An Introduction to Artificial Intelligence

Artificial Intelligence is, in our opinion, the most exciting area of computer science. Artificial Intelligence (or AI, for short) is the name given to any attempt to have computers gain attributes of the human mind.

Of course, this is a very vague statement, and much argument has happened over what exactly constitutes AI (mathematicians and scientists hate vague statements). There are essentially two schools of thought: Strong AI and Weak AI.

Weak AI philosophers believe that computers, as advanced as they may get, will only be able to seem intelligent. The responses of the computer may seem like intelligent actions, but the Weak AI theory insists that the computers are just mindlessly manipulating data to produce "intelligent" actions. Strong AI philosophers believe that computers someday can be as intelligent as humans.

Of course, the heart of the discussion is this: What is intelligence? Normally, we declare that humans are the standard for intelligence, but then, isn't human intelligence, at the very basic level, just a bunch of mindless chemical reactions? Both the Weak AI and the Strong AI have strong supporters, many of them quite fanatic.

Aside from the philosophy of the Strong AI and the Weak AI debate, there is also another, related, division in AI: Connectionism and Classicalism.

Classicalism is the AI that is found in Chess programs, weather diagnostics, and language processors. Classicalism, also known as the top-down approach, approaches AI from the standpoint that human minds are computational by nature, that is, we manipulate data one piece at a time (serially) through a built in circuitry in our brains. So, much of the classical approach to AI consists of things like minimax trees, preprogrammed databases, and prewritten code. Expert systems are another name for Classical AI.

Connectionism is the newer form of AI. The problem with classicalism, connectionists say, is that it is too unlike the human mind. The human mind can learn, expand, and change, but many of the Expert systems are too rigid and don't learn. Connectionism is the AI that the media really likes, which is why it contains many famous names like Neural Networks and Parallel Processing. Connectionism seems a step closer to the human mind, since it uses networks of nodes that seem like the human brain's network of neurons.

Connectionism, however, also has its flaws. Connectionism is many times inaccurate and slow, and currently connectionism has failed to reach higher level AI, such as language and some advanced logic, which humans seem to pick up easily in little time. Human intelligence isn't built from scratch, like the Connectionist systems often are. So, for those higher-level AI, Classicalism is by far the better suited. Connectionism, however, is quite successful at modeling lower level thinking like motor skills, face-recognition, and some vision.

In the end, both viewpoints are valid in both debates. Regardless of the viewpoint, however, the most important goal for AI is that it helps us understand the mechanisms of the human mind.