Back to List
AI Outpaces Human Mathematicians: ChatGPT Disproves Erdős Conjecture and Logical Intelligence Achieves Autoformalization in Lean
Research BreakthroughArtificial IntelligenceMathematicsLean

AI Outpaces Human Mathematicians: ChatGPT Disproves Erdős Conjecture and Logical Intelligence Achieves Autoformalization in Lean

In a significant shift for the field of mathematics, AI tools are now successfully identifying counterexamples to long-standing conjectures, a phenomenon described as humans being "outcounterexampled." On May 20, 2026, ChatGPT disproved Erdős’ Unit Distance conjecture in discrete geometry using a theorem by Golod and Shafarevich. While human mathematicians initially verified the proof, the focus quickly shifted to formalization. By May 26, 2026, Logical Intelligence—a company led by Mike Freedman and Yan LeCun—successfully autoformalized the entire proof in the Lean theorem prover. This milestone was subsequently verified by experts Kevin Buzzard and Thomas Browning. The event highlights the accelerating role of AI and interactive theorem provers in establishing mathematical rigor and the rapid transition from human-led discovery to AI-driven formalization.

Hacker News
Share

Key Takeaways

  • AI Discovery: ChatGPT successfully disproved Erdős’ Unit Distance conjecture on May 20, 2026, utilizing the Golod-Shafarevich theorem.
  • Rapid Autoformalization: Within less than a week of the discovery, the company Logical Intelligence autoformalized the entire proof into the Lean theorem prover.
  • Expert Verification: The formalized proof was reviewed and confirmed by prominent mathematicians, including Fields Medallist Mike Freedman and Lean expert Kevin Buzzard.
  • Shift in Mathematical Rigor: The event marks a turning point where AI tools are not only generating proofs but also automating the formal verification process, outpacing traditional human methods.

In-Depth Analysis

The Disproof of Erdős’ Unit Distance Conjecture

The landscape of discrete geometry was significantly altered on May 20, 2026, when ChatGPT identified a counterexample to Erdős’ Unit Distance conjecture. The core of the AI-generated argument relied on the application of a profound theorem in number theory developed by Golod and Shafarevich in the 1960s. This discovery was initially met with a mix of skepticism and intrigue within the mathematical community. However, several human mathematicians who were granted early access to the argument provided testimonies supporting its validity. This event serves as a primary example of "outcounterexampling," where AI identifies flaws in long-standing mathematical assumptions that human intuition had previously failed to uncover.

The Transition to Formalization and Lean

For mathematicians like Kevin Buzzard, who has spent nearly a decade advocating for the use of interactive theorem provers, the discovery by ChatGPT raised an immediate question regarding formalization. While human experts believed the argument, the history of technical errors in human-checked mathematics has led to a growing reliance on systems like Lean. The challenge was to move the ChatGPT-generated paper from natural language into a machine-verifiable format. This transition is critical because it moves the proof from a state of "human belief" to a state of "mathematical certainty" verified by code. The author notes that his own journey into Lean began nine years prior due to a lack of trust in the technical details of human-generated mathematics, making this AI-driven breakthrough a culmination of that shift in perspective.

Logical Intelligence and the Speed of Autoformalization

The most striking aspect of this development was the speed at which the formalization occurred. On May 26, 2026—just six days after the initial announcement—Mike Freedman, the Chief Science Officer for Logical Intelligence, informed the community that their system had successfully autoformalized the paper. Logical Intelligence, co-founded by Turing Award winner Yan LeCun, utilized AI to bridge the gap between the informal ChatGPT output and the rigorous requirements of the Lean mathlib. This autoformalization was not merely a theoretical exercise; it was verified by Kevin Buzzard and post-doc Thomas Browning, who confirmed that the system had indeed translated the proof accurately. This rapid turnaround suggests that the bottleneck of manual formalization is being overcome by specialized AI systems.

