The annual Turing Test is held each year at the University of Reading in England. The test is designed to further the ability of programmers to create computers which use artificial intelligence, or AI. This year, there were five entries competing in the test. The goal is to create an “artificial conversational entity” which can carry on human-like conversations up to 5 minutes in length.
In effect, the computers attempt to fool human beings into believing they are having a conversation with another human, and simply a machine. A group of interrogators are selected to converse with the machines at a distance by text and judge their responses according to how appropriate and human their conversation is perceived.
This year, the winning entry was “Elbot,” the creation of scientist Fred Roberts. Elbot managed to fool 25% of the human interrogators at the test. While that may not sound like a lot, it actually represents a major breakthrough in the area of artificial intelligence. Developing a computer which can carry on a conversation with the human is not as easy as it sounds. There are an infinite number of variables in human conversation — and a conversation could change in any direction at any given time.
Further, there are complicated subjects to deal with: irony, sarcasm and other forms of humor are notoriously hard to define, much less program into a machine. So when you consider it in this light, it’s easy to see how convincing even 25% of the human judges that the computer was a real human being, is quite a feat indeed for any machine.
The Turing Test is attempting to raise the bar in the field of artificial intelligence, and even though computers aren’t yet smart enough to fool a majority of people, AI has reached a stage where it is clearly improving.
The test itself involves what is known as a parallel-paired comparison, and was designed by brilliant British mathematician Alan Turing back in 1950. Dr. Turing theorized that if a machine could carry on a text-based conversation with the human, without the human being aware it was communicating with a machine, then the computer in question with be said to be involved in “thought,” and must be attributed with a certain level of what humans refer to as “intelligence.”
This year’s winner, Elbot, is the highest functioning computer conversationalist in the history of the Turing Test, and has set the bar pretty high with a 25% success rate — a rate it will need to defend next year against a new field of challengers.
Related Website Directory Categories:
Congratulations to the scientist Fred Roberts.
Comment by Computers — October 22, 2008 @ 12:10 pm