This is a little esoteric, but, also, extremely important: No matter how cleverly we program a Von Neuman machine, internally it's still rules-based and completely deterministic. (I realize I'm ignoring the fuzziness of floating point numbers.)
Reality, however, always exhibits degrees of indeterminacy.
So do neural nets.
One can easily extrapolate the rest of what I was going to say.
TY for such nice work you guys are up to!