Grr… I only have to update my blog to get annoying spam comments, so I’ve just been deleting the latest lot. I’ve now turned on “image verification”, where you have to type in some slightly garbled text from an image to prove that you’re a human and not an evil spam-spreading piece of computer programming.

But what will happen if programs are developed sophisticated enough to do any such task thrown at them? What if artificial intelligence becomes

But it takes something very sophisticated, like a human brain, to think on a high enough level to be able to make mistakes, as opposed to just automatically following a set of instructions. Of course, if the instructions are flawed then you’ll get the wrong answer, but the program will do what it is supposed to.

So here’s how you could devise a new test to outwit a super-intelligent AI. Set a super-fiendish Su Doku puzzle as a test, which an automatic programme will easily be able to solve but any normal human will find impossible, and make making a mistake the criteria for testing humanness. Perhaps in future actual stupidity will be the best way of testing that something isn’t artificially intelligent.

Tags: Spam artificial intelligence