Solomonoff’s prior can’t predict something uncomputable, but I don’t see anything obviously uncomputable about any of the 3 statements you asked about.
Yes. Anything that can be represented by a turing machine gets a nonzero prior. And its model of itself goes in the same turing machine with the rest of the world.
Solomonoff’s prior can’t predict something uncomputable, but I don’t see anything obviously uncomputable about any of the 3 statements you asked about.
Right. But can it predict computable scenarios in which it is wrong?
Yes. Anything that can be represented by a turing machine gets a nonzero prior. And its model of itself goes in the same turing machine with the rest of the world.