UW students stress-test Ai2's AutoDiscovery science agent in 10-team challenge
allen_ai · x · 2026-09-15
Ai2 detailed how University of Washington students stress-tested AutoDiscovery, its AI research agent that analyzes datasets, proposes hypotheses, runs experiments and ranks findings by Bayesian surprise. From 25 proposals, 10 teams joined the GenAI for Science: Ai2–UW Materials Challenge, surfacing a possible flaw in battery aging measurement, an odd pattern in silicon polymers, and recurring features in light-responsive proteins. The class centered on interrogating AI output—checking evidence against literature and separating real discoveries from coincidence or data issues. Free credits extended through Dec 31.
Related event: 25 UW Teams Stress-Test Ai2's AutoDiscovery Science Agent(4 posts)→
More from Apps
- Cohere releases Parse 5, pitching cheaper document parsing — cohere · 2026-09-15
- Whatever Happened to AI Browsers That Were Going to Kill Chrome? — cdrfrk · 2026-09-15
- Dev builds handwriting canvas where AI sees your math notes, plus a Tom Riddle diary mode — whizzkidme · 2026-09-15
- Google DeepMind's WeatherNext 3 brings turbine-height wind and solar forecasts to grid operators — GoogleDeepMind · 2026-09-15
- Prompt-to-LUT workflow: generate custom .cube color grades and test in CapCut PC — AIwithGhotai · 2026-09-15
- Crator pitches itself as the Lovable for ERPs, building on ERPNext in minutes — Scobleizer · 2026-09-15