The Turing Test
I'm killing time before I head down to London tomorrow morning and I'm in a mood that can only be described as super-bored and frisky, so I thought I'd write about an idea that's been gestating in my head for a few months now.
Firstly a quick history lesson.
Alan Turing was a British mathematician and computer scientist who is famous for his pioneering work with computers, not to mention his ground breaking work during the second world war at Bletchley Park in the field of code-breaking. This is the man who broke the Enigma code, so you know he was a smart guy. He was also a practicing homosexual in an era when it was criminalized for which he was eventually prosecuted, but that's by the by.
Turing was fascinated by the idea of AI (artificial intelligence) and he postulated that if a human could converse normally with a computer, without ever realising that it was not a human they were conversing with, then that computer would have reached a state of true artificial intelligence. This idea is now known as the Turing Test.
A few months ago I was messing around on the website Omegle, which is like a one to one chat room that connects you with another random user. The idea is that you can strike up a conversation and hopefully learn something about someone else somewhere in the world. Well it got me to thinking about the Turing Test and whether I could really be sure that the people I was chatting to were really people and not just computer programs, when it struck me; all these people who have been trying to beat the Turing Test have been going around it in COMPLETELY THE WRONG WAY and I'm here to set them straight.
One thing we all do very well is converse, this means we can spot very easily when a conversation seems unnatural (in the same way, we can always recognise when a person in a film is CG since we are hard wired to recognise other human traits) so it is extremely difficult to write a program that can imitate a human conversation, which is what everyone is attempting to do to beat the test. My proposal is that instead of trying to converse with you, the program should hurl random comments, catchphrases and abuse at you. My guess is that you would instantly come to the conclusion that you were talking to a person. A dick-headed person. But a person nonetheless.
I don't know why scientists waste their time with any of that "curing cancer" or "Stem-cell research" crap when there are truly pressing questions like this going unanswered!
Firstly a quick history lesson.
Alan Turing was a British mathematician and computer scientist who is famous for his pioneering work with computers, not to mention his ground breaking work during the second world war at Bletchley Park in the field of code-breaking. This is the man who broke the Enigma code, so you know he was a smart guy. He was also a practicing homosexual in an era when it was criminalized for which he was eventually prosecuted, but that's by the by.
Turing was fascinated by the idea of AI (artificial intelligence) and he postulated that if a human could converse normally with a computer, without ever realising that it was not a human they were conversing with, then that computer would have reached a state of true artificial intelligence. This idea is now known as the Turing Test.
A few months ago I was messing around on the website Omegle, which is like a one to one chat room that connects you with another random user. The idea is that you can strike up a conversation and hopefully learn something about someone else somewhere in the world. Well it got me to thinking about the Turing Test and whether I could really be sure that the people I was chatting to were really people and not just computer programs, when it struck me; all these people who have been trying to beat the Turing Test have been going around it in COMPLETELY THE WRONG WAY and I'm here to set them straight.
One thing we all do very well is converse, this means we can spot very easily when a conversation seems unnatural (in the same way, we can always recognise when a person in a film is CG since we are hard wired to recognise other human traits) so it is extremely difficult to write a program that can imitate a human conversation, which is what everyone is attempting to do to beat the test. My proposal is that instead of trying to converse with you, the program should hurl random comments, catchphrases and abuse at you. My guess is that you would instantly come to the conclusion that you were talking to a person. A dick-headed person. But a person nonetheless.
I don't know why scientists waste their time with any of that "curing cancer" or "Stem-cell research" crap when there are truly pressing questions like this going unanswered!
Comments