Several AI programs, from simple to very hard, following the same guidelines of general AI

[Richard White]

Hello,
My goal is to implement several programs, from dummy level to
difficult one, following the same requirements:
- Perceive a given universe
- Associate and record
- Rate recorded schemes
- Reuse successful schemes
- Highlight exceptions from previous records
- build concepts
- Act on the perceived universe.

From my point of view, all these elements are compulsory in order to
get a general intelligence. Maybe several elements are missing. How to
know it? Just test in a restricted universe.
The very beginning will be a kind of neural network with a few
neurons, exploring all the topics above.
The second step will prove efficiency around natural language learning
(both analysis and synthesis).
A further intermediate goal is teaching an AI the game of Go. This AI
will be directly connected to Kiseido Go Server on the Internet as a
bot player. There will be a total difference with present bot players:
This bot will begin at a dumb beginner level (30k) and will learn from
other players thanks to its linguistic capabilities on this restricted
domain. Commented games in natural language will take place in this
learning project. Given the complexity of this game, even a partial
result will be a fruitful test of general intelligence. Testing on a
restricted universe allow better understanding of involved mechanisms.
This is really huge. Any question? Are-you interested?