The Lonely Planet guide to Artificial Intelligence (Part 1)

By: Shane Dempsey

Anybody with a passing interest in Science Fiction will have some idea what the terms Artificial Intelligence or AI means. However, even among AI experts this can mean different things to different people. Most contentious is the use of the word 'intelligence' as this is quite difficult to define. AI is informally categorised into two classes: strong AI and weak AI.

Strong AI assumes that computers can be made to think with the same or greater facility than humans. Weak AI simply states that computers can merely approximate a subset of less demanding human thinking.

Weak AI concerns making computers more useful but denying that they will ever think as we understand the term. Weak AI is everywhere from washing machines to speech recognition systems. The famous French philosopher, Descartes, once said I think therefore I am. This kind of reflexive chicken or egg definition nicely illustrates our difficulty in deciding what thought actually is. To provide some kind of objectivity in our assessment of whether a machine is really thinking Alan Turing, one of the founding fathers of computing and Artificial Intelligence, proposed the Turing test.

In the Turing test, a human judge is placed in one room, and the machine or another human is placed in another room. The judge may ask questions or answer questions posed by the computer or human in the other room. All communication is done through a terminal by typing.

The judge does not know if the subject that they are talking to is a human or a computer before the conversation begins. If the judge is talking to a computer and they are fooled into thinking that the machine is a human then the machine has passed the Turing test. The test is flawed however as the AI machine creator is forced to mimic human details such as the imperfections of human typing, grammar, syntax etc. Contrived spontaneity is another tool that has been employed to trick the judge. The test is also focused on Natural Language Processing (NLP) but the assumption that intelligence can only be expressed within the constraints of human language is contentious to say the least.

It's unclear how some of this chicanery advances AI so the Loebner Prize [1] was created in 1990 as an alternative test of AI. In recent years the prize has been won by Richard Wallace's chat both named Alice [2]. Indeed Instant Messaging IMBots, which pretend to be Instant Messaging buddies are currently being developed for fun and profit by academic and commercial organization capitalizing on widespread coverage of Instant Messaging software such as MSN Messenger and AOL IM. [3]
As you can imagine, Artificial Intelligence is a wide and fascinating research area. Future columns will discuss AI areas such as Neural Networks, Artificial Life, Genetic Algorithms and their real-world applications.

[1] Loebner Prize home page, http://www.loebner.net/Prizef/loebner-prize.html
[2] The A. L. I. C. E Artificial Intelligence Foundation, http://www.alicebot.org/
[3] Create AIM bots & AIM Expressions. http://www.runabot.org