I’m always curious about attempts to challenge the Turing Test, in which machines convince humans they’re human through intelligent interaction. I remember using an ancient text-to-speech programme that attempted this – Dr Sbaitso – extremely badly in the early 90s.
I’d love a world in which there were useful simulacra of humans able to support humanity in its day-to-day dealings (Skynet, natch), but have never yet come across a useful implementation of this technology.
Of of my many blog subscriptions pointed me to Cleverbot.com, an “AI” simulator whose intelligence is predicated on human knowledge available on the Internet. The bot warns you that its insights are based on what people say and think around the Internet and therefore might potentially be offensive.
It took me three questions to get to a Chuck Norris reference. I reckon five questions in I’d be onto Nyan cats and Dogforts.
To the creators of Cleverbot: if you were going to pick a compendium of human knowledge, as wonderful as the Internet is, you might need to sling in some algorithms that limit the frequency with which people mention Chuck Norris, pirates and ninjas, hipster (and other) kitties, things-made-of-bacon-that-normally-aren’t, and a thousand other memetic ideas. Despite the fact that people talk about these things at extreme length with surprising regularity online, they don’t in real life and therefore Mr Turing and his test will remain undefeated.
Unless, of course, that’s the whole idea. Imitate and Internet Savant and people won’t know the difference. It could be inspired!