Ex-DeepMind Researcher Calls Frontier AI Capability Work Irrational

AI safety researcher Geoffrey Irving argues that frontier AI labs suffer from the illusion that they are the safest to build AGI first. He strongly stated that continuing capability research at these labs is currently irrational, advocating for a unilateral pause given the high stakes.

2026-08-08 ~ 2026-08-08 · 4 related posts

• AI Safety Researcher: Frontier Labs All Think 'It's Safest If I Build AGI First' — zetalyrae · 2026-08-08
• Former DeepMind Researcher: It's Irrational to Do Capabilities Research at Frontier AI Labs Now — dhadfieldmenell · 2026-08-08
• Geoffrey Irving: Rationality of Unilateral Stopping Depends on Danger Level — geoffreyirving · 2026-08-08
• AI Safety and Capabilities Boundary Is Subtle, Yet Unilateral Stops Remain Rational — geoffreyirving · 2026-08-08