Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

AI claimed a solution to the 167-year-old Riemann Hypothesis, and a human did too

An AI system has claimed a solution to the 167-year-old Riemann Hypothesis while a human mathematician advanced a competing solution, TechCrunch reports, leaving mathematicians stunned. Neither argument's content, authorship, nor verification status was disclosed, so there is no basis yet for treating the conjecture as resolved.

Published
AI 宣称解决 167 年未解的黎曼猜想,人类数学家同日也给出方案
Image source: techcrunch.com

An AI system has claimed a solution to the Riemann Hypothesis, a problem open for 167 years, and a human mathematician has put forward a solution of their own, leaving mathematicians stunned, TechCrunch reports.

Bernhard Riemann posed the conjecture in 1859. It concerns the location of the nontrivial zeros of the Riemann zeta function, and its truth governs the error term in the distribution of prime numbers; it has long ranked among mathematics' most important open problems.

The word the report uses is claimed, and that matters. In mathematics a claim and an accepted proof are separated by a long verification process in which peers check each step, and a single unproved lemma can bring the whole argument down.

The coincidence is the more interesting part: an AI result and a human result surfacing in the same news cycle. That shifts the discussion from whether a machine found the answer to how much verification each argument needs, and whether the two approaches converge.

Over the past two years AI's role in mathematics has expanded from assisting computation to proposing conjectures, searching for counterexamples, and drafting proofs. Mathematicians remain cautious for a practical reason: formal verification is expensive, and model-generated arguments can contain plausible-looking gaps that resist completion.

The report does not disclose the content of either argument, who produced them, or how far verification has gone, so there is no basis yet for saying the hypothesis has been resolved. The signal worth noting is that AI output is now entering discussions of a top-tier open problem rather than staying in competition problems and lemma proofs.

What to watch: whether both arguments are published for peer review, whether anyone attempts machine checking in a system such as Lean, and how quickly the community issues a first public assessment.

Why it matters

Whatever the outcome, the episode moves the AI-in-mathematics debate from contest problems to a landmark open problem, where the decisive variable is verification rather than generation.

MathematicsFrontier Research
Back to realtime news

Nearby Updates

All

11/21, 04:04

RYPTASK Launches Managed AI Agent Solution Aimed at MSPs

RYPTASK has launched a managed AI agent solution built for managed service providers and their clients, according to an announcement carried by EIN News. The notice does not disclose which models it runs, how it is priced, or when it becomes generally available, leaving the resellable agent product as the confirmed direction.

11/21, 02:15

METR's updated time horizon plot: exponential gains continue, judgment lags

METR has published an updated version of its widely cited and widely misread time horizon plot, adding longer and more open-ended tasks to probe how models handle work that cannot be scored automatically. The report finds capabilities on verifiable tasks still improving exponentially, while models remain below human level on judgment-heavy work such as designing evaluations or scoring philosophical arguments.

11/21, 04:49

Elvis Presley's estate sues Anthropic for $46.6 million

The estate of Elvis Presley has sued Anthropic for $46.6 million, according to Wisevoter, adding to the wave of copyright disputes over generative AI training data and outputs. The report does not describe the specific claims, the works involved, or Anthropic's response.

11/21, 01:53

Google's Gemini clears the static from firefighter radio, cutting Watch Duty alerts to under 30 seconds

Google detailed on November 20 how a Gemini-powered tool built with the nonprofit wildfire app Watch Duty transcribes noisy fire radio into clean, searchable text and flags new dispatches in under 30 seconds. Word error rate on degraded VHF audio fell from 38% to 22.34%, and volunteer reporter satisfaction with transcription rose from 35.3% to 92.3%.