Do you have adblock enabled?
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Buzzfeed)   Sorry everyone, a computer did not just pass the Turing test   (buzzfeed.com ) divider line 85
    More: Followup  
•       •       •

13352 clicks; posted to Main » on 09 Jun 2014 at 1:11 PM (1 year ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



85 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | » | Last | Show all
 
2014-06-09 12:33:37 PM  
img.fark.net

Yes it did.
 
2014-06-09 01:12:23 PM  
I've know plenty of people who could fail the Turing test.
 
2014-06-09 01:15:01 PM  

GameSprocket: I've know plenty of people who could fail the Turing test.


Just off the top of my head, there was a certain anti-evolution troll that we all knew and loved here who seemed suspect.
 
2014-06-09 01:15:23 PM  
Whats that? Something news reporters spouted wasn't accurate or completely truthful? Shocker
 
2014-06-09 01:16:52 PM  
I also think that impersonating a 13 boy is a kind of low bar, since they are basically just seething cauldrons of testosteronone-fueled anger and lust. Really not hard to simulate.
 
2014-06-09 01:19:46 PM  
Where is the Obvious tag?

You can teach a computer to say "how does that make you feel" but you can't make it care about how you feel.
 
2014-06-09 01:20:37 PM  
www.ritholtz.com
 
2014-06-09 01:20:54 PM  
Ah-ha!  That's exactly what Skynet WANTS you to think!  Wake up Sheeple!
 
2014-06-09 01:21:28 PM  

Mad_Radhu: I also think that impersonating a 13 boy is a kind of low bar, since they are basically just seething cauldrons of testosteronone-fueled anger and lust. Really not hard to simulate.


"Hey what's up? Want to see a movie?"

"Fark you, dad! I hate you!"

*checks Turing box*
 
2014-06-09 01:23:25 PM  

Mad_Radhu: I also think that impersonating a 13 boy is a kind of low bar, since they are basically just seething cauldrons of testosteronone-fueled anger and lust. Really not hard to simulate.


When I first heard the news, my initial reaction was that it was a scam.  Basing the AI on a 13-year old boy doesn't prove that a computer can think, it just proves a 13-year old doesn't, which we already knew.
 
2014-06-09 01:23:53 PM  
Wake me up when one passes a Voight-Kampff test.
 
2014-06-09 01:26:02 PM  

Mad_Radhu: I also think that impersonating a 13 boy is a kind of low bar, since they are basically just seething cauldrons of testosteronone-fueled anger and lust. Really not hard to simulate.


Well, yeah, especially when it's English-speaking judges appraising someone who is supposedly a native Ukrainian speaker.  The subtleties of communication that native speakers naturally have (idiomatic expressions, etc) don't have to be programmed in.  Although, at this point with Big Data, Watson can do a pretty good job on Jeopardy even when the clues can be assumed to have non-literal meanings much of the time.

All the same, the scientific method would say that a case-control experimental design would be the best way to demonstrate how good this AI is.  The same judges would pose the same questions to the AI and to a real, average 13 year old Ukrainian boy and then we'd see if they could guess correctly if it was man or machine.
 
2014-06-09 01:26:05 PM  

Mad_Radhu: I also think that impersonating a 13 boy is a kind of low bar, since they are basically just seething cauldrons of testosteronone-fueled anger and lust. Really not hard to simulate.


Otherwise the Politics tab comment tabs would pass the Turing test.

lisasimpsonlaughsnort.wav yeah like that'd happen.
 
2014-06-09 01:27:57 PM  
Jesus farking christ, why does this have to be a farking listic.... oh. Buzzfeed. Internet for morons. Carry on.
 
2014-06-09 01:34:20 PM  

goatan: FTFA "The supercomputer claim seems to have originated in a press release issued by the University of Reading."

What's the bet Kevin Warwick was behind that claim.


Wouldn't be the first time that Warwick made wild claims via press releases only to later have them found to be just so much bull shiat.
 
2014-06-09 01:34:38 PM