[Question] Re: TD & Intelligence

[Mousheng Xu <moushengxu@xxxxxxxxxxxxxxxx>]

Hi,

	Frankly speaking, I know almost nothing about TD nor Robert Jasiek's
f(situation). But my understanding is that the fundamental of the
difficulty of go programming comes from the game space. If you want to hack
on patterns, you have "uncounterable" patterns; if you want to apply
knowledge, you have "unlimited" knowledge to incoorperate into; if you want
to train your neurons, you need so many neurons.

	Now here are my two fundamental questions about ANY AI approach:

	1. How fast does your system learn? It's imaginable that a fish can evolve
to a monkey, but how many years does it need? You don't expect to live
forever, do you?
	2. How much information do you need to store? If you use NNW, how many
nodes/layers do you need? If you use Genetics Algorithm or Evoluationary
Strategy, how many "DNA molecules" do you need? You don't want to use the
whole universe as your hard drive, do you?

	If you don't have these answers in mind, hmm, think it again. :)

	Thanks for reading.

-- Mousheng Xu