On Mon, 25 Jan 1999, Wei Dai wrote:
> It's not at all clear to me what counterfactuals have to do with
> consciousness. So what if counterfactuals have no meaning for strings?
And it's not at all clear to me that static strings, without
dynamics or decisions, could have anything to do with consciousness.
Stalemate. We may never agree on this one. I think more people
would support my position than yours, but that's not saying much.
> Ok, I didn't realize that you wanted to evolve the neural nets in a
> (simulated) natural environment. Presumably it would be the same
> environment as we and our ancestors faced. If you're right and this really
> is the shortest program to output the state of an intelligent being, then
> I have no objection to the possibility that my consciousness is actually
> produced by such a program.
No, that's not what I said. For the shortest program, the
simulated environment would no doubt be vastly simplified compared to what
we observe. But the basic elements of competition for survival and
the advantages of communication would be there. It will take many
generations to evolve intelligence, of course.
> Well it would take me a long time to write the program, but the result
> shouldn't be too long. (Of course you do know that finding the shortest
> program to output a string is a noncomputable problem?) Of course I don't
> have time to do it (unless someone wants to give me a research grant?) so
> I can't show it to you. But your own proposal is even less complete.
You know as well as I do that an approximate solution is good
enough for our purposes and much easier to find. To be fair, the length
of both the form you proposed (and which I had earlier rejected), and of
the competitors such as the neural nets in the simple environment, should
be estimated. Maybe a good (and neutral) computer scientist could give us
those estimates without too much work.
My proposal to find a definition of implementation of a
computation is much more conservative than your approach, since it also
applies to regular theories of physics, and is in the established
tradition of computationalism. I admit it requires more work but I don't
think it's hopeless; and if it is, that's just as important to know.
> > Structure labeled how? By connections and interaction strengths
> > among the neurons? For the AI, the pattern of spatial locations of the
> > momory elements would be much of the needed info. Not so the human.
>
> Yes, what's wrong with by connection and interaction strengths?
Just that it would take a big program. Of course the neurons must
also be numbered in some sensible way.
> > That's just one choice of model among the many possibilities. And
> > what about the computations that never terminate?
>
> What other models do you have in mind? As for the computations that don't
> halt, first of all we don't have to worry about the ones that don't
> repeat, since they contribute zero measure. For the ones that do repeat,
> we can count their contribution, but I think it would be negligible.
On the contrary. Some programs will just rewrite the same part of
the tape, and those we can neglect for now. But some will keep printing
out copies of the same string end on end, and the ones that halt will have
zero measure compared to those, since more copies means more measure.
This gives a slight advantage to short strings.
Here is a serious problem for you. The Turing machines start out
will all possible programs, that is, with all possible infinite strings on
the tapes. At any finite time, no matter how large [and therefore in the
limit], the fraction of the tape that even the repeat printers have
covered is equal to zero. So the printed strings have zero measure
compared to the infinity of random strings on the as yet untouched part of
the tape, even though the fraction of those that are conscious is small.
Check.
- - - - - - -
Jacques Mallah (jqm1584.domain.name.hidden)
Graduate Student / Many Worlder / Devil's Advocate
"I know what no one else knows" - 'Runaway Train', Soul Asylum
My URL:
http://pages.nyu.edu/~jqm1584/
Received on Wed Jan 27 1999 - 10:24:36 PST