from Hacker News

I define intelligence as the ability of an agent to optimize its environment to suit its preferences in a variety of contexts. In the context of abstract environments this means dealing with constraint optimization and reinforcement learning. In partially observable environments means using perceptual learning techniques to make sense of the environment.

In order to create an intelligent agent I have constructed a two part system consisting of a system for dealing with optimization under complete information using techniques from order theory, operations research, and optimization theory as well as a system for optimization under incomplete information that uses the perceptual learning capabilities of OpenCV. You can read more about my work on my blog:

http://lisp-ai.blogspot.com/