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

P-Zero Research published a new benchmark designed to assess how well AI models can independently design experiments and evaluate their own results. The benchmark tracks progress on a capability that sits at the intersection of autonomous research and self-improvement, two areas long considered frontier challenges in AI development.

Anthropic co-founder Jack Clark responded to the benchmark results by moving his prediction for self-improving AI forward to 2027. Clark has been a prominent voice on AI timelines and safety, and his role at Anthropic places him at the center of one of the industry's most closely watched labs.

Self-improving AI refers to systems capable of iteratively enhancing their own capabilities without direct human intervention. Achieving this milestone would mark a significant shift in how AI systems develop, potentially accelerating progress across domains but also raising substantial safety and governance questions.

The benchmark's rapid improvement trajectory suggests that autonomous experimentation capabilities are advancing faster than many experts anticipated. Whether 2027 is realistic remains a matter of debate, but the shift in a leading AI figure's timeline underscores growing confidence that self-improvement is no longer purely theoretical.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.