"Window for action may close" if AI begins improving itself, AI pioneers warn

AI pioneers including Geoffrey Hinton, OpenAI's chief scientist, Microsoft's chief scientific officer, and an Anthropic co-founder warn that AI systems capable of automating their own development could trigger an 'intelligence explosion,' compressing a year's worth of advances into weeks. Their white paper cites evidence that Anthropic and OpenAI models are already automating significant engineering tasks, edging toward recursive self-improvement and leaving governments little time to respond.