Perhaps I am beating a dead horse for this community, but this is a very lucid explanation of what probabilistic/statistical reasoning cannot do. Namely: first order logic. There's really no way of encoding relations or quantifiers into purely Bayesian inference, which actually makes it quite weak in terms of model building.
Further, integrating probability and logic is a huge unsolved problem! We actually have very little idea how to combine our two greatest successes in formalizing rationality.
I found this tremendously clarifying, though not immediately useful. But it has definitely broadened my thinking.