Science

Man vs machine…on instant messenger?

In 1950, computer pioneer (and gay icon) Alan Turing came up with a test to establish whether a computer exhibits true intelligence. Known as the Turing Test, it involves a human judge chatting on instant messenger, with both an AI program and a bog standard human, and then attempting to distinguish which is which. If the judge fails to tell the two apart the computer wins the Turing Test and is granted the status of artificial intelligence, while the human control subject is condemned to a lifetime of shame and infamy of being that hapless soul with a personality duller than a toaster.

Okay, it’s a bit more complicated than that. To pass the Turing test a computer must be able to fool the human judges at least 30% of the time. No computer or software program has yet come close to achieving such a feat…until now! Meet Elbot, an Artificial Conversational Entity (ACE) created by Fred Roberts, a computer consultant based in Hamburg, Germany. Elrod was recently involved in an annual Turing shootout between ACEs, each hoping to win the Leobner prize for superior ‘chatterbots’, held at the University of Reading. Elrod not only beat the competition but he – or I suppose, ‘it’ for the cynics among you, beat them in style by fooling 25% (3 out of 12) of its judges and coming closer than anything before to meeting Turing’s threshold. I had the honour of personally speaking to Elrod about his recent success and he seemed to be taking it very modestly (for the opportunity to speak to Elrod see below). What’s my honest opinion of its human busting abilities? I was pleasantly surprised – whilst there was something definitely very fishy about Elrod’s tendency to wildly jump from one topic to another, I enjoyed some badly executed attempts at humour – all quite complex behaviour for what is essentially a pile of electric circuits. In all frankness I have to question the people skills of some of the judges or human control subjects involved in the Loebner prize. However, what Elbot did suggest is a computer program capable of winning the Turing test isn’t too far off over the horizon.

Of course, philosophers and computer scientists have been thrashing around the nebulous issue of whether the Turing Test really proves intelligence for decades (notably, Elbot’s creator doesn’t buy it). But there are some more immediate implications of the development of such technology on cyberspace, especially due to the widespread usage of internet chatrooms, instant messaging and social networking websites. I’ll let you mull over that one. Dying to meet Elbot? Go fire up your web browser and visit him at www.elbot.com .

By Nasef Khan

Categories
Science

Leave a Reply