Re: Identity

Eliezer S. Yudkowsky (
Fri, 20 Nov 1998 15:43:21 -0600

John Clark wrote:
> Eliezer S. Yudkowsky Wrote:
> >How would you define "identity"?
> I think we'd need to start with the word "identical". It two things can be
> exchanged and no change can be detected by the outside world then the two
> things are identical. An unproved and unprovable but nevertheless reasonable
> corollary is that it doesn't matter internally either.

I am not getting into this. We already did this discussion two years ago. Besides, Turing identity is totally irrelevant to the issue of whether a given Turing process should be considered a singleton or a community, and you know it.

No offense. But if there's any AI-literate being on the planet who doesn't have strong, unarguable opinions on Strong AI, I would like to meet him.

As for myself, of course, I have always held that the laws of physics are non-Turing-computable and there is no observer-independent answer to the general form of the question "Does process A instantiate computation B"? Likewise, you are a strong computationalist: Everything is Turing-computable and the phrase "non-Turing-computable" is synonymous with ghosties and ghoulies. The chance that either of us will change opinions without totally new arguments equals the probability that we will both become fanatic Calvinists.

Why bother?

--         Eliezer S. Yudkowsky

Disclaimer:  Unless otherwise specified, I'm not telling you
everything I think I know.