The Information Bottleneck EP19: AI in Finance and Symbolic AI with Atlas Wang
Dec 10, 2025
Atlas Wang, a faculty member at UT Austin and research director at XTX Research, dives into the captivating realms of symbolic AI and quantitative finance. He unveils how neural networks can learn symbolic equations through gradient descent, challenging our understanding of AI reasoning. In finance, he demystifies high-frequency trading’s complexities, highlighting the struggle against market noise and the pursuit of predictive accuracy. Atlas advocates for robust time-series models and suggests that researchers should explore finance for the unique opportunities it presents.
AI Snips
Chapters
Transcript
Episode notes
From RL To Fast Decision Trees
- Atlas converted a learned RL congestion controller into a decision tree and achieved 400–500x CPU speedups in a constrained environment.
- He also distilled a visual RL agent into symbolic parsers that grounded patches into objects for simple tasks.
Symbolic Forms Are Ultimate Compression
- Clean, low-dimensional symbolic forms are the strongest compression of a neural model because they map to human-readable knowledge.
- Atlas argues symbolic equations are a deeper form of compression than pruning or low-rank factorization.
Run Theory And Experiments Together
- Pursue theory and empirical work in parallel instead of waiting for one to finish.
- Atlas Wang recommends developing rigorous proofs while running practical experiments to accelerate progress.