Artificial Intelligence

Any discussion of artificial intelligence, or A.I., must inevitably start with the question of what exactly intelligence is. Unfortunately, it is not an easy matter to decide. Intelligence is normally defined as the ability to recognize relationships and to build upon them. However, computers can often do that better than humans, yet they are not therefore considered more intelligent. Desires, goals, and preferences are also important, as is a sense of self- awareness, when we talk about what we mean by artificial intelligence. The scope and depth of a program's coverage of all of these attributes determine which of the two main sorts of artificial intelligence it belongs to.
Weak A.I. is the main type of artificial intelligence that exists today. Weak A.I. programs do not attempt to mimic human consciousness or encapsulate the full range of human mental activity. Instead, they attempt to perform one particular problem-solving task very well. The most obvious example of such a program is the chess-playing computer Deep Blue, which, in May of 1997, became the first computer to defeat a current world champion in a standard tournament match. Deep Blue is clearly more intelligent than humans when it comes to chess, but it just as clearly has no greater consciousness that would allow it to compete with us in any other area. Other examples of weak A.I. include computerized grammar checkers, e-mail spam filters, and Internet chat bots. Because these sorts of programs are limited to specific tasks, and because they have become so familiar to us, they are not often considered artificial intelligence programs by most members of the general public. Nevertheless, they all represent considerable advances in the A.I. field, and form the best examples of the progress computer scientists have made towards creating thinking machines.
Strong A.I. is the other type of artificial intelligence and is what most people think of when they hear the term. Strong A.I. refers to computers that have a wide range of general cognitive abilities, including consciousness or self-awareness. No strong A.I. programs actually exist today, but scientists continue to work on developing one that works. At present, there are two main approaches to the creation of strong A.I. The first involves attempting to build a computer that is modeled after the human brain. The main problem with this approach is that scientists do not yet have a complete understanding of the human brain, so that any models based on it must necessarily be flawed. In addition, the human brain is so complex that it is virtually impossible to create a computer model based on it with today's processing technology. The second approach involves trying to create a strong A.I. program based on building up existing computer programs. This approach has the advantage of allowing scientists to make progress on strong A.I. software without having to
first develop much mere powerful hardware, but also raises the interesting question of Whether or not they would even recognize success: a strong A.I. program that was not modeled after the human brain might not manifest its intelligence: ina manner noticeable to its programmers.
Scientists and philosophers have long debated exactly how a computer might prove to us that it had developed genuine intelligence, yet no solid consensus exists. Indeed, we often find it difficult to judge another human being's level of intelligence, so it is perhaps unsurprising that we find measuring a computer's simulation of that ability nearly impossible. ■

(A) One method for gaging the success of a strong A.I. program is called the Turing Test. ■
(B) First proposed in the 1950s, a Turing Test works by having a judge or series of judges engage in a written conversation with hidden test subjects, some of whom are human and some of whom are actually computers. ■
(C) The theory is that a computer that could be mistaken for a human being by another human being would have to be considered intelligent. ■
(D) While this test has certainly spurred programmers to create much more advanced programs, many doubt its efficacy. For a computer to pass the test, it must have broad, generalized knowledge, but human experts participating in the test have often been misidentified as computers for having too much knowledge of a particular topic. Moreover, even if a computer could talk about things in exactly the same way we do, it would still lack desires or goals, which, for many, is a key element of true A.I.