The concept of the Turing test has been cropping up a lot on the news lately, due to the current film about Turing and a competition hosted at Bletchley Park. For some reason it sparked off a whole bunch of thoughts.

First of all a pedantic thought. The news keep taking about how scientists are trying to create a machine that has intelligence, but really it is just a program isn’t it? As far as I know, all, or most, entrants to these competitions are coded on machines that are just computers and not special hardware like a positronic brain. But that leads to the speculation that maybe we won’t acheive AI without developing some sort of new hardware architecture to run it on.

I do also wonder whether the test itself needs a bit of updating. It basically involves people communicating via a typed interface and deciding whether they are ‘talking’ to a computer or a person. The thing is that you can look at a lot of the comunication via typed comments on YouTube comments, or readers’ comments on the online newspapers and half of them don’t make sense. It must be possible to pass the Turing test with just pseudo-AI rather than real AI. In the same way that standard IQ tests do not measure intelligence but just one’s ability to do IQ tests, the Turing test could just measure a program’s ability to do the Turing test.

At the heart of it all is the suspicion that the people running these tests now are missing the point a bit. If you go through all the instances of AI in science fiction stories and films there is never any doubt that it is artificial. A dog, a chimpanzee or a dolphin is intelligent but isn’t going to pass a Turing test. Do people working on AI really just want to create a machine that is just as intelligent as humans, but no more, and with the same sort of intelligence? What if they create something that is differently intelligent?

It must be theoretically possible to create something that counts as intelligent but can’t pass the test because we don’t understand it? Imagine if an ant scientist created a human brain: it would be more intelligent than the ant and yet unable to communicate with it.

Most intriguing is the idea that somebody has already cracked it and there is an AI out there acting as an intelligent bot out on Twitter or Facebook. I think we must all follow at least one person on Twitter who we haven’t met. How do we know they are actually a person?