~/talk

Foundations for AI+Science

Talkat PyTorch Conference 2025 · 2025-10-22..10-23· [machine-learning]
links:none on file
slides:none on file

Introduces neural operators that map between continuous function spaces so models run at any resolution and learn from discretized samples, claims the 2021 AI weather model trained on 40 years of weather data runs tens of thousands of times faster than traditional solvers and enables calibrated probabilistic ensembles in ForecastNet3; reports AI-designed devices including a catheter that reduces bacterial contamination 100-fold and open‑source PyTorch libraries (Neural Operator library, Lean Dojo) that couple physics constraints and formal proof verification for science.

AI-written summary · about

  1. 0:00Welcome
  2. 0:41Language limits for science
  3. 3:13Neural operators for physics
  4. 6:46Physics-informed learning
  5. 8:11Neural operators architecture
  6. 9:38AI weather models
  7. 11:48AI in medicine and chips
  8. 13:32Continuous modeling
  9. 13:54AI for math reasoning
  10. 15:40Hardware for AI training
  11. 17:29Low-rank gradient training
  12. 18:55Hardware-aware learning
  13. 19:16Future of AI and science
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.