Artificial intelligence – 25% of testers fooled
For the first time, one “chat bot” managed to fool 25% of its human conversationalists that they were chatting to another human. The annual Loebner prize ($3000, so they’re not in it for the money) was won by Elbot, created by Fred Roberts, this last Sunday. The “Turing Test” model sets the interrogator behind a wall, and has them converse with an unknown human-or-machine on the other side. A machine that fools 30% of people is deemed (by Turing) to have passed, so we’re almost there, you’d think.
This is what I got from a quick chat with Elbot:
E: Hello dear user and welcome to Elbot! Just wondering, could you enter a single word that describes your energy level?
Variable
E: Well, my biorythms are all mixed up. Maybe because I’m not biological.
Way to give the loebner prize away!
E: Do you think you will win the Loebner prize?
oh sure
E: Please say that in some other way. I am not able to process ironic statements.
And so on, but much more random as time went on. Not sure I’d want to meet the 25% that were fooled, but you can make your own mind up by chatting to Elbot right now: http://www.elbot.com/
