ArtificialIntelligence, also known 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 this has been achieved.

Strong ArtificialIntelligence research is an attempt to create computer-based intelligence that can truly reason and solve problems. Such an AI would sentient, or self-aware. Initial efforts in this area led to Prolog, SymbolicManipulation, NeuralNetworks and other techniques. However, no real progess has been made, and the easiest way to create an entity with human intelligence currently takes 9 months.

In contrast, a weak ArtificialIntelligence can't actually reason and solve problems, even if acts as if it were intelligent. Modest progress has been achieved in this area.

The scope of ArtificialIntelligence changes as computers get more powerful. Things that were assumed to require intelligence often turn out to be doable with heuristics, or even deterministically. As an example, machine vision (recognising objects, driving cars, etc) used to be thought to require intelligence. However, since image processing methods have advanced and current computers have orders of magnitude more processing power than just a few years ago, simplistic heuristics can be applied to this problem with fairly good success.

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.