OpenAI announced ten AI-generated math advances from Astra, its next major LLM — each resolving or advancing a “long-standing open problem,” gathered into a nearly 250-page paper that the company says cost about $2,000 in tokens. The press release framed the results as breakthroughs on problems that “have been open and seen no progress on the main result for at least a decade.” Then mathematicians actually read the paper.
Two of the most exciting results turned out to lean on preexisting work without properly citing it. The improved bound on high-dimensional sphere packing rests on an argument OpenAI presented as its own that first appeared in a 2016 paper by Stephen Miller of Yeshiva University and a collaborator. Miller is blunt: “They are running roughshod over the work of others who came before them in a deliberate way… It seems completely systematic to me, and it points to research misconduct.”
The other flags a group-theory result — a construction of a group that lacks soficity, a property of being approximable by simpler groups. Francesco Fournier-Facio of Cambridge was stunned until he “engaged with this breakthrough as I would if a human had written it”: the key step combines ideas from a 2016 paper and a 2019 paper, the latter co-authored by Andreas Thom of Dresden, who called the move on MathOverflow “creative and at the same time elementary.” Fournier-Facio credits OpenAI’s mathematicians for trying to attribute the work correctly — the failure sits upstream, with “the big PR machine that wants to sound as impressive as possible and does not care about being 100 percent accurate.” OpenAI has since made its press-release language more accurate and says it will make “small updates… consistent with standard academic practice.”
The episode is the AI divide in miniature: superhuman patience for assembling existing puzzle pieces reads as novelty until a human checks the provenance. If OpenAI wants its models treated as participants in high-level research, the field’s answer is that participation carries the field’s norms — citations included. As Fournier-Facio puts it, “OpenAI is now fully participating in high-level research, so they should be held to the same academic standards that we are.”