Assume: all data are results of a computation. No physical evidence against this! Solomonoff weaker: he only assumes computable probabilities.
A priori probability of data goes down with costs: ?x costs > O(n) P(x) ? 1/n
E.g., on your computer: most data computed within a few ms, few data require hours, fewer require weeks..
Add flesh: which are minimal costs for x?
Back to J. Schmidhuber's Speed Prior page