← Back to Model Beat
Research·21h ago·all news from October 25, 2026

Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027

Anthropic co-founder Jack Clark has moved his forecast for the arrival of self-improving AI to 2027. This shift follows new research indicating that models are becoming increasingly proficient at autonomously designing and evaluating complex experiments. The updated timeline reflects a growing technical consensus that AI development may soon reach a threshold where systems can significantly accelerate their own advancement.

Covered by 1 source

Related stories

ResearchOpenAI Resumes Training AI, But Without Internet AccessOct 23 · 11 sourcesResearchSecuring our frontier model training environmentOct 23 · 2 sourcesResearchUber brings ChatGPT Work to all employees with OpenAIOct 22 · 2 sourcesResearchTurning Database Grading into RL Rewards: Post-Training Agents with ThinkingBoxOct 22