From: Ben Goertzel (ben@goertzel.org)
Date: Wed Jan 01 2003 - 08:59:18 MST
> > 1) Automaton. Fixed learning about X,Y,Z...No training.
> > 2) Learner. Learns X,Y,Z.....
> > 3) Meta Learner. Learns to learn X,Y,Y....
> > 4) Meta-Meta Learner. [snip]
> >
>
> Nice Levels. 3) is the one I'm interested in, and targeting the
> curriculum for. 1) is like DBMS. 2) is like an expert system, 'narrow'
> AI. 4) can be bootstrapped from 3) and so we (unenhanced humans) will
> never build it.
Michael,
It's a peripheral point, but one that should be clarified: Colin's level 2
is NOT like an expert system, because typical expert systems do not learn
based on experience at all. They just do logical reasoning based on a fixed
set of rules, encoded by humans. It's more accurate to say that his level 2
is like conventional machine-learning AI (feedforward NN's, statistical
learning systems, GP, etc.). A simple form of learning-to-learn (Colin's
level 3) may also be present in machine learning algorithms, to the extent
that the algorithms may adapt their parameters over a series of learning
trials.
-- Ben
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:41 MDT