2023-09-11

• More Yudkowski (Rationality: From AI to Zombies). The structure of the book is that he wrote 1 blog post every day for like 200-something days. Then the posts became popular and someone turned them into a book. Now, I’m reading that book. I read roughly 1 of the blog posts per day. I’m trying to write about them to share what I find interesting and help solidify the concepts in my own mind. Today, Yudkowski talks about Bayes’ Theorem. He seems to have a real soft spot for this mathematical idea. And I can see why. He makes the argument that Bayes’ Theorem forms the most compelling/succinct description of the scientific method. Via Bayes’ Theorem, we get a definition of both what a theory is/does and also what constitutes evidence. Yudkowski’s ideas make sense to me. It’s definitely exciting stuff. Makes me want to go read the books Yudkowski cites as foundational (e.g. E.T. Jaynes’ probability books). Ok, so what’s so great about Bayes’ Theorem? Yudkowski argues that it connects “rational inference” to “physical causality”. I take this to mean that it describes how to think about what you see in the world around you in a way that makes sense. It provides both sets of important guardrails around thinking: how to do it properly and how to not do it improperly. Of course, many of the details have to be figured out by the practitioner. But having an overarching framework for thinking, if it really does make sense/work, would be quite a big deal.

---

---

---