Sam Altman acknowledges the mess. After months of publishing a torrent of machine-written mathematics, the OpenAI chief told the Cheeky Pint podcast this week that he understands why researchers are struggling to absorb it.
His response is that the firehose will widen. “I think, on net, the world deserves to understand how the universe works,” he said, adding, “But I do have empathy for the method.” He also said the company has spent hundreds of millions of dollars on mathematics and is holding back more findings.
That last point is the awkward one. Mathematicians speaking to The Verge this week say they are frustrated by the reverse problem, accusing OpenAI of gatekeeping after three results were withheld. So the lab is being criticized both for publishing too fast and for not publishing enough.
The common complaint is craft. Hundreds of results have landed this month, many attached to brief write-ups that researchers describe as poorly written and hard to verify. Attribution is thin, context thinner.
Altman’s stated position is that public access to fundamental knowledge “outweighs” the disruption, and that no one should sit on a discovery. He conceded the communication needs work.
That leaves an unusual editorial problem for a laboratory. The output is not the bottleneck. The explanation is.
