If you were trying to write a computer program as intelligent as a human
being, the Turing test would help you decide whether you had succeeded.
In this test, named after the English mathematician who invented it,
a human judge sends questions to two computer terminals. One terminal
is hooked up to another human being, the other to a computer program.
If the judge cannot decide from the answers which is the human being,
the computer program is said to be intelligent.
Although many people have tried, scientists are still a long way from
writing a computer program that can pass the Turing test. I have made
a few attempts myself, and in 1996 my program was judged good enough to
win an international competition.
I am a postgraduate student in the Centre for Intelligent Information
Processing Systems at the University of Western Australia. At the Centre,
we are trying to build computers that use the same techniques as human
beings to do complicated things such as seeing and hearing.
The aim of my PhD project is to develop a computer program that can learn
a language. We can measure how well the program has learned the language
by seeing whether it can guess the next word of a sentence as well as
a human being can.
For example, if you were given the sentence "The cat sat on the.....",
you could do a pretty good job of guessing the next word. A computer,
on the other hand, doesn't know what a cat is. It is very difficult
to get a computer to perform as well as a human being.
Working with computers and communication requires a lot of maths and
physics, but maths is much easier when you understand why you are using
it and how it applies to the real world.
|