What Is Physical AI

Any system that perceives the real world and interacts with it.

Ideas I want to discuss

• Biology parallel
• architecture idea found relatable to hardware-software co-design of computer architecture by onur mutlu
• Parallel processes as default paradigm interacting with each other
• Keep systems engineering in mind and I also want to learn about this as we go
• I want add anti fragility as dimension so we talk not just about robustness but adaptability at all layers

Processes are almost always incremental and iterative, maybe that's how processes in our robots be.

Prerequisites

Introduction and architecture first two posts should be accessible to anyone with enough curiosity to handle the content and with multi domain surface level knowledge like decent idea of STEM related fields, for upcoming hub and spoke posts, hub posts should be good overview of the physical ai stack layer but might require more in depth curiosity on readers side and then spoke posts are technical research survey and commentary for people wanting to pursue research or have an overview of the field.

What and why

This is my personal roadmap and how i think we should approach physical ai, there could be different opinions and I would try to also mention some very popular ones that are currently industry standard and try to make an argument about tradeoffs, my own learning journey and perspective that I'm making while trying to understand the problem from multiple facets and coming up with best way to comprehend it.

This guide will also act as overview to the field who has no STEM background but are curious enough to look things up and big picture for those who usually work with few layers or 1 but not all, for them to see how everything connects, systems engineering perspective.

This is not a robotics primer — please checkout robotics primer for that.