Re: Proof by construction, again (was Re: [sl4] prove your source code)

From: Matt Mahoney (matmahoney@yahoo.com)
Date: Sun Jul 20 2008 - 11:02:18 MDT


--- On Sun, 7/20/08, Tim Freeman <tim@fungible.com> wrote:

> So you stand by your bold assertion that backing up computers is
> impossible. The backup media are defined to be part of the computer
> that's being backed up, so the loss of previous contents of the backup
> media is a failure of the backup. Odd.

No. I stand by my assertion that backing up a computer does not disprove that an agent cannot know its own source code. How do you know the backup is correct? You could make a second backup and compare the results, but that doesn't prove the process is correct. You could reload your computer from the backup, but then how do you know nothing is changed?

An agent making a copy of itself would have the same problem. It could not verify that the copy is exact. It has to trust the copy process.

> >More generally, two machines cannot know each other's states because
> >then each machine would have more information than the other (as
> >measured by Kolmogorov complexity).
>
> The state of a machine is a function of the machine and the
> observation time. Your statement is incoherent because
> you're omitting the observation times.

Observation time is irrelevant. I am arguing based on information theory. Each machine would have to have a copy of the other machine's state in addition to its own state.

-- Matt Mahoney, matmahoney@yahoo.com



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