|From the Movie I,Robot|
The most well known test for AI is the Turing Test, originally described by Alan Turing in 1950 as a way of answering the question, "Can machine's think?" The basic idea is that a human
interrogator would ask questions to two players, one being a machine and the other being a human. The interrogator would then have to make the determination as to which player is the human and which is the machine. Turing proposed that a machine could be said to think if that the machine could imitate a human to the point where an interrogator could not reasonably distinguish it from a human based on its responses.
Each year the Loebner Prize competition is held in an attempt to find a machine that can "think" based on the Turing Test standard. To date, no machine has been able to yield results in this annual competition that are "indistinguishable" from a human. In other words, no machine is currently known to "think" based on this standard.
Another well known test of computer intelligence is how well they can play chess (the topic referred to in Ars Technica's article). Almost since the inception of the study of AI, chess was thought of as a great test of machine intelligence. The reasoning? Exhaustive search in chess is VERY computationally expensive. It's so expensive in fact that even for a computer to successfully compete in chess, it must have some level of intelligence to make decisions with imperfect information outside of search (although faster processing and increased parallelism does make more search possible - part of the point made in Ars Technica's article); conducting a search on every possible outcome is not a feasible solution.
And that really is the root of what intelligence is: the ability to use knowledge and understanding to solve problems without perfect information. Sometimes we call it intuition. Sometimes we call it experience. But whatever you call it, it's the reason why we can understand language even when someone speaks with an unfamiliar accent. It's also the reason why chess players can make good moves even when they don't know (or consider) every outcome.
Intelligence Reduces the Need for Search...
Allen Newell and Herbert A. Simon discussed this in Computer Science as Empirical Inquiry: Symbols and Search. They said that intelligence reduces the need for search. And when you think about it, it's true. How often do we perform searches of every possible scenario before making decisions in our lives? For most of us, the answer is rarely. Instead, we try to find solutions to daily problems by relating those problems back to similar experiences. Sometimes that relationship is strong and we are able to make good, informed decisions. Sometimes that relationship is weak and as a result we might be uncertain of our decision or we might seek out advice from another person who had a more closely related experience.