Universal mix
Solomonoff (1978): x=010011010010..., next bit? Assume: bits are drawn from recursive measure µ - there is a program p that reads x, computes µ(x), halts. Build Bayesmix M(x)= weighted sum of all enumerable measures.
M dominates all µ: M(x) ? cµ µ(x) for all x
Solomonoff / Kolmogorov: x “simple” if M(x) high / K(x) low (almost but not quite the same)
Back to J. Schmidhuber's Theory of Everything page