Deciphering Human Motion AI Models Using Sparse Autoencoders (SAEs)
mathildepapillo · x · 2026-08-05
A researcher utilized GoodfireAI's newly released Silico tool to break down and analyze general-purpose human motion AI models.
The study shows these models can meaningfully detect Parkinson's gait impairment. Using Silico, the researcher was able to look under the hood and figure out exactly which aspects of a person's walk drive the model's predictions.
Related event: Silico Tool Decodes Gait AI to Detect Parkinson's Disease(2 posts)→
More from Research
- Goodfire Launches Silico: AI Agents Can Autonomously Run Experiments for Days — mathildepapillo · 2026-08-05
- Impressive Progress on NIST Robot Benchmark: Models Master Complex Contact-Rich Tasks — chris_j_paxton · 2026-08-05
- Visualizing Qwen2.5-VL's "Thoughts" Using Goodfire Silico — ninamiolane · 2026-08-05
- Google's AI Overview Flips Answer Based on Single arXiv Preprint — sayashk · 2026-08-05
- Cursor Releases Mixture-of-Kittens Megakernel for MoE, Claims Nearly 2x TFLOP/s — CapnHat · 2026-08-05
- OpenADMET Launches 3rd Challenge: Predicting CYP Inhibition — CatAstro_Piyush · 2026-08-05