Anthropic AI extends Riemann Hypothesis verified range after nonexpert prompt
Anthropic's new AI just made a real dent in the legendary Riemann Hypothesis, a puzzle about prime numbers that's stumped mathematicians for more than 150 years.
The million-dollar prize for a full proof is still out there, but this AI managed to extend the range where the hypothesis holds true, all after an Anthropic staff member without significant mathematical training prompted it to "take a real stab" at the hypothesis, then left it to coordinate the task across the following day and a half.
AI split tasks among 60 agents
The AI split its tasks among 60 mini-agents (some came up with insights, others helped out or checked work, and a few even wrote up the first draft of the research paper).
The finding was confirmed by two of Anthropic's in-house mathematicians and formalized using the open-source proof assistant Lean.
This win joins other recent moments where AI has cracked tough math problems, and it's got researchers talking about how tech could totally change how we do math in the future.