Re: [sl4] Rolf's gambit revisited

From: Matt Mahoney (matmahoney@yahoo.com)
Date: Mon Jan 12 2009 - 11:21:10 MST


--- On Mon, 1/12/09, Benja Fallenstein <benja.fallenstein@gmail.com> wrote:

> Hi Matt,
>
> On Mon, Jan 12, 2009 at 3:23 PM, Matt Mahoney
> <matmahoney@yahoo.com> wrote:
> > Simulating past versions of each other does not allow
> > 2 way communication.
>
> Why not?
>
> * You say something (to yourself).
> * You simulate the other AI until *it* has simulated *you*
> enough to see what you said.
> * You simulate it some more, until it says something (to
> itself) in return.
> * You say something in return (to yourself).
>
> Do you see a reason that wouldn't work?

If X simulates Y, then X must know Y's state. That knowledge is part of X's state. Likewise, if Y simulates X, then knowledge of X's state must be part of Y's state. The only way this can happen is if X = Y.

Perhaps you could write a simple program of two agents simulating each other and prove me wrong.

-- Matt Mahoney, matmahoney@yahoo.com



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