Expert: George Moustris - 7/16/2006

Question
How is "Artificial Intelligence" defined or determined? I've talked to a psychologist that feels we are far from understanding what our own human intelligence means. Or accurate ways in which it can be measured (academics not being so).

Answer
Hi James,

your psychologist is partly right, but this is not what AI is mainly about. AI can be divided into 2 main goals. First is to build conscious machines (strong AI) and second is to build intelligent ones. Intelligence cannot be strictly defined but a qualitative measure is rather common knowledge. You CAN see if a machine behaves intelligently or not without being able to specify exactly how intelligent it is.
Ultimately, intelligence in machines is doing complex things in a timely, organized and purposeful manner. Solving problems, learning from experience, producing knew knowledge, understanding context etc.
Human intelligence is just one paradigm we try to mimic. But  it is not the only one, and in fact it is the oldest one in AI but is rather deprecated. New paradigms are found in nature, in the way that nature solves problems. For example, the way that ants search for food (Ant colony optimization), the way that organic beings evolve (Evolutionary strategies, Genetic algorithms), the way that flocks move around (Particle swarm optimization) etc.
Other important paradigms are found in the way the brain works (Neural networks) and in the way humans think (Fuzzy logic). So, we don't have to know exactly what human intelligence means in order to mimic it.
On the other hand, conscious machines is what AI ultimately is all about. Here things get very tough. This is where we need to find out what consciouness means for humans in order to reproduce it. Current models are very simple and rather inadequate.

Best
G.