Volume IV: Physical AI Systems
Machine learning systems that sense and act in the physical world.
Note
In development. This volume is being written now and changes quickly as I iterate. Chapters are added, reorganized, and rewritten often, so please do not cite or teach from it yet. Feedback is welcome through the book feedback issue forms.
About This Volume
Volume IV covers machine learning systems whose outputs are physical actions. It begins at the causal boundary, the point where a command becomes current in a winding and mass begins to accelerate, and follows the consequences of that boundary through the design of the body, the brain, and the nervous system that connects them.
The governing constraint is irreversibility after actuation. An advisory prediction can be reconsidered before it commands an actuator, though errors may still have consequences when acted upon. A mistimed motor command delivers energy into an environment that will not roll back. Every design decision in this volume is shaped by that asymmetry.
Structure
This volume follows the content layout shared by all four volumes:
vol4/
├── index.qmd volume home page
├── frontmatter/ author note, about, prerequisites, notation
├── parts/ part openers and part summaries
├── NN_chapter/ one directory per chapter: NN_chapter.qmd plus its images and data
└── backmatter/ references, appendices, glossary
Chapter order comes from books/config/_quarto-pdf-vol4.yml, and books/shared/STRUCTURE.md lists it for every volume.
The proposed Physical AI course is a working instructor draft. Its hardware studio plan is in labs/vol4/README.md.