The latest Turing Test attempt: Try again
It seems that not a day goes by without someone announcing a new revolution based on some form of deep learning, or the announcement that some organization has invested lots of money and/or people in the field of artificial intelligence. Therefore, it was inevitable that we’d eventually hear that a program had passed the Turing test.
Reports of the chatbot Eugene Goostman were all over the news, and after reading the first article about it, I could immediately tell that we were not dealing with ‘intelligence’, but clever maneuvering:
• Only 33% of the judges took the computer for a human. Usually it takes at least a majority to pass a test of this kind.
• The test itself was only a five-minute conversation. Most anyone can pretend to be something they’re not for five minutes, and I think we’ve all been tricked into thinking someone was smarter than they actually were for five minutes (or more).
I’d be curious to know what the other 20 judges who tested the program thought. Why choose to simulate a boy of 13? That’s easy—as others have pointed out, this is also a sort of trick, as we wouldn’t expect a 13-year-old to know everything. In any case, if to trick the system is enough to pass, then someone could do the same from the point of view of a two year old. But I don’t think that’s the point—I least I hope it’s not.
Joking aside, I think it’s curious, not a revolution: Certainly different technologies are becoming more efficient and “smarter” (not to mention able to resolve an increasing number of complex problems effectively and usually inexpensively). But for something really clever, please, try again in a few decades.