OpenAI claims a general-purpose reasoning model found a counterexample to Erdos’s unit-distance bound [D]

**What Happened:**
OpenAI has claimed that one of its general-purpose reasoning models discovered a counterexample to Erdős’s conjecture regarding the unit-distance problem in planar geometry. The model found a construction disproving the expected upper bound of \( n^{1+O(1/\log \log n)} \), challenging a long-standing mathematical belief.

**Why It Matters:**
This claim is significant because it represents the first time a general-purpose AI has autonomously generated a proof that could have implications for known conjectures in mathematics. If verified, this would mark a milestone as the model not only found a counterexample but also did so without any specific guidance tailored to the problem at hand.

– **First Autonomous Proof:** This suggests that advanced AI models are capable of conducting genuine research and discovering new mathematical truths on their own.
– **Complexity Handling:** The discovery required handling intricate geometric configurations, indicating the model’s ability to reason about complex problems beyond its initial training objectives.
– **Verification Needed:** While OpenAI has provided a proof and some details about how it was found, further verification is necessary to confirm both the validity of the counterexample and the process by which the AI arrived at this result.

These points underscore the potential for AI in mathematical research but also highlight the need for rigorous validation.

Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.

Please enable JavaScript in your browser to complete this form.

Name Email

Name *

First

Last

Email *

Submit