AI Learns at the Wrong Abstraction Level: Predict Latents, Not Tokens
Machine Learning Street Talk · rss · 2026-08-11
Machine Learning Street Talk (MLST) hosts statistical physicist Matthieu Wyart for a deep dive into the theoretical foundations of deep learning and abstraction.
Key discussion points include:
- Advantages of Deep Networks: Language and images have a hidden hierarchy of parts within parts. Depth allows networks to recover these coarse-grained variables and escape the curse of dimensionality.
- Predicting Latents: Wyart argues that predicting raw tokens is inefficient; predicting latent representations instead could drastically improve sample efficiency.
- Limits of Machine Creativity: Where current systems fall short of genuine scientific invention.
- Interdisciplinary Perspectives: The conversation covers diffusion models, neural scaling laws, Chomsky's context-free grammars, and physics-inspired machine learning theories.
More from Research
- Netflix details its production LLM judge: hundreds of thousands of recommendations scored weekly — omarsar0 · 2026-08-24
- Nature Comment: Provenance, not interpretability, grounds trust in autonomous science — gabepgomes · 2026-08-24
- New Architecture RHEA: Train 1B Model on 8GB VRAM — zemondza · 2026-08-24
- Trained two 16M-param models to do generative CAD with real physics — debreuil · 2026-08-24
- Claude model helps discover complex structure on S^6, solving 60-year-old math problem — Singularitarian · 2026-08-24
- Study: Agents read instructions/notes 60.5% of the time, rarely touch API docs — dair_ai · 2026-08-24