From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Sun May 05 2002 - 21:56:35 MDT
Ben Goertzel wrote:
>
> One more thing.
>
> I am quite perplexed by your notion that concepts need to be *named* to be
> useful.
Actually, it currently looks to me like "words" have certain characteristic
qualities which "ideas" do not. That's why I use "concept" to denote a
word, rather than an idea.
> Don't you believe in content-addressable memory???
If by that you mean generalized association between similar things, sure, I
believe it. I also believe in words.
> This is one of the design principles that Novamente inherits from attractor
> neural nets (ANN's), actually.
>
> Semantic nets, which you dislike, involve addressing concepts by name.
Because semantic nets allow "address by name" and use it exclusively, it
does not follow that a real mind does not use address by name.
> ANN's involve addressing concepts by *specifying part of their contents* or
> *specifying a collection of related entities*.
I see... they can't handle indexical relationships between an auditory sound
and a concept, then?
> In Novamente (an integrative design), *some* concepts are addressable by
> name, but *all* concepts are addressable ANN-style (because concepts are
> represented by "maps", which are much like attractors in ANN's).
Words and phrases are addressable by name and association; unnamed ideas and
episodic memories are addressable only by association. Once an idea gets a
name or a phrase, it becomes possible to easily build up thoughts involving
that idea, which adds a great deal of complexity to the idea. It is also
possible and probable that ideas which are indexed by name are treated
differently by the brain than ideas which have no names. Introspectively, I
have often had the experience that giving an idea a name (or a phrase), or
hearing someone else do it, causes the idea to coalesce; this could be
triggered either by the ability to more efficiently create thoughts which
reference by the idea, or it could indicate the invocation of new cognitive
subsystems on that idea, or both.
-- -- -- -- --
Eliezer S. Yudkowsky http://intelligence.org/
Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT