X-Message-Number: 7961
Date:  Thu, 27 Mar 97 19:47:55 
From: Mike Perry <>
Subject: Reply to Thomas Donaldson

Thomas Donaldson (#7946) wrote:

>Finally, for Mike Perry: if you decide that everything is 
>symbolic I clearly can have no discussion with you other 
>than to ask you what you mean by "symbolic". And so I 
>will do that now. Please explain how you came to that 
>conclusion and what you mean by "symbolic".
>
Maybe "digital" would have been a better word. Processes 
are digital, in my view. Finite processes (bounded in space-
time and energy content) and objects have finite descrip-
tions--in symbols, hence "symbolic." Sorry for any confu-
sion. (Such processes should also be emulable in digital 
devices, i.e. computers--this follows not only because they 
have finite descriptions, but because the necessary transi-
tions from one state-description to the next are Turing 
computable, and in fact finite state machine computable, 
given the assumption of boundedness.)
>
>I would say, myself, that something is symbolic when we 
>treat it as a representation or symbol of something else.
>
To me, O and X are "symbols" but not necessarily represen-
tative or "symbolic" of something else. They are marks that 
can be formed into long strings to encode information. It 
might be tempting then to say that, if we did this, the infor-
mation is necessarily "without intrinsic meaning"--i.e. 
whatever meaning it would have must be "assigned" from 
the outside. To me, though, the information could be as-
signed simply by the structure of the language itself, i.e. the 
recurring patterns and regularities one sees. In my last 
posting (#7952), for example, I noted the string

OXXOXXXOXXXXXOXXXXXXXOXXXXXXXXXXX

which encodes the first few prime numbers: 2, 3, 5, 7, 11. 
So something "meaningful" is here, that should not require 
an explanation from the outside. 

To flesh out a scenario I have in mind for a self-contained 
language with intrinsic meaning, we might imagine a long 
"broadcast" or bit stream made up of just our two symbols, 
which occur over and over, but not in a totally random 
fashion. By the regularities that occur, an intelligent creature 
should be able to infer "meaning" without having to be 
instructed by anything beyond the bit stream itself. Mathe-
matical theories in particular could be encoded in an intelli-
gible form that should be understandable to a hypothetical 
alien creature who had no talents beyond knowledge of how 
to do mathematics and express the results in some sort of 
reasonable digital form (not necessarily the same as ours!). 
So the "meaning" would be an intrinsic property of the 
language, and not something that had to be added from the 
outside. Sure, the two bit-symbols all by themselves might 
be said to have "no intrinsic meaning"--but considered as a 
whole the language, and the expressions within it, would 
have such meaning.

In particular, one feature of the language could be to intro-
duce definitions, which is a powerful tool for expressing 
ideas in mathematics (and reasoning about them too). In this 
way we could have arbitrary strings of symbols *stand for* 
other things in the language, but the meaning of these 
"stand-in" strings would again be assigned entirely within 
the language, not from the outside. (Appropriate juxtaposi-
tions of patterns of symbols, reiterated occasionally, and 
other such coincidences, would convey the idea that one 
pattern is to stand for another one, say a shorter pattern for a 
much longer one that is used with some frequency.) Mathe-
matics would be a starting point for other ideas, maybe with 
computer science next in line, then physics. From there you 
could go to chemistry, biology, and ultimately even politics 
:-). (One complication might be that physics hasn't really 
been reduced to "first principles" yet; maybe string theory 
will do that, or we might have to "wing it" with what we 
have now, but with the quantum state it seems that physics 
could be reduced to appropriate digital form.) Perhaps very 
tough, but feasible in principle, I think, and maybe best for 
the hypothetical intelligence that really knows nothing about 
"where you're coming from" but is good at processing bit 
streams. (Maybe this approach could even be used to train a 
future computer!) 

Thomas also wrote (#7956) 
>
>The problem with symbols is that they have no 
>INTRINSIC meaning to anyone. Some other person must 
>be present to interpret the symbols produced by a 
>computer program. That interpretation, one way or 
>another, connects them somehow to something in the 
>world... but it is not the computer which does that, it is its 
>user.

Again, I think you could design a language in which the 
meanings of the expressions would be intelligible intrinsi-
cally to a thinking being, and not require "interpretation" 
from the outside.

>One problem with the Turing Test is that it plays on our 
>human tendency to react to symbols --- that is, the 
>language produced by the person or computer on the other 
>side --- quite automatically, as if they automatically meant 
>something to the object/person producing them.

Generally it is assumed, for the Turing Test, that you are 
conversing in English or another natural language. Such 
languages would make poor candidates for the sort of uni-
versal language I've imagined. We have too many words 
like "tree" and "politics" whose meaning could be plenty 
hard to decipher if you don't already know. Instead I pro-
posed starting with what seems like a more intrinsically 
intelligible foundation--mathematics--and working from 
there. You could then make the case that expressions in the 
language would "automatically mean something." If a 
language could have such intrinsic meaning, as I think it 
could, it makes more credible the idea that a digital com-
puter could have "awareness."

Mike Perry

http://www.alcor.org

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=7961