AXRP - the AI X-risk Research Podcast 36 - Adam Shai and Paul Riechers on Computational Mechanics
Sep 29, 2024
Adam Shai, co-founder of Simplex AI Safety, dives into the realm of computational mechanics and its application to AI safety. He explores how computational mechanics can improve our understanding of neural network models, especially in predicting outcomes. The discussion covers the intriguing world models that transformers create and how fractals emerge in these networks. Shai also highlights the potential of combining insights from quantum information theory with computational mechanics to enhance AI interpretability.
AI Snips
Chapters
Transcript
Episode notes
Computational Mechanics Basics
- Computational mechanics, rooted in physics, helps predict random processes.
- It clarifies "world models" by examining next-token prediction implications.
Computational Mechanics vs. Bayesian Inference
- Bayesian inference requires defining objects for updates.
- Computational mechanics helps understand these objects by compressing past data for future prediction.
Epsilon Machines and Data Presentations
- Epsilon machines are minimal for prediction, not generation.
- Different presentations of data are suitable for answering various questions.