[EMAIL PROTECTED] (Juergen Schmidhuber) writes:
>
> Hal:
>
> >What about the fact, which I mentioned earlier, that probabilities are
> >based on Kolmogorov complexity, but that is uncomputable, or at least
> >it would require a super-TM to compute it.  However we do observe
> >probabilities and their effects, hence K. complexity is being computed,
> >hence there must be a super-TM somewhere to compute it.
>
> The precise probabilities are never really computed. 

Perhaps not, but approximate ones are computed and we observe the effects,
aren't they?  Approximate probabilities based on approximations to the
K. complexity of a string are no more computable than precise ones.
There is no fixed bound B which allows you to compute the K. complexity
of an arbitrary string within accuracy B.

Hal

Reply via email to