Re: tech: digital physics

Eliezer S. Yudkowsky (sentience@pobox.com)
Sun, 16 May 1999 21:39:14 -0500

hal@rain.org wrote:
>
> Eliezer S. Yudkowsky, <sentience@pobox.com>, writes:
> > One might say that Special Relativity says that "real" things occupy a
> > continuum between rules and data, while in Turing machines things are
> > separated into rules and data. Another of the many reasons to
> > best-guess that the physical Universe is noncomputable. And also a
> > rationale for Power-class technologies like Greg Bear's "descriptor
> > theory", or being able to change the laws of physics: I think we'll
> > find that there's no rigid distinction between substance, law, and
> > metalaw, and that all three can be changed. Well, now I'm getting
> > Eganic, so I better sign off.
>
> I wouldn't go this far. I view relativity as being a concrete and really
> rather prosaic theory, just one which does not map too well to our
> mental models. It is a problem with our minds, though, not a problem
> with reality. You have spacetime, you have matter, and you have rules
> for how they influence each other. Our conventional philosophical views
> of the nature of reality are compatible with relativity, just not our
> specific intuitions about how space and time behave.

Well, thanks for agreeing with me about everything else, but...

Actually, I would probably say much the same thing if I came out of the blue and made a statement like that. The real reason I think that "laws are stuff" is that back when I was trying to figure out how to define "instantiation" for conscious entities, the one thing I kept running into was the idea that you could define any sort of consciousness apart from the algorithms composing it - two identical algorithms, one of which was conscious, the other wasn't; consciousness as epiphenomenon. I hate this idea. Any theory in which this idea shows up is wrong.

I intuit that the same holds true of reality. A process which is "real" will differ materially, in results and output, from a process which is "not real". Now, you can have one Turing algorithm which is "real" and one that is "not real", and they'll both work exactly the same way; reality is an optional add-on, just a matter of instantiation. Thus I don't believe in the Turing formalism. It's no small wonder that the Turing formalism can't answer questions about instantiation, or about the First Cause, when it treats reality as an epiphenomenon.

Special Relativity keys into that, but only vaguely. I could be wrong - I probably am wrong - about the blur between Turing tick and Turing data representing the blur between the laws of physics and material. But it's still one of the major intuitions, perhaps *the* intuition I use to understand ontology: Anything real can be detected, tested for, and modified. That goes for the laws of physics and the presence or absence of "reality", just like it goes for anything else.

-- 
        sentience@pobox.com          Eliezer S. Yudkowsky
         http://pobox.com/~sentience/AI_design.temp.html
          http://pobox.com/~sentience/singul_arity.html
Disclaimer:  Unless otherwise specified, I'm not telling you
everything I think I know.