On 9/6/06, YKY (Yan King Yin) <[EMAIL PROTECTED]> wrote:
On 9/5/06, Fredrik Heintz <[EMAIL PROTECTED]> wrote:
> [re learning]
> I do not see the point in only allowing a single learning paradigm. I
> believe people learn by different methods, so why shouldn't an AI do
> the same? Both supervised learning using induction (which in some
> sense both NN and ILP does) and unsupervised learning by trial and
> error, like reinforcement learning are useful methods. Some of the
> most successful learning methods have been deviced by combining RL and
> NN (see for example the work of M. Riedmiller et al
> http://www.ni.uos.de/).
I think even if NN+GA or NN+RL become very powerful, using them to evolve an
AGI would still be too slow.
And inductive approaches have problems with overfitting and thereby
lack of generality. They can find a pattern that very closely match
your examples, but if you give it a radically new example it will
utherly fail to generalize. Therefore the approach is more sensitive
to what test cases you choose, and what attributes you classify on. I
would like to see different approaches as complementary. Use each
method where it works well and do not try to squeeze everything in one
mold.
But maybe I miss your point.
My approach can be termed "declarativism", first proposed in John McCarthy's
paper "Programs with Common Sense" 1958, 1968, where the architecture itself
is static, and what is changing is the knowledge.
I have nothing against GOFAI and my point has nothing to do with NN/GA
being more inspired by biology (neither that you should consider it
more nor less interesting because it has analogies in biology). As a
matter of fact, I like the symbolic route and believe it is necessary
Also my architecture has yet to be integrated with acting/planning.
Which is a very tricky business :)
/Fredrik
-------
To unsubscribe, change your address, or temporarily deactivate your subscription,
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
