‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest drop

Key Points:

• OpenAI released nearly 400 AI-generated mathematical results across 700+ manuscripts covering diverse fields, overwhelming mathematicians who estimate it may take years to fully understand and verify the work.
• While some results appear impressive and could merit top-tier journal publication or major awards, many manuscripts lack formal verification, contain inconsistent quality, and have been criticized for poor explanations and insufficient attribution.
• The flood of AI-generated mathematics has disrupted the research community, with some researchers' entire programs reportedly "obliterated," causing anxiety especially among early-career mathematicians reliant on open problems for their work.
• OpenAI engaged with the Advisory Group on Mathematics and Artificial Intelligence (AGMAI) and made some improvements, such as providing more transparency and committing to workshops, but did not fully adhere to recommendations like disclosing models and prompts or thoroughly formalizing proofs.
• Mathematicians broadly acknowledge AI’s potential to advance the field but express concern over the rapid, large-scale releases that shift the burden of verification and contextualization onto the community, creating social and professional challenges amid ongoing uncertainty about the future of mathematical research.