Interpretability paper finds a faithful internal world model inside taxiGPT
__nmca__ · x · 2026-09-23
The new paper "World Modeling in Transformers" (with authors from MATS) asks whether AI models can recover world models purely from data.
In taxiGPT, the authors found a faithful internal map of the world, and traced the model's failures to how that map is stored in superposition. The finding directly counters the common "LLMs don't model the world" argument — empiricists looked inside and found a little model of the world.
Related event: Paper: Transformers Do Have Faithful World Models, Just Scrambled(3 posts)→
More from AGI Musings
- Investor: AI diffusion will be slowest in consumer software — Oura, Strava, Labcorp still have no real AI strategy — MartinGTobias · 2026-11-12
- On AI Progress: When Models Lag on Your Load-Bearing Skill, You Can't See Them Improving — dfrsrchtwts · 2026-11-12
- Slack's David Hoang: agents beat people on coordination, proposes 'boundary agents' — davidhoang · 2026-11-12
- CS student mourns coding flow states he'll never experience in the AI era — chrissyykat · 2026-11-12
- Ford's 370% Turnover Led to $5 Days — But ASI Never Quits — krishnan · 2026-11-12
- Against 'verifiable vs. fuzzy': the real axis is how expensive tasks are to grade — dfrsrchtwts · 2026-11-12