Re: Defining symbolic AI, what it really is and what it's not?


Traveler wrote:
On 26 Apr 2006 12:55:24 -0700, "heiska mikko" <mikk2lnx@xxxxxxxxxxx>
wrote:

Traveler wrote:
On 25 Apr 2006 21:16:12 GMT, curt@xxxxxxxx (Curt Welch) wrote:

Yeah, the last thing you want to do is see symbols as something evil for AI
just because one approach to AI that has the word symbol in its name seems
to have stalled.

In another post Joachim Pimiskern seemed to define 'symbol' this way:

"A symbol is a partial state of a system
that stands for something else. "

A symbol is usually defined as using one pattern to represent another.

Is it "worse than evil" to have AI that uses "partial states of a
system
that stand for something else"? What is your definition for symbol and
symbolic AI?

Symbolic AI, also known as representationalism, is the idea that
intelligence is the manipulation of symbols. It is nonsense for the

brain does not contain symbols in it.

How can you know that?

Intelligence is 100% signal processing. A signal is just a temporal
marker that announces an unnamed or unlabeled event. All signals
(neural spikes) are alike. A receiving neuron does not care about the
meaning of a spike, its source or its destination.

That does not contradict symbols.
You could say similar things about computers.
Symbols are in higher level, in patterns of signals and memory
markings.

This is why
intelligence is non-symbolic. It takes inteligence to create a symbol,
not the other way around.

If so, then what other AGI's intelligence components can do apart from
symbol processing? What else those components do? What are the other
categories?

You haven't proved that symbol processing can't have the ability
to reliably achieve something very intelligent with modest time
and memory. I don't mean any particular kind of symbol processing
(like those that are in AI-textbooks), but symbol processing in
general, something from the abstract, mostly unseen to humans, class
of every possible variation of symbol processing AI.

I claim that certain kind of symbol processing is intelligence.

Maybe you are calling "crap" just one set of symbolic AI versions, e.g
a subset of all possible versions of symbolic AI? How do you know that
there can't exist any good versions? Don't you have to provide
a mathematical proof for that?

I call it crap because it's crap, all of it. Symbolic processing has
nothing to do with intelligence as we have come to understand it from
observing intelligent biological systems. Any claim to the contrary is
either ass kissing or cluelessness. That anybody in the AI business
would want to kiss the asses of the GOFAI crowd at this late date, is
pathological.

rhetoric, circular



Mikko Heiska 23:37

.