RE: The problem of cognitive closure

From: Mitchell Porter (
Date: Fri Mar 16 2001 - 06:07:30 MST

--- Ben Goertzel <> wrote:

> The sources of ideas underlying the Webmind AI
> design are
> 1) introspection, as you describe
> 2) knowledge about the human brain
> 3) knowledge gained from practical computer science
> experience
> & algorithm theory

These correspond to the three ontological aspects
I mentioned: (in reverse) computation, physics,
and the noetic 'something else'.

> > Pragmatic consequence: Seed AIs need to be
> > philosophically sophisticated *by design*.
> >
> I don't think that philosophical sophistication can
> be
> embedded in an AI system by design. I think that
> the capability for
> philosophical sophistication is there in any highly
> intelligent
> system, and that the practical manifestation of this
> capability in
> an AI system has to be encouraged by the system's
> ~education~.

Well, when you get to the point of, say, discussing
whether or not one should be a Platonist, you may
have to do some retooling. I would be very surprised
if whatever representational system you're using
has the capacity to express even just all the *known*
philosophies. But if you're adding fundamental new
predicates to your AI's knowledge representation
language, that seems a little drastic to be called

And the main point of making 'philosophical
sophistication' a design principle up there with
Friendliness and modularity is to avoid the
brute-force imposition of incorrect 'solutions'
determined by locked-in philosophical errors.
(This is possible because problems of physics
and computation, to some extent, are separate
problem domains in which one could achieve
superintelligence while being oblivious to
anything else.)

The lesser point is to avoid physically harmless
philosophical errors, especially to Hard Problems
and Ultimate Questions and so on. The lesser point
is what warrants this being something other than
just a subgoal of Friendliness.

Do You Yahoo!?
Get email at your own domain with Yahoo! Mail.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT