Re: [sl4] Rolf's gambit revisited

From: Peter de Blanc (peter@spaceandgames.com)
Date: Tue Jan 06 2009 - 11:17:58 MST


Benja Fallenstein wrote:
> Matt was replying to a scenario where the two AIs know each other's
> source code, and claiming, as I understand, that they cannot possibly
> conduct a dialog by simulating each other. IMHO that makes it clear
> that his statement meant something more specific than you're making it
> out to be.

Hi Benja,

I think "X simulates Y, therefore K(X) > K(Y)," is pretty unambiguous in
its meaning, and it's wrong. Also, two AIs certainly could conduct a
dialog by simulating each other, if we're talking about Turing machines
with infinite tapes.

> Matt, as far as I can see you're wrong, though; K(X) = K(Y), no
> contradiction. Yes, if X "did something besides simulating Y," in a
> certain sense, then you would have K(X) > K(Y).

"Kolmogorov complexity" does not refer to the same thing as the English
word "complexity." In particular, doing more things can actually
decrease your Kolmogorov complexity. A program that outputs Wikipedia
would be pretty long, but a program that outputs every string, including
Wikipedia, can be very short.



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT