Neo Labs and The Limits of AI Math & Science

· Source: AI Supremacy · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Mathematics & Computational Sciences · Depth: Expert, extended

Summary

On May 20, 2026, an OpenAI reasoning model disproved the 80-year-old Erdős unit distance conjecture, a problem in discrete geometry, combinatorics, and number theory. The AI found an infinite family of point configurations where unit-distance pairs grow like n to the power of 1.014, polynomially faster than Erdős's conjectured rate, though still below the n^(4/3) upper bound. This breakthrough involved synthesizing three existing mathematical techniques, not inventing new ones, and was verified by nine human mathematicians. Concurrently, the generative AI economy generated \$110 billion in sales over the past 12 months, with an annualized run rate exceeding \$175 billion. This period also sees the rise of "Neo Labs," highly funded AI startups like Mirendil and Engram, pushing frontiers in areas like "self-accelerating" AI and persistent "learned memory," amidst concerns from the Bank for International Settlements about the 2025–2026 AI investment cycle and a 20% increase in consumer hardware prices due to HBM supply chain issues.

Key takeaway

For AI Scientists and Directors of AI/ML evaluating advanced AI capabilities, recognize that while models like OpenAI's can achieve significant breakthroughs by synthesizing existing knowledge, their outputs demand rigorous human verification. You should prioritize developing robust internal validation processes, as current AI systems lack inherent self-correction. Furthermore, consider investing in "Neo Labs" focusing on physical AI or world models, as these specialized approaches are driving the next wave of scientific and industrial innovation beyond traditional LLMs.

Key insights

AI can synthesize existing mathematical concepts to achieve novel scientific breakthroughs, but human verification remains critical.

Principles

Method

OpenAI's model disproved the Erdős conjecture by combining Ellenberg-Venkatesh techniques, the Golod-Shafarevich theorem, and Hajir-Maire-Ramakrishna methods, specifically by letting the number field's degree grow without bound.

In practice

Topics

Best for: Research Scientist, AI Scientist, Investor, Director of AI/ML

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Supremacy.