This only seems true because of your practical experience with what we humans call computer science. As I mentioned in a previous post, there is a mathematical solution to this (Solomonoff Induction applied to AIC) which has never been solved as a matter of practical implementation. We know it is solvable (i.e. we could develop a software expression of it), but our view of computer science is colored by the fact that no one has. If you want to make a mint and become famous, that is a problem worth some effort.
Your objection only really stands because no one has yet solved a problem that we know to be solvable. In this sense, the objection is weak because it will obviously collapse the second some bright person publishes a decent universal implementation of the mathematics and it gets incorporated into the body of knowledge we call "computer science".
Weak or not, my objection is valid at the present time. However, I think it is unsolvable. Yes, you can compare two algorithms to see which is more efficient, but you cannot determine whether there is a better algorithm unless you can write a better one yourself.
This leads me back to the basic problem with these 'rules' of Wolfram. The problem is not so much description as it is of creation. It should be noted that while Wolfram believes that such can be done - which is the real test of such a theory - he himself does not claim that he has achieved it.