Welcome!

Thanks for stopping by! If you like what you read, tell your friends! If you don't like what you read, tell your enemies! Either way, please post a comment, even if it's just to tell us how much we suck! (We're really needy!) You can even follow us @JasonBerner! Or don't! See if we care!







Tuesday, July 28, 2009

One Singular Sensation (Continued)--Plus, a Bit of Shatner

A few more thoughts on Turing tests. (Feel free to skip back to "One Singular Sensation," our post of 7/26/09, if you need a refresher.)

We reported that, as of now, no machine has passed a Turing test--that is, no machine has successfully "tricked" a human interlocutor into thinking he or she is interacting with another human being. Keen readers may object. "What about those customer-service helplines that have a recorded human voice responding to your vocal commands? I've sometimes thought I was talking to an actual human being."

True, but for how long? We admit to an occasional lapse in similar situations. We've found ourselves speaking cordially to the infallibly polite "person" on the other end of the line. But the illusion is quickly dispelled. We realize that we are talking to a recording--and we immediately drop any pretense of politeness. A machine truly passes the Turing test only if it conducts an entire conversation with someone who is then unable to discern the true nature of the other "speaker."

In a "classic" Turing test, the conversation is conducted through text. One would think that this would make it easier for the "machine." After all, on a screen there is no telltale "robotic" tone of voice. For all you know, the Solipsist could actually be a robot could actually be a robot could actually be a robot could actually be a robot could actually be a robot could actually be a robot-----

-----Sorry, crossed wires.

But seriously, one wonders why it's apparently so difficult for a computer to "trick" a human if all the human has to go by is words on a screen. We imagine that the Turing programmers have managed to overcome any basic glitches; we assume computers are not responding to questions like "How are you today?" with answers like "Cream cheese enchiladas!" So what's the problem?

(Digression: Mmmmm. . . . Cream-cheese enchiladas. EOD)

We suspect a clever computer programmer, trying to pass a Turing test, would program some intentional imperfections. A few well-chosen misspellings or typos would go a long way toward establishing "humanity." On "Star Trek: The Next Generation," one distinguishing feature of the android, Data's, speech was a complete lack of contractions. So, reasonably assuming that anyone sitting down to challenge a computer to a Turing test is a full-fledged Star Trek fanatic, we advise programmers to instruct their charges to use lots of contractions. And if you can convince a computer to make with the funny, that would help a lot, too.

Still, it's an interesting comment on the human condition that, as we head towards an ever-more automated and infallible future, it is our little quirks, flaws, and errors that will come to be seen as the mark of humanity.

***********************************
Apropos of nothing, we thought Sloppists might get a kick out of this. It's a clip of master thespian William Shatner doing a dramatic reading of Sarah Palin's "farewell speech." What's funny is Shatner. What's tragic is that he is reading Sarah Palin's actual words, VERBATIM! Enjoy.


No comments:

Post a Comment