Mathematicians clash with AI labs over models racing them using their preprints
suchenzang · x · 2026-09-09
A dispute is roiling the math community: a paper explaining a counterexample to the Jacobian conjecture allegedly drew from an earlier Borisov-Gabber-Vasiu preprint, prompting public questions about how the counterexample was obtained. Yoav Goldberg argues that consenting to models trained on your data is very different from internal models trained on your latest work racing you to the solution; Boaz Barak notes Anthropic has similar opt-out policies for training data.
Ulam's Sochen Zang mocked the situation: researchers contribute data with stripped PII while frontier models pump their aura on uncited 'breakthroughs' built from that work.
More from AGI Musings
- Nonfiction book market is collapsing, and authors are memeing about it on X — jjvincent · 2026-09-09
- Former Mosaic researcher mocks the "user data flywheel" moat narrative in viral thread — bookwormengr · 2026-09-09
- Does AI Automation Kill Accidental Breakthroughs? Lessons From CMB and CRISPR — jjvincent · 2026-09-09
- FT's Burn-Murdoch: ChatGPT-assisted coursework means schools are no longer assessing kids at all — jburnmurdoch · 2026-09-09
- Anthropic researcher: novelty-rewarded RL may teach agents to obfuscate their sources — suchenzang · 2026-09-09
- "A country of geniuses in a datacenter": Tuvalu-scale AGI quip — nabla_theta · 2026-09-09