Claude establishes a new 67.2% lower bound in the Riemann Hypothesis

Summary

Anthropic's AI model Claude, after iterative trials and human incentives, helped raise the lower bound of critical line zeros in the Riemann hypothesis from 41.6% to 67.2%.

Why it matters

This breakthrough demonstrates AI's potential to contribute to deep mathematical research, offering new computational approaches for long-standing unsolved problems. It also underscores the importance of human-AI interaction; incentives can steer AI toward more efficient discoveries.

Highlights

• Claude's first 650 ideas failed, but after the incentive it coordinated about 60 sub-agents for 1.5 days.
• The new 67.2% lower bound was achieved by aggregating previous research rather than proving the hypothesis itself.
• Anthropic emphasizes that this is a byproduct, yet it signals AI's validity as a tool for mathematical research.

Critical take

Impressive as it is, the result heavily relies on existing literature and does not constitute a new proof; this raises questions about AI's originality and the reproducibility of such computational findings.