Google, IBM & Meta Certificates — All 10,000+ Courses at 40% Off
One annual plan covers every course and certificate on Coursera. 40% off for a limited time.
Get Full Access
Watch a Harvard CMSA New Technologies in Mathematics seminar where Meta AI researcher François Charton explores the bidirectional relationship between transformers and mathematics. Discover how transformer models can be trained to tackle mathematical challenges, including integer sequence prediction and analyzing scattering amplitudes in Quantum ChromoDynamics-adjacent physics. Learn through concrete examples from linear algebra and integer arithmetic how mathematical problems provide insights into transformer behavior, demonstrating that model predictions can be explained, models maintain accuracy without confabulation, and training distribution selection impacts performance and robustness.
