RE: [Artificial Intelligence Group] Learning algorithm testing environments...
> >So the nature and design of test will have to depend on what you;)
> >intend to teach.
> Okay, so what do you do when the answer is "anything and everything"?
I would narrow "anything and everything" down to a smaller, more manageable
subset and start there. Once it can learn "anything and everything" within
the context of that subset you then expand the subset, and thereby expand
its learning capabilities. Do this enough and eventually it will be able to
learn "anything and everything" within the context of "the known universe".
The human brain, as I understand it, does not pop out into the world ready
to learn just any concept. You have to start small, build a foundation, and
then, once you have a framework to support yourself on, go on to learn
higher, more advanced, concepts.
[Non-text portions of this message have been removed]