals. The clash pits Meta’s case for company-by-company safeguards against fears that the race is moving too fast.
Zuckerberg Rejects an AI Pause as Rivals Push for a Slowdown

Zuckerberg Rejects an AI Pause as Rivals Push for a Slowdown
The argument sharpened after an Anthropic researcher resigned last week, warning that AI companies were “gambling with our lives.” That alarm helped propel Anthropic chief executive Dario Amodei’s call on Saturday for a global slowdown, a position later joined by OpenAI’s Sam Altman, Elon Musk and Google DeepMind’s Demis Hassabis.

On Tuesday, Meta chief executive Mark Zuckerberg broke from that emerging coalition. He accepted that AI safety demands caution, but rejected an industry-wide pause as the remedy. “Every lab has the responsibility and incentive to move at the pace required to train its models safely,” he said, arguing that Meta had made that commitment and rivals could do the same.

His case rests on voluntary, company-by-company restraint rather than a shared speed limit. Zuckerberg pointed to Meta’s Muse AI agent, saying the company delayed its release for several months to address safety and security concerns. “We didn’t call for everyone else to do this before we would,” he said. “We just did it as part of our day-to-day work because it was clearly the right thing for people and for us.”

That framing challenges Amodei’s central concern: that no single lab can safely police a race in which competitors fear falling behind. Zuckerberg’s response is that safety work should be embedded in normal product decisions, not used to halt everyone else’s development. He also criticized labs focused on “recursive self-improvement,” urging them to concentrate on serving people instead.

Behind the safety dispute lies a harder geopolitical and commercial argument. Zuckerberg has warned that “any policy that slows American model releases – even by a month – could add significant risk to American leadership while letting foreign models race ahead.” Meta, aiming to close ground on Anthropic and OpenAI, is preparing to release its most powerful model yet, internally called Watermelon, in the coming weeks.

https://foxvector.com

Write a comment