Why I agreed to join AGMAI

Guest post by Martin Hairer, cross-posted on Tao’s blog on 22 Sept 2026, the day after AGMAI (Advisory Group on Mathematics and Artificial Intelligence) was announced with nine members. Hairer says the group is “genuinely independent of OpenAI and any other of the so-called” frontier labs, that members are unpaid and that technical support (IT, legal, communications) comes from the IAS, and that “our overarching aim is to represent as well as we possibly can the interests of the mathematical community”. He admits the group may fall short and its advice may be ignored. His reasons: falling publication standards around recent AI results, the reality of AI’s impact on research, the chance to use OpenAI’s belated outreach, and consistency with the “A Severe Misalignment of AI in Mathematics” statement (nearly 8,000 endorsements by then): “it would be rather hypocritical of myself to then chicken out at the first opportunity”. About 50 comments, many critical (confidentiality terms, composition, leverage over AI companies).