[mjevans]

My issue with the second statement, about knowing the exact state of the universe at any given reference time is that by definition the information within that state would require the entire space of the universe to store with sufficient detail to make an accurate prediction of future states. (One might also assume it would require a real universe's worth of processing power to compute a new state as well.)

I believe it's impossible to completely isolate any segment of that universe (E.G. to make it smaller and thus predictable within the capability bounds of a larger universe) without literally removing it from that universe. That no matter what every part of an existing universe interacts with every other part, even if very, very, indirectly.

As for the question of free will: I believe the biology is largely deterministic. For me, that leaves the main set of questions in the direction of all of the elements that might happen between, outside, or otherwise beyond our current understanding of how the universe works. I feel that if there is any actual freedom in free will that is where it comes from; otherwise it's just the RNG being too complex to understand completely masking the lack of actual choice.

[chr1]

The fact that nothing can propagate faster than light speed, allows to isolate segments of universe from one another. Of course you can not simulate the whole universe exactly, but simulating a part of it, like completely closed room, or a different smaller universe, is still interesting and useful.

Wolfram proposes an interesting solution to the question of free will, that does not require any randomness: computational irreducibility. It is the hypothesis that for some computations there is only one way to perform. That is if you try to predict what an AI will chose, your only option is to create an exact copy and let that copy to make the choice.