OpenAI’s Math Solutions Fall Short of Mathematicians’ Standards
· Technology · TechCrunch
OpenAI released hundreds of AI-generated solutions to complex math problems, consulting an advisory group to avoid past controversies. The solutions lacked human understanding, with only 42% of proofs formalized as recommended. The group, AGMAI, had urged frontier labs to stop testing proprietary models on open math problems—a practice OpenAI continues. Only ten of the 719 manuscripts included the model's chain of thought, raising concerns about accountability and meaningful human engagement with the results.
Why it matters
This critique highlights the gap between AI's mathematical outputs and the rigor expected by human experts, potentially delaying the field's trust in AI-assisted proofs.
Read the original report — TechCrunch
Join us on Telegram
Breaking news the moment it lands. At 10,000 members we ship the Android app.