On Thu, Oct 11, 2018 at 2:46 PM John Rose <[email protected]> wrote:

> > A string is random if there is no shorter description of the string.
> > Obviously this depends on which language you use to write descriptions.
> > Formally, a description is a program that outputs the string. There are no
> > "shades" of randomness. A string is random or not, but there is no general
> > algorithm to distinguish them in any language. If there were, then AIXI and
> > thus general intelligence would be computable.
> >
>
> Exactly! But you don't know generally if there is a shorter description do 
> you? So the compression agent does its thing and executes programs and tries 
> to find out. Until it runs out of resources for whatever reason. In absolute 
> case there is only random and non-random. In a real world there are shades of 
> randomness I think? Something appears random to one agent but not another?

Encrypted data appears random if you don't know the key. But it is not
random because it has a short description (compressed plaintext +
key). Kolmogorov proved that there is no general algorithm to tell the
difference. If there was, then I could describe "the first string (in
lexicographical order by increasing length) that can't be described
using a million bits" even though I just did.

This is why there is no simple, general algorithm for intelligence. We
predict sequences by compressing them into short programs, then run
the program to predict the next symbol. By Occam's Razor, the shortest
program gives the best prediction. But there is no way to know whether
you found it or if there is a better theory.


-- 
-- Matt Mahoney, [email protected]

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T55454c75265cabe2-M66735a9428e04abdf4ace22f
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to