Re: Failure of AI a prediction of Neal Stephenson's "The Diamond Age"

Eliezer S. Yudkowsky (
Wed, 16 Dec 1998 14:43:10 -0600

Anders Sandberg wrote:
> I think FTL travel is impossible (without shortcuts and tricks like
> wormholes).

Well, nobody (except the folks on Star Trek) was claiming that you could just get out and push. This is like saying the sky is blue unless it isn't.

> So? Suppose it really turns out that there are souls (of the kind
> people commonly believe in). Would that invalidate extropianism? Not
> the least, it would just mean we have to adjust our plans. Like for
> example study how to make AI with souls.

Yes, this is exactly the point I was trying to convey last year during the debate about Turing computability; even if the Atheist's Worst Case is true and there are noncomputable indestructible structures interfacing our neurons, this does not mean that these structures are outside physics, logic, or the scientific method. At absolute worst they might be intrinsically incomprehensible to our intelligence, but even so all the specifications for producing any effects involved would have to be encoded in a few megabytes of DNA. So there are no emotional/political/social reasons for preferring Strong AI or the Church-Turing thesis to their negations; they are simply open questions to be resolved by experiment.

I do happen to believe reality is noncomputable; but I don't think it's intrinsic to our reasoning methods (to consciousness, perhaps, but not to intuitions or emotions). In the event that I'm wrong about the latter, we would simply dissect neurons until we could duplicate the effect in silicon. None of this presents any sort of impassable obstacle.

--         Eliezer S. Yudkowsky

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