Hmmm… What matters is the structure you can use to represent a hypothesis space for your purposes, not its size in some silly representation. If you can denote 3^^^^3 states as X and get away with it, it doesn’t matter that the number of states is 3^^^^3 and not (horror!) 3^^^3.
Do you mean that they can’t use Solomonoff’s prior? It’s easy for a computer to be Bayesian about a very simple universe, no?
A sufficiently small, discrete universe with known physics? Yes. But not in real life. All real-world hypothesis spaces are exponential or larger.
Hmmm… What matters is the structure you can use to represent a hypothesis space for your purposes, not its size in some silly representation. If you can denote 3^^^^3 states as X and get away with it, it doesn’t matter that the number of states is 3^^^^3 and not (horror!) 3^^^3.