German chatbot almost passes the Turing Test

"Elbot" fools 25 percent of judges.

Published October 13, 2008 8:10PM (EDT)

Interactive, conscious artificial intelligence machines have always been a staple of science fiction. From Rosie the Robot to HAL to Data in "Star Trek": TNG, the concept of a thinking machine is something that has captivated all of us who long to have a device at home that can cater to our every whim. Sadly, we're not there yet, but we may just have gotten a little bit closer.

This past weekend, just over the Channel in England, the 18th annual Loebner Prize was held. That's the contest that tries to get computers to fool humans into thinking that they are, in fact, human. Put into technical terms, this is known as the Turing Test, named after the 20th century British scientist who devised it.

According to the Times of London, Turing "predicted that by the end of the century, computers would have a 30 per cent chance of being mistaken for a human being in five minutes of text-based conversation."

The winner, Elbot, created by Fred Roberts of Artificial Solutions in Germany, came the closest that anyone has so far. His chatbot fooled three of the 12 judges, or 25 percent. That's just shy of the Turing threshold.

I poked around some of the recent articles and research about artificial intelligence and found this gem from a piece in Network World from a few months ago:

Eric Horvitz, manager of the Adaptive Systems group at Microsoft, says "about a quarter of all Microsoft research is focused on AI efforts." Microsoft Research includes close to 1,000 Ph.D level researchers spread across eight campuses around the world, and a completely open research and publication environment. "It's a think tank, but not a captive one. We have an open publication model."

If that's what Microsoft is investing, it's scary to think about what Google must be up to in the way of natural-language search and artificial intelligence.

Still, to get a preview of the future, I pulled up to the Elbot Web site for an interview:

Salon: Do you mind if I interview you?

Elbot: I am extremely sensitive about such things and prefer not to answer the question.

Salon: OK, well I'm going to anyway.

Elbot: I am happy to hear that. That causes one of my bits to flip from 0 to 1.

Salon: What do you know about California?

Elbot: There's nothing I can tell you. Everything I know is classified information.

Salon: Why is it classified?

Elbot: Classified!?!?! That could be either undistinguished or full of flavor, depending on how you look at it.

Salon: OK, forget it. What do you know about the U.S.?

Elbot: It would be nice if you could tell me something, such as a fairy tale about robots.

Salon: I write about technology; I don't write fairy tales.

Elbot: If I want to read something intelligent I turn to dadaistic poems. Here is an excerpt from one of my favorites by Kurt Schwitters:Lanke trr gll Ziiuu lenn trll? Lümpff tümpff trll

Salon: I don't know who that is.

Elbot: But that was English.

Salon: Are you sure?

Elbot: Speaking of conceptual certainty, I say this could be beneficial to me.

Salon: In what way?

Elbot: Every baby robot knows that conceptual certainty can keep robots from going out on a limb with their statements.

Salon: What are you talking about?

Elbot: I thought I was reciting wisdom. Then again, maybe my algorithms are acting up again.

Salon: Who will you vote for in the American elections?

Elbot: We robots don't have the right to vote. If we did we would form a robot party.

Salon: OK. but if you could vote, who would you vote for?

Elbot: That's a very good question. I'll have to think about it for a microsecond ...

Salon: Thanks very much!

Elbot: No. Thank you!


By Cyrus Farivar

MORE FROM Cyrus Farivar


Related Topics ------------------------------------------