OpenAI turns to elite mathematicians to avoid fumbling again

A run of spectacular mathematical results turned into a reputational black eye for OpenAI, and now the company is asking humans — mathematicians — to help it find a less disastrous way forward.

On Monday OpenAI announced an independent panel, the Advisory Group on Mathematics and Artificial Intelligence (AGMAI), meant to advise the company and other AI labs on how they interact with mathematical research and the wider community, including how new results are presented and released. Its abrupt arrival caught many mathematicians off guard. Researchers told The Verge the group is a decent first step, yet many were left with basic questions: what will it actually do, how much sway will it have, will OpenAI really listen — and can such a small band of stars speak for everyone?

Who is on it

The nine members are drawn from Stanford, Harvard, Oxford and Cambridge, with multiple Fields Medals and MacArthur "genius" grants among them. The group will be hosted by the Institute for Advanced Study (IAS) in Princeton, once the intellectual home of Einstein, von Neumann and Oppenheimer.

OpenAI stresses the panel's independence: members can offer unsolicited advice, comment publicly, and make their recommendations public. They are unpaid, and the group can change its own membership. Responding to what he called a "torrent of misinformation" around its formation, member Martin Hairer said the mathematicians signed no restrictive agreements beyond ordinary confidentiality, and that IAS provides technical support.

The doubts

Not everyone is convinced. Simon Machado of ETH Zurich called the whole thing "shrouded in mystery" and asked how the nine were chosen. Francesco Fournier-Facio of Heriot-Watt University said the panel feels like "an ivory tower" and questioned whether it reflects those whose work and livelihoods OpenAI's trophy hunt is upending. "They're amazing mathematicians, but I don't know if they are the people I want to represent me in more political questions," Machado said.

Hairer concedes the group is hardly a representative sample, but insists all nine are acting in good faith. He also harbors no illusions: the labs, he wrote, will surely try to spin whatever the group says to suit a PR machine that "dwarfs anything we could possibly come up with."

The real test is ahead

Announcing the panel, OpenAI casually noted that its still-unreleased model — the one said to have resolved the Navier-Stokes Millennium Prize problem — has now "resolved more than 100 long-standing open problems across most areas of mathematics." Clearing that logjam is the group's first stated priority: advising OpenAI on how to coordinate the release of a large trove of results.

Kevin Buzzard of Imperial College London wonders why OpenAI needed such an elite body at all: what the community wanted was not proofs of hard theorems but "better understanding of our subject." "It's not entirely clear to me that you need a committee of brilliant people to hammer the point home."

Source: The Verge