Anthropic’s latest experiment with Claude has offered a glimpse into how artificial intelligence could contribute to difficult mathematical research without actually solving one of mathematics’ most famous unsolved problems. An unreleased research version of Claude was asked to make a serious attempt at the Riemann hypothesis, a problem dating to 1859 that carries a $1 million prize for its solution. Claude did not prove the hypothesis.Instead, during its attempt, it produced a new result concerning the proportion of zeros of the Riemann zeta function that are known to lie on the critical line. Anthropic says Claude improved the longstanding lower bound from 41.6% to 67.2%, with the work subsequently examined by mathematicians and formally verified using Lean.About The AuthorHey there, i am a technology enthusiast with a deep passion for gadgets, consumer electronics, emerging technologies, and the fast-paced world of digital innovation. Constantly exploring the latest tech trends, product launches, and industry developments, I enjoy translating complex technological advancements into engaging and accessible stories for readers. My interests span smartphones, wearables, artificial intelligence, smart devices, and the broader technology ecosystem. As I begin my journey as a Tech Journalist at Gadgets Now, I am excited to contribute to a platform that informs millions of readers, combining my passion for technology with storytelling to deliver insightful, accurate, and timely tech coverage.The Riemann zeta function: Why does it matter?The Riemann zeta function is important because of its deep connection to the distribution of prime numbers, which are fundamental to number theory and modern cryptography. According to Anthropic, the function’s zeros contain increasingly fine information about how primes are distributed.The Riemann hypothesis states these solutions lie on a vertical line in the complex numbers. Tremendous math has been put forth but no one has proven the Riemann hypothesis wrong and no one has managed to prove it.Researchers have instead established results about the zeros and gradually increased the proportion known to lie on the critical line. Anthropic says that proportion had reached 41.6% before Claude’s work pushed the lower bound to 67.2%.How Claude approached this problemClaude’s approach was less like a single calculation and more like an extended research process involving multiple AI agents. Anthropic staff member Jarred Sumner initially asked Claude to “take a real stab” at the Riemann hypothesis, allowing the model to determine the mathematical direction. The first effort generated about 650 ideas without success.After another prompt, Claude spent roughly a day and a half coordinating around 60 subagents. Together, they executed about 2,400 shell commands, produced hundreds of Python scripts and conducted thousands of numerical checks against known zeta zeros. Anthropic says the agents also reviewed one another’s arguments, searched for counterexamples and examined previous research before arriving at the new result.What are its findings? Claude’s important finding was not a solution to the Riemann hypothesis, but an improvement to a related lower bound. Anthropic says Claude combined results from mathematicians Baluyot, Goldston, Suriajaya and Turnage-Butterbaugh with earlier work by Enrico Bombieri to establish that at least 67.2% of the relevant zeros lie on the critical line, compared with the previous 41.6% bound.More articles by AuthorTrending StoriesThe key mathematical step involved constructing a suitable space of functions and using a quadratic-form argument that simultaneously accounted for positive- and negative-definite subspaces. Anthropic mathematicians Levent Alpöge and Ralph Furman examined the work, while Claude also produced a Lean formalization that passed a standard validation tool.Why Claude’s failure may be more important than a solutionThe significance of the experiment lies partly in what Claude did not accomplish. Anthropic explicitly says it does not expect the techniques used in this result to lead directly to a proof of the Riemann hypothesis. The value instead comes from demonstrating that a language model can navigate a substantial body of existing mathematical literature, combine techniques from separate lines of research and identify a potentially useful connection that was not the original objective.That distinction matters because mathematical research is rarely limited to finding an answer to a clearly defined problem. Researchers often build on partial results, reformulate questions and discover unexpected consequences while pursuing a different goal.Claude’s experiment suggests that AI systems may increasingly participate in this exploratory stage of mathematics, where identifying promising combinations of existing ideas can be as important as performing calculations.Anthropic said Claude’s result emerged as an unintended byproduct of the original attempt to attack the Riemann hypothesis. The model itself was initially skeptical about making meaningful progress, apparently reflecting its learned understanding of how difficult the problem is. Continued prompting encouraged it to persist. The resulting workflow also included independent reviews, searches for existing work and attempts to reproduce the result from scratch, providing safeguards against a plausible-looking but incorrect mathematical claim.What the experiment says about AI’s mathematical capabilitiesThe episode points toward a changing role for AI in advanced mathematics. Claude was not simply asked to calculate an answer; it generated hypotheses, delegated tasks, wrote code, tested numerical evidence, examined research papers and subjected its own conclusions to scrutiny. Anthropic says 54 papers were downloaded during the checking process to determine whether the finding had already been made, while human mathematicians subsequently examined the result.Still, the episode should not be described as AI solving the Riemann hypothesis. That claim would go beyond the evidence. The hypothesis remains open, and Anthropic itself does not argue that Claude’s technique will resolve it. Instead, the experiment demonstrates a narrower but potentially significant capability: AI can sometimes combine established mathematical ideas in ways that produce new results.For mathematicians, that could make AI useful not as an autonomous replacement for researchers but as a high-speed collaborator capable of exploring large numbers of possible approaches. Claude’s Riemann zeta experiment therefore offers a more measured picture of AI progress in mathematics: not a machine conquering one of the discipline’s greatest problems, but a system showing that the boundary between mathematical assistance and mathematical discovery may be moving. FAQsWhat is the Riemann hypothesis and why is it significant?The Riemann hypothesis is an unsolved mathematical problem dating back to 1859 that proposes all zeros of the Riemann zeta function lie on a specific vertical line in the complex plane. Its significance lies in its connection to the distribution of prime numbers, which are fundamental to number theory and modern cryptography.How did Anthropic's Claude contribute to mathematical research regarding the Riemann hypothesis?Claude improved the lower bound of the proportion of zeros of the Riemann zeta function known to lie on the critical line from 41.6% to 67.2%. Although it did not solve the Riemann hypothesis, Claude's approach involved collaborative efforts by AI agents, generating numerous hypotheses and combining insights from prior research.What does Claude's experiment suggest about the role of AI in mathematics?Claude's experiment suggests that AI can play a significant role in the exploratory phase of mathematical research by navigating existing literature, combining techniques from various fields, and identifying new connections, rather than merely calculating answers. It highlights AI's potential as a collaborative tool for mathematicians.end of article