OpenAI has helped establish an independent advisory group of mathematicians to review how AI-generated mathematical discoveries are evaluated, communicated and released. The group arrives after criticism from prominent mathematicians over the speed at which OpenAI has been announcing results from an internal model capable of tackling major open problems.
The Advisory Group on Mathematics and Artificial Intelligence is hosted at the Institute for Advanced Study in Princeton, New Jersey. OpenAI describes it as a bridge between the company, the mathematical community and the public, giving outside mathematicians more influence over how significant AI-generated results are presented.
Group Will Review Results but Cannot Slow Research
The nine-member group will assess the significance of mathematical results and advise OpenAI on how their release should be coordinated. Members are unpaid, can publish their own views and control membership independently of OpenAI.
The group itself stresses that its role is advisory. It has no decision-making power at OpenAI or other AI companies, and OpenAI has explicitly said it will not advise the company on how quickly its internal mathematics research should progress.
Its immediate task is unusually large. OpenAI says the internal model behind its recently published Navier-Stokes result has also resolved more than 100 additional open problems spanning much of mathematics, creating questions about how those results should be reviewed and released.
The announcement follows OpenAI’s September publication of an AI-generated proposed solution to the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. The release drew scrutiny from mathematicians over attribution, verification and the pace at which AI-generated discoveries were being made public.
Mathematicians Have Called for More Deliberate Releases
Earlier this month, 25 Fields Medal winners signed an open letter warning that competition between AI labs could damage mathematical research if companies rush to announce solutions without allowing enough time for verification, proper attribution and integration into the wider mathematical literature.
The signatories argued that mathematical progress involves more than producing correct proofs. Researchers also need to identify new techniques, explain their significance, connect them with previous work and transmit that knowledge to other mathematicians.
Only one member of OpenAI’s new advisory group, Institute for Advanced Study professor Camillo De Lellis, also signed that letter. The wider independent group says it is willing to provide recommendations to any AI company whose models could significantly affect mathematics, rather than operating solely for OpenAI.
Its first challenge will be helping determine how a potentially large backlog of AI-generated mathematical results can be communicated without overwhelming the researchers expected to verify, interpret and build on them.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.