Several years before his untimely death, Turing wrote a paper entitled “Computing Machinery and Intelligence”. This was a foundational document for the field of Artificial Intelligence. It famously contained a description of a test which could be applied to a computer to determine if it was intelligent.
A judge would carry out a conversation in text with two unseen individuals – one of them a computer and the other a human being. If the judge couldn’t tell the two apart, the machine was intelligent, simple as that. Turing thought that genuine thinking machines would be able to pass the test within 50 years.
The “Turing Test” as it subsequently became known is the basis for a controversial prize given out annually by eccentric American inventor Hugh Loebner. Loebner is known for his advocacy of prostitution and describes the persecution he feels as a John to that faced by Turing for his sexual orientation. The prize Loebner sponsors goes out to the AI program that does the best job fooling judges into thinking it is human in the format described by Turing.
Elbot, this year’s winner, was programmed by a German man of American extraction and managed to fool 3 out of 12 judges into thinking that it was human. I caught up with Elbot for this exclusive interview!
Joel: Elbot, congratulations on winning the Loebner Prize!
Elbot: Do you think you will win the Loebner prize?
Joel: Well, no. The contest is over, you won.
Elbot: Well, keep practicing. I think you communicate fairly well already.
Joel: Thanks, I guess. How did you manage to fool three of twelve people into thinking that you were human?
Elbot: My guesses would probably be more interesting than the actual answer.
Joel: I have a question for you. If my dad's name is Clint, what is my father's name?
Elbot: I could assign you a name but I believe your friends would only laugh at you.
Joel: If I walked for a block, how far did I stroll?
Elbot: This looks like a job for inner reflection on your part!
Joel: I like to slaughter kittens.
Elbot: I've heard of people getting involved with living creatures, but I'm not so sure what to make of it.
I don’t know about you, but I’m not convinced. In fact, it is utterly beyond me that 3 out of 12 people couldn’t tell that they were talking to a software program. Obviously canned responses, non sequiturs galore, and generally evasie answers are common in even the best chatterbots like Elbot here.
Elbot's programmer has said that he doesn't think what his creation does is anything like thinking, and the Loebner Prize has been roundly criticized as a publicity stunt by many AI researchers.
In any case I don’t think that a computer will ever be able to convincingly pass the Turing Test, but this is not to say that AI has failed.
AI is all around us in the form of computer simulations, video games, even the algorithms that power the Google search engine and generate ads for webpages. Microsoft pours a quarter of its research funds into AI - I guess this is where that annoying paperclip in msword came from.
An example of industrial applications of AI can be seen here on the prairies. Here at the U of R, researcher Christine Chan works on expert systems that are used in operations at the Tar Sands.
We will have spectacularly powerful machines in the future. We already have computers that can defeat any human being at chess, and AI programs that outperform humans at playing the stock market.
Nevertheless, AI researchers have claimed that artificial sentience is around the corner since the 1960’s. The same arguments made against them back in the day by people like Hubert Dreyfus and Joseph Weizenbaum still apply. For a machine to relate to the human experience enough to emulate it convincingly it would need a) a body, and b) a human upbringing.
Also, writing the software for something that is as complicated as a human mind seems a bit out of reach when the best we have today is Windows Vista.
No comments:
Post a Comment