Passing the Turing Test: Killing Machines Now Indistinguishable From Humans

by Ann Finkbeiner


Poor Alan Turing proposed a test by which you’d know whether The Machines are thinking: converse with someone you can’t see and who might be a human or might be a machine, and you’ll always know which. Test after test, we always know; machines are inferior conversationalists. But recently from the IEEE, the Institute of Electrical and Electronics Engineers-an extremely large, ruthlessly intelligent, highly organized professional association-comes troubling news. Change the test from conversing to killing, and all hell breaks loose: machines are indistinguishable from humans.

Computer scientists tinkered with a video game, “Ultimate Tournament 2004,” in which controlled avatars must kill or be killed. Some avatars were controlled by humans, some by bots. The bots were programmed not to act like a human but to think like one. And then the gamers had to figure out who was trying to kill their avatars, other humans or bots. They couldn’t. A computer scientist, no doubt grinning like a maniac, said, “There’s only a slender gap between the humans and bots now.”

Granted that the IEEE is less interested in killer bots than in artificial intelligence, and in particular-I think, but don’t take my word for it-whether AI should follow the rules of human cognition or of machine learning. Never mind. The information is out there now and can’t be recalled. You have been warned. Meanwhile, The Machines are known to be working on their conversational skills.

Ann Finkbeiner is a proprietor of The Last Word on Nothing, and is newly the author of A Grand and Bold Thing: An Extraordinary New Map of the Universe Ushering In A New Era of Discovery.