I still have question, as a non-expert who wants to develop his own long-term learning path (or at least have my own educated opinion about the current fast-paced AI hype):

To become more familiar with Reinforcement Learning (RL), is it useful to turn my attention to basic Agent-Based Modeling first (like the NetLogo App does it) in order to learn some basics about RL first, and then add in more math and ML stuff gradually?

It seems to me like most prospective data-scientists move on a path like this: Linear regr. -> Supervised ML -> Unsupervised ML -> NNs -> DL -> RL .

Agent-Based Modeling seems undertaught.

Maybe you should post again, with "author here" at the beginning, in a more popular HN discussion-thread on MuZero? According to Algolia HN search, this thread has the most comments, n=76) currently (Dec 25 20202) https://news.ycombinator.com/item?id=25519176