 If computers are so smart, why do <em>we</em> have to understand <em>them</em>? Making machines more  <a href="http://www.cs.wisc.edu/~finton/int.html">intelligent</a> is the goal of Artificial Intelligence.  To me, the essence of intelligence is the ability to learn and adapt, to learn to act appropriately in order to reach our goals. Reinforcement learning treats this problem in the general case where  the system has outputs to control actions that can change its environment,  and it has inputs through which it senses its environment.  It also has an input  for reinforcement, which is a weak kind of feedback which can be expressed as a positive or negative number.  So, instead of having a teacher to present the system with input/output pairs, the system instead receives "thumbs up" or "thumbs down" at irregular intervals.    <P>
