Penguin

ArtificialIntelligence, very often abbreviated as AI, is defined as making a machine behave in ways that would be called intelligent if a human were so behaving. The TuringTest is a famous criterion for whether a system is intelligent. Efforts in this area led to Prolog, SymbolicManipulation, NeuralNetworks, ExpertSystem?s, and other techniques, which were very important contributions to the field of ComputerScience. As such, even if we never succeed in creating ArtificialIntelligence, the research is giving us valuable results.

Outside ComputerScience the term ArtificialIntelligence is used almost exclusively as a marketing buzzword, meaning anything a computer does for the first time. The first Accounts Receivable system was called ArtificialIntelligence, as was the first Compiler. Many systems that just react to changes by following simple sets of rules are called "intelligent" in MarkeTroid lingo.

Computer scientists distinguish two goals in AI research.

  • A Strong ArtificialIntelligence would be able to truly reason and solve problems, and would be sentient, ie self-aware. No progress has been made in this department. After so many decades, the easiest way to create an entity with human intelligence takes 9 months. (Which maybe isn't so bad - it's more fun that way, too.)

  • In contrast, a Weak ArtificialIntelligence acts as if it were intelligent, but can't actually reason and solve problems. Such AIs employ heuristics and rule sets to make decisions, but act "intelligently" by trying to refine their rules and heuristics automatically based on their assessment of the success of their reactions. Modest progress has been achieved in this area. ExpertSystem?s are the most major achievement.

The scope of ArtificialIntelligence changes as computers get more powerful. Things that were assumed to require Strong ArtificialIntelligence turn out to be doable by a Weak ArtificialIntelligence, or even deterministically. An example is machine vision, ie recognising objects, driving cars, etc. As image processing methods have advanced, and with the explosion of processing power in current computers, simplistic heuristics have become viable and achieve fairly good success.