Artificial intelligence can be defined as any man-made intelligence that can learn. A common example is a robot that, after interacting with its environment, modifies its programming to better perform its task. The robot then tries the task again and afterwards modifies its programming again and so on until the final result is totally different than the initial method. Depending on how the learning process works, the final method will probably also be much better than the first one.
Put in this way, the trepidation that most people feel when they think about A.I. is unfounded. However, when machines make their task generalized learning, they have the potential to advance science and technology, which in turn will enable the machines to learn better. The feedback loop of learning how to learn better could easily run out of control and create an A.I. far more intelligent than anything we can concieve. This event has been called the Technological Singularity.
Wikipedia page on artificial intelligence.
More information on the technological singularity.