Industry Impact

The ability of AI to both discover counterexamples and autoformalize them has profound implications for the future of the AI and mathematics industries. First, it validates the utility of Large Language Models (LLMs) in high-level theoretical research, moving beyond simple coding or text generation into complex logical reasoning. Second, the success of Logical Intelligence demonstrates that the integration of AI with interactive theorem provers like Lean can drastically reduce the time required to verify new mathematical knowledge. This could lead to a new era of "verified-by-default" mathematics, where the traditional peer-review process is supplemented or even replaced by automated formalization. For the AI industry, this represents a move toward "Logical Intelligence"—systems that are not just probabilistic but are capable of producing and verifying absolute truths.

Frequently Asked Questions

Question: What specific mathematical conjecture was disproved by AI?

ChatGPT disproved Erdős’ Unit Distance conjecture in discrete geometry. The proof was based on the Golod-Shafarevich theorem from the 1960s, which was used to construct a counterexample to the conjecture.

Question: What is "autoformalization" in the context of this news?

Autoformalization refers to the process where an AI system automatically translates a mathematical proof written in natural language (like the paper generated by ChatGPT) into a formal programming language (like Lean) that can be checked for absolute correctness by a computer.

Question: Who were the key figures involved in verifying the AI's work?

The AI-generated proof and its formalization were reviewed by several notable figures, including Fields Medallist Mike Freedman (Chief Science Officer at Logical Intelligence), Turing Award winner Yan LeCun, and mathematicians Kevin Buzzard and Thomas Browning.

Read Original Article

Related News

DeepSeek V4 Flash 0731 Achieves Breakthrough ARC-AGI Scores with High Cost-Efficiency
Research Breakthrough

DeepSeek V4 Flash 0731 Achieves Breakthrough ARC-AGI Scores with High Cost-Efficiency

DeepSeek has unveiled the latest benchmark results for its V4 Flash 0731 model, demonstrating exceptional performance on the ARC-AGI (Abstraction and Reasoning Corpus) benchmarks. The model achieved a peak score of 89.0% on the ARC-AGI-1 Semi-Private benchmark and 61.4% on the ARC-AGI-2 Semi-Private benchmark under 'Max effort' conditions. Notably, DeepSeek has optimized these reasoning tasks for extreme cost-efficiency, with costs ranging from $0.02 to $0.04 per task. The results highlight the model's ability to handle complex logical reasoning through three distinct variants—Max, High, and Low—each offering a different balance of accuracy and computational intensity. These findings, verified across 120 tasks in the ARC-AGI-2 Public Eval, position DeepSeek V4 Flash 0731 as a significant contender in the pursuit of advanced machine reasoning.

AI Tutoring and the 'TutorMoments' Challenge: When Should AI Help or Hold Back?
Research Breakthrough

AI Tutoring and the 'TutorMoments' Challenge: When Should AI Help or Hold Back?

The emergence of 'TutorMoments,' a project by the Allen Institute for AI (AllenAI) hosted on Hugging Face, highlights a critical frontier in educational technology: the timing of AI intervention. While modern Large Language Models (LLMs) are optimized for immediate helpfulness, effective pedagogy often requires 'holding back' to allow for productive struggle. This analysis explores the core question posed by the TutorMoments initiative: whether AI tutors can discern the optimal moments to provide assistance versus when to remain silent to foster independent problem-solving. By examining the tension between being a 'helpful assistant' and a 'transformative educator,' we delve into the technical and pedagogical implications of this research for the future of personalized, AI-driven learning environments and the shift toward more sophisticated, Socratic digital tutoring systems.

Microsoft Research Unveils Orchard: A New Open Framework for Scalable Agentic AI Systems
Research Breakthrough

Microsoft Research Unveils Orchard: A New Open Framework for Scalable Agentic AI Systems

Microsoft Research has announced the development of Orchard, an open framework specifically designed to address the challenges of scalable agentic AI. Authored by a prominent research team including Baolin Peng and Jianfeng Gao, the project focuses on providing a robust infrastructure for autonomous AI agents. As the industry shifts from simple conversational models to complex, multi-agent systems, Orchard aims to provide the necessary scalability and openness required for broad implementation. The framework represents a strategic move by Microsoft to standardize the development of agent-based architectures, ensuring that AI systems can operate efficiently at scale while remaining accessible to the global research and development community through an open-source approach.