AI Safety: Intelligence Explosion Timeline & METR's External Oversight Vision
AI safety researchers are flagging that limited superintelligence may already be here and an intelligence explosion could happen within a year — while METR releases a framework for external parties to detect when AI developers are losing control of their agents.
What we know
An account on X RT'd the view that 'limited superintelligence is already here' and that Jack Clark believes an intelligence explosion could occur within a year — possibly controlled, but uncertain. Simultaneously, another account on X RT'd METR's new public framework: how would anyone outside an AI company know if it was losing control of its internal agents? METR is proposing collaborative assessment processes to surface this evidence openly. Noah Smith added that no AI video app has matched Sora's UX for social creativity — another account on X — gesturing at the uneven pace of AI progress across domains.
By the Global Eye 24 desk, with AI assistance. Updated as the story develops. Spot an error? Write to corrections@globaleye24.com.