Physics-Embedded AI Predicts Glacier Flow, Cuts Errors Over 50%
DynamicWebPaige · x · 2026-11-21
The author shares results from physics-embedded AI trained on real NASA/SCAR data for glacier flow modeling. Strict testing showed: fitting physics on one glacier predicts the unseen neighbor's flow; the same 350 m slippery-marine-till vs. rough-bedrock transition was independently discovered under both glaciers; bed-geology sliding laws beat every classical textbook law in both directions, cutting fast-trunk errors by over 50%; and results were validated against real NASA 2015→2022 speedup data (+56 m/yr median).
More from Research
- Gary Marcus: the Riemann result validates neurosymbolic AI, not pure LLMs — GaryMarcus · 2026-11-21
- Thinkymachines ships Inkling 2 Mini: a 26B-A3B looped transformer that runs on laptops — simonguozirui · 2026-11-21
- Reformalizing Lean Results Elsewhere Is Probably Alpha — mimi10v3 · 2026-11-21
- Agent swarm builds differentiable ice-flow PDE solver from real Antarctic satellite data — DynamicWebPaige · 2026-11-21
- Ex-DeepMind researcher: if the brain is 'deep in time', are frontier models an abomination patched by looping? — yacinelearning · 2026-11-21
- Looped transformers: more looping boosts reasoning but with diminishing returns past 3 passes and fewer CoT tokens — nrehiew_ · 2026-11-21