Anthropic's Jack Clark predicts AI self-improvement by 2027

anthropic co-founder jack clark has moved his prediction for recursive self-improvement in ai to 2027. the prediction is based on the tasteval benchmark developed by p-zero research, which measures an ai system's ability to design experiments and interpret results. the benchmark has shown a steep improvement curve, with capabilities roughly doubling every three months since december 2025. the best model tested, anthropic's opus 5.5, outperformed human experts at a fraction of the cost. this rapid progress could change the direction of ai research and funding.