Oh, but for a Bayesian super intelligence it’d be easy! That’s actually true in a way: Bayes theorem and (actual science-oriented, rather than bullshit handwavey) statistics involving it are about how conditional probabilities interact, including, overly simplified, how to get from prior probabilities (what i find how probable before looking at data) to posterior probabilities (what i find how probable conditioned on data). Basically what it means to throw “Bayesian” into the word soup (because a stew requires more substance) is that the fairy tale is not about some blank slate figuring out everything from minimal data and first principles, but that it’s about a model that starts accurately and doesn’t have to adjust much (because otherwise by necessity it’d need more data - otherwise it’s not Bayesian). God I hate this