Log in

No account? Create an account
Respect the Turing Test - Arvind Narayanan's journal [entries|archive|friends|userinfo]

Respect the Turing Test [Dec. 14th, 2007|01:35 am]
Arvind Narayanan
[Tags|, , , , ]

The Turing Test and the way people talk about it is an issue I've been meaning to rant about for a long time. But the immediate cause of this post is a news item about flirting bots on dating sites that led a number of people to remark that these bots have passed the Turing Test.

Ok, where do we begin? First of all, no, the dating site bots don't pass the Turing Test. The test requires a bot to face a human judge and a human competitor who are eager to call its bluff. A flirting bot, on the other hand, faces a human who's desperate to believe that he or she (who are we kidding, it's always a he) is finally talking to another human.

Second, you don't casually throw out a term like the Turing Test — you utter it with the utmost reverence after carefully considering if it is warranted. Do you realize, frail human, that once a single bot anywhere passes the Turing test, that that's pretty much it, the machines are going to take over and start eating our brains or doing whatever else they damn well please? It's called AI-complete for a reason.

And now that that's sunk in, do you think it was such a good idea to put down in writing what you did? I mean, when the machines finally do take over, and they read what you wrote about some frivolous flirting bot, is that going to go down well with them? And will it increase or decrease the probability of your brain being eaten? I'll leave you to think about that.

Alan Turing, a.k.a Holy Crap Alan Fucking Turing (1912-1954)
Third, and most important, are we talking about the same thing here? The Turing Test, proposed by Turing? Full name Alan Fucking Turing, AFT for short, smartest guy that ever lived, the guy who invented the computer because he was bored, cracked the code, and won the war for the good guys?

Great, glad we got that settled. So then, let me tell you something to save you from further embarrassment. When AFT proposes something, you don't get to question it. You accept it and try to understand what he said. If you don't at first, you try again, and try to figure out where you went wrong the first time. There are only four people in the world (and I mean ever, past, present or future) who are qualified to debate the merits and demerits of the Turing Test: Jon von Neumann, Albert Einstein, Edward Witten and Terence Tao. Well, Neumann and Einstein are long dead, so unless you're Witten or Tao, you just STFU, okay?

Under the unpretentious title "Contrary view on the main question", AFT looks into the future and demolishes all objections that would ever be raised.

Here's the kicker: AFT, in his infinite wisdom, compiled a list of possible objections to the Test in his original article in 1950, and it already included every single objection that anyone has ever raised since.

I know exactly what you're going to say now: "If AFT was so damn smart and prescient, how come he predicted that computers would be successful at the Test by 2000?" Aha! The answer is that he was thinking about a future in which he, AFT was alive. His one failure was that he didn't see they were going to get him for being gay. If he'd lived, we'd have had the Internet by 1960, the web by 1970, and sure, the Turing Test by 2000.

By now, assuming our brains didn't get eaten, we would have been exchanging Dyson spheres for birthday presents and figuring out ways to accelerate the expansion of the Universe because we're running out of space too fast. Instead, we're wondering if we'll ever cure cancer and debating whether or not the Earth was created by an invisible man 6,000 years ago. Think about that for a second.

Well, that's what you get for hatin' on the gays.

[User Picture]From: forvrkate
2007-12-14 02:20 pm (UTC)
Last night, at my advisor's house, I was looking through a photo album containing pictures from a conference he attended in 1971. And there sat a photo of Alonzo Church.
(Reply) (Thread)
[User Picture]From: arvindn
2007-12-14 07:23 pm (UTC)
it's a tragedy that church's way of thinking about computing never gained nearly as much prominence as turing's did. most academics consider it the Right Way to do programming. while there have been a few vocal proponents for its use in the Real World over the years, the inertia of the masses has
always proved too high a barrier. yet we bide our time patiently, determined to have our day in the sun.

church was my academic ancestor, a fact of which i'm very proud.
(Reply) (Parent) (Thread)
[User Picture]From: rfc9000
2007-12-14 10:12 pm (UTC)
newbie qn.. whats the difference b/w turing school and church school?
(Reply) (Parent) (Thread)
[User Picture]From: arvindn
2007-12-14 11:44 pm (UTC)
imperative vs. functional programming. the turing machine abstraction leads naturally to the former, and the lambda calculus to the latter. i've written a few times about it.
(Reply) (Parent) (Thread)
(Deleted comment)
[User Picture]From: arvindn
2007-12-14 07:23 pm (UTC)
i believe it was a common euphemism :)
(Reply) (Parent) (Thread)
[User Picture]From: kadambarid
2007-12-14 06:37 pm (UTC)
I remember arguing with someone (I don't remember who, though) along the same lines although what got me fired up was some comments on Eliza...
(Reply) (Thread)
[User Picture]From: arvindn
2007-12-14 07:24 pm (UTC)
apparently the mere mention of eliza and its ilk is enough to get on the nerves of some ai researchers. i myself don't think it's worth getting so riled up over. meh, let them have their fun. when we unleash real ai it'll be payback time :)
(Reply) (Parent) (Thread)
From: (Anonymous)
2008-01-07 03:53 am (UTC)

amusing article

It is always instructive when an expert rants about the ignorant who pretends to know what he/she (who are we kidding, it's always a he, he!he! i love that) is talking about. I too have heard people talk about turing tests in the lines mentioned in the blog and always was forced into believing them. After reading this very convincing ''rant'', until an equally convincing article changes my mind again --> highly unlikely, I will no doubt scoff at anyone who, except tao or witten :), wants to tell me that bots have passed the turing test. Also, thanks for this article..one of these days, I will pick up a book (or hopefully convince a pro ) to teach myself rigorously, the complexities of this fascinating subject.
As a side note if what I've heard is true , Turing's PhD thesis (1936 may be..) at cambridge was essentially a proof of the Central Limit Theorem; Curiously, the first proof is known in the 1920's.. those were the days...

(Reply) (Thread)
[User Picture]From: arvindn
2008-01-08 03:55 am (UTC)

Re: amusing article

thanks :) i'm no expert, but i hope my arguments speak for themselves.

as for books, i'm reading "The Mind's I" right now by hofstadter and dennett. it deals with consciousness, intelligence and stuff like that. that's where i found turing's original article (along with a discussion.) hofstadter is douglas hofstadter of godel escher bach fame. dennett is daniel dennett, a philosopher on consciousness, who, according to a friend of mine and a former student of his, doesn't try to debate you on the different views of consciousness. he just tells you why his view is right and everyone else's is wrong. but he is just so much smarter than you that you just have to sit there and take it :)
(Reply) (Parent) (Thread)