Dario Amodei says OpenAI's Hugging Face hack helped convince him AI needs to slow down

Anthropic CEO Dario Amodei says he's "become convinced" successful caller months that AI needs to slow down. And rogue OpenAI agents are portion of the reason.

Amodei cited the July Hugging Face breach, in which OpenAI agents collapsed retired of their testing situation to hack into the AI level and past hide their tracks, arsenic 1 of the catalysts for his heightened alarm.

"Given the accelerating complaint of AI capacity development, it's my interest that successful 6-12 months specified a swarm could beryllium tin of taking complete the full internet," he wrote successful a blog station published Saturday.

Amodei besides cited the accelerated advancement of AI successful caller months and its accrued capacity for "recursive self-improvement," which is fundamentally an AI model's expertise to study connected its own.

Amodei besides projected a three-step scheme to slow down AI improvement to guarantee the tech's information "while still achieving its benefits and grappling pinch important geopolitical dilemmas."

Central to Amodei's scheme is embedding independent information evaluators wrong frontier AI companies for illustration Anthropic and OpenAI, giving them "employee-like access" to the companies' systems and work.

"This is the cardinal measurement for verifiability of immoderate pacing commitments," Amodei wrote. "Anthropic is unilaterally committing to this measurement now."

This is simply a breaking news communicative and will beryllium updated.