Compact Proofs of Model Performance via Mechanistic Interpretability 2024-06-24T19:27:21.214Z

Comment by rajashree (miraya) on «Boundaries», Part 3a: Defining boundaries as directed Markov blankets · 2023-04-02T23:18:22.985Z · LW · GW

Some notational bugs in the bulleted list in "Definition part (a): "part of the world" that come up in trying to communicate about this with someone:

In the first bullet, $W\in W$ should probably be $w\in W$ both for consistency with later bullets, and so that it does not seem like $T_W$ is subscript-indexed over some $W\in W$. (I'm inferring that subscript W is just part of the name of $T_W$.)

In bullet 6 about $\text{Fut}\left(w\right)$, the sentence should end with $w\in W$, not $w\in W$.

The last three bullet points should maybe use $w_t$ instead of $W_t$, or else maybe all the lowercase $w$s should be made capital.