The New Scientist last week gave us a breathtaking exclusive about intelligent chatbot programs called ChatNannies that could pose as children in a chatroom and detect any paedophiles trying to groom them.
However, his claims of chatbots so intelligent they appear to be human (effectively passing the Turing Test) are quite an achievement, and his claims that he has 100,000 nanies online and has helped police investigations with it look dangerously on the side of fantasy. Cameron Marlow managed to get an interview with the bot but the conversation showed the bot was apparently capable of making tricky moral decisions and understood the concepts of betrayal, friendship and loyalty, making it light years ahead of the current state of the art.
Now suspicion has fallen on the creator, Jim Wightman, including an excellent investigation in this blog into his Usenet postings that revealed he (allegedly) makes spurious claims to have written newsreading software, is a Holocaust denier, and has made death threats. Wightman’s own responses in the comments section of the blog post aren’t exactly measured either.
Though perhaps no indicator of his coding abilities, this weird stuff doesn’t help this guy’s case that he is genuine and onto something big. Certainly to me it looks like a big hoax (seeking attention rather than money, as he’s making no commercial pursuit here), though the Guardian’s ‘Bad Science‘ columnist Ben Goldacre is being given a demonstration, so he’s certainly ambitious. It’ll be interesting to see what happens on the product’s release date on April 2nd and whether his claims actually stand up in the real world.
(via As Above, and I thoroughly recommend reading the two blog entries I mentioned above for their detailed analysis)
Update: Ben Goldacre’s latest column discusses it, and Wightman’s reluctance to have it tested on a standalone, isolated machine, which further raises suspicion. Oh, and last week’s NTK provides a succinct summary of why it’s probably a load of rubbish.