I think he does a pretty good job of explaining a theory in which that isn't the case - that a) in order to predict (compute) "you" at at point, you have to ac

I think he does a pretty good job of explaining a theory in which that isn't the case - that a) in order to predict (compute) "you" at at point, you have to actually "run" you at every time-step up until that point; this is specifically analogous to NP-completeness: result verification is identical to result generation; which could be meaningfully rephrased as "there's no shortcut to figuring out what you will do than going out and doing it." b) There's a gap between non-determinism and randomness. This reminds me of the basics of Kolmogorov complexity: the number of bytes to transmit a random string is the same as the size of the random string, except instead of "random" you have "non-deterministic". You are a non-random non-deterministic string; there is no shortcut in computing you than actually doing it... with the rest of the universe. (Which may mean there's no way to transmit you, either...?)