Turing test result

xilman · 2014-06-09, 12:02

According to http://www.telegraph.co.uk/technolog...-is-human.html a program has managed to fool a third of the human judges that it was a 13-year old boy.

There's a bit of journalistic hyperbole, IMO, in that I'd want a success rate greater than 1/3 but it's progress nonetheless.

retina · 2014-06-09, 13:21

Quote:

Originally Posted by xilman

There's a bit of journalistic hyperbole, IMO ...

Your assessment is somewhat hypobolic.

xilman · 2014-06-09, 14:22

Quote:

Originally Posted by retina

Your assessment is somewhat hypobolic.

I dispute that claim. Hypobole requires that several alternatives be provided whereas I made only a single simple claim.

I believe I employed "meiosis", perhaps best defined as "rhetorical understatement". It is a subset of irony with a distinctively Socratic flavour.

Nonetheless, I enjoyed your cute pun and faux etymology.

Brian-E · 2014-06-09, 16:02

According to the author Alex G. Bell. there was an instance of a machine running the ELIZA program which passed the Turing test way back in the 1960s in a "conversation" which apparently lasted 10 minutes. Admittedly, this program could never pass the test nowadays, and its success then was presumably attributable to the relative unfamiliarity with the capabilities of computer programs back in those days.

I posted it here once before:

http://www.mersenneforum.org/showpost.php?p=174326

Quote:

Alex G. Bell, in his 1978 book "The Machine Plays Chess?", recalls an amusing account of one of the very earliest programs (ELIZA) which could apparently carry on a conversation in English even though it had no idea what it was "talking" about. In this incident Eliza "passed the Turing test with flying colours".

Apparently a student went into the AI office of his technology institute in 1968 to use one of the online teletypes as they were known in those days. There was one apparently free but the terminal had a notice taped on it reading "computer time reserved by Dan Bobrow".

The student sat down and tried to link to Bobrow's home terminal. Thinking he had succeeded, but having in fact linked to Eliza, he had the following conversation:

Quote:

Student: May I use your computer time?

Answer: Do you want to use my computer time?

Student: Yes.

Answer: I understand.

Student: Are you coming in tonight?

Answer: Why do you want to know if I am coming in tonight?

Student: Because I want to use the computer!!!

Answer: I see.

Student: Well?

Answer: What do you mean, "Well"?

Student: Well, what's your answer?

Answer: Let's explore further why you want to use the computer.

Bell recounts that this conversation continued for a further 10 minutes with increasing frustration on the student's part, after which he rushed to a telephone and dialled Bobrow's home number, rousing him from sleep.

Quote:

Student: What the heck do you think you're doing?

Bobrow: What do you mean, what the heck do I think I'm doing?

Batalov · 2014-06-09, 19:20

Turing's paper specifies no such arbitrary limit as "30%". Also, writing a code that pretends to be a Cartman (or a Beavis/Butt-Head) is too easy. Which is in a way what the program's authors did (together with manipulistic play on the Ukrainian/Russian current relations this rather stinks. "Eugene Goostman, a poor Jewish boy from Odessa", - oh, please, can we get some more stereotypes?* I bet that they trained their initial AI to pretend to be Borat.)

I suspect that the Turing's question was: "can we distinguish a clever machine from a clever human", rather than "can we distinguish a machine that pretends to be a moron from a moron?" I am surprised that they were able to "defeat the second question" with only 33% accuracy.

______________________
*Not to say that that stereotypes are true. I know people from Odessa, and among them, some poor Jewish boys, too (or grown-up boys ;-). The stereotype is of course a well-chosen one: it has a life of its own.

ewmayer · 2014-06-09, 21:29

Quote:

Originally Posted by xilman

According to http://www.telegraph.co.uk/technolog...-is-human.html a program has managed to fool a third of the human judges that it was a 13-year old boy.

Yes, but does it come in a blow-up-doll model?

R.D. Silverman · 2014-06-09, 23:01

Quote:

Originally Posted by ewmayer

Yes, but does it come in a blow-up-doll model?

Try

http://shopoza.com/

LaurV · 2014-06-10, 03:04

Quote:

Originally Posted by Batalov

I suspect that the Turing's question was: "can we distinguish a clever machine from a clever human", rather than "can we distinguish a machine that pretends to be a moron from a moron?"

Wonderful! I want to use this as a motto :D

LaurV · 2014-06-10, 03:07

Quote:

Originally Posted by R.D. Silverman

Try

http://shopoza.com/

Sh!t, you should say that is NSFW!

OTOH, I am surprised to see what kinda links our RDS knew...

kladner · 2014-06-10, 05:34

Quote:

Originally Posted by LaurV

Sh!t, you should say that is NSFW!

OTOH, I am surprised to see what kinda links our RDS knew...

I'm not surprised, even though this link only demonstrates awareness of the site. Why should you suppose that RDS is locked in the proverbial ivory tower, and is utterly naive about the big bad world outside?

retina · 2014-06-10, 05:48

Quote:

Originally Posted by LaurV

OTOH, I am surprised to see what kinda links our RDS knew...

In a world where Google exists it is not hard to find a link to almost anything.

2014-06-09, 12:02	#1
xilman Bamboozled! "𒉺𒌌𒇷𒆷𒀭" May 2003 Down not across 2⁵·331 Posts	Turing test result According to http://www.telegraph.co.uk/technolog...-is-human.html a program has managed to fool a third of the human judges that it was a 13-year old boy. There's a bit of journalistic hyperbole, IMO, in that I'd want a success rate greater than 1/3 but it's progress nonetheless.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Modifying the Lucas Lehmer Primality Test into a fast test of nothing	Trilo	Miscellaneous Math	25	2018-03-11 23:20
Incorect result of DC test	Miszka	Information & Answers	22	2014-07-16 22:46
The cmd Turing test	cmd	cmd	113	2011-02-06 15:46
Simplest universal Turing machine	davieddy	Science & Technology	4	2007-11-26 04:59
Church-Turing Thesis	Prime Monster	Math	8	2005-07-12 05:35

2014-06-09, 19:20	#5
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 22164₈ Posts	Turing's paper specifies no such arbitrary limit as "30%". Also, writing a code that pretends to be a Cartman (or a Beavis/Butt-Head) is too easy. Which is in a way what the program's authors did (together with manipulistic play on the Ukrainian/Russian current relations this rather stinks. "Eugene Goostman, a poor Jewish boy from Odessa", - oh, please, can we get some more stereotypes?* I bet that they trained their initial AI to pretend to be Borat.) I suspect that the Turing's question was: "can we distinguish a clever machine from a clever human", rather than "can we distinguish a machine that pretends to be a moron from a moron?" I am surprised that they were able to "defeat the second question" with only 33% accuracy. ______________________ *Not to say that that stereotypes are true. I know people from Odessa, and among them, some poor Jewish boys, too (or grown-up boys ;-). The stereotype is of course a well-chosen one: it has a life of its own.