AI and number theory
May 27 to May 30, 2026
at the
American Institute of Mathematics,
Pasadena, California
organized by
Tudor Achim,
Brian Conrey,
David Farmer,
Alex Meiburg,
and Michael Rubinstein
Original Announcement
This workshop is motivated by the very general questions: What problems in research mathematics
are amenable to attack by AI, and how can general AI be improved to be more
successful, and more widely applicable, to research mathematics?
The workshop will focus specifically on problems in number theory. Two categories within that area will receive particular attention:
-
Improving bounds in theorems, both at the level of main terms and
lower order terms
-
algorithms: improving existing algorithms and implementing new ones
Each problem in these areas could potentially serve as a benchmark
for testing the capabilities of general AI. An outcome of the
workshop will be a curated list of interesting problems, several of
which have been converted into official benchmarks with an indication
of the current limits of AI. The list will serve as a permanent and
growing resource for research mathematicians and AI developers.
Material from the workshop
A list of participants.
The workshop schedule.
Workshop videos
The problem list and a selection of those problems as benchmarks in lean.
