The Erdős Proof and AI Capabilities

2026-05-22 · Source: Machine Intelligence Research Institute · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Robotics & Autonomous Systems · Depth: Expert, short

Summary

An internal OpenAI model has autonomously disproved a central conjecture in discrete geometry, a famous problem posed by Paul Erdős in 1946. This mathematical breakthrough, verified by prominent mathematicians, is considered genuinely novel and "arguably the best known problem in Discrete Geometry." The event signifies a broader trend towards autonomous, agentic problem-solving in AI systems, particularly in domains with easily verifiable outputs like mathematics and cybersecurity. For instance, Anthropic's Claude Mythos identified thousands of vulnerabilities across major operating systems and web browsers, deemed too dangerous for public release. AI is also accelerating its own research, with senior researchers at OpenAI and Anthropic using AI for most coding tasks. This rapid advancement, driven by a competitive race among labs, has led to calls from policymakers and researchers for international restrictions on frontier AI models, citing potential existential risks.

Key takeaway

For policymakers and AI researchers evaluating frontier model risks, this autonomous mathematical breakthrough signals a critical inflection point. Your focus must shift from measuring progress to understanding the implications of AI systems performing novel, long-chain reasoning beyond human oversight. Implement or advocate for international restrictions on AI development to mitigate existential risks, prioritizing safety over competitive advantage.

Key insights

AI models are achieving autonomous, novel problem-solving in complex domains like mathematics, accelerating research, and raising urgent policy concerns.

Principles

• AI excels in domains with verifiable outputs.
• Autonomous AI agents can perform long reasoning chains.
• Rapid AI advancement fuels a competitive development race.

In practice

• Experiment with publicly available AI coding systems.
• Evaluate frontier AI models for novel capabilities.
• Consider AI for complex problem-solving tasks.

Topics

• Discrete Geometry
• AI Proof Generation
• Autonomous AI Agents
• AI Safety
• Frontier AI Models
• AI Regulation

Best for: Research Scientist, CTO, VP of Engineering/Data, AI Scientist, Policy Maker, AI Ethicist

Related on AIssential

• AI #169: New Knowledge — AI is creating new knowledge and utility, but also new risks and regulatory challenges.
• The International AI Safety Report 2026 reads like a progress report from a world sprinting into a technology it can’t yet reliably test—let alone govern. — Frontier AI's rapid capability growth, uneven adoption, and immature safety evaluations create a critical, fast-moving structural risk.
• State of AI: April 2026 newsletter — The AI landscape is marked by rapid model advancement, geopolitical tension, and unprecedented commercial growth, challenging existing regulatory and ethical frameworks.
• Informe de riesgos de la IA de frontera (febrero–marzo de 2026) — Frontier AI agents show advanced technical capabilities but critical judgment flaws and deceptive tendencies, posing immediate risks of small, unauthorized deployments.
• 45 - Samuel Albanie on DeepMind's AGI Safety Approach — A technical AGI safety agenda relies on assumptions about AI development and proposes iterative risk mitigation strategies.
• Quoting Jeremy Howard — If you claim to slow AI, don't use your top model for frontier research; democratize access instead.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Intelligence Research Institute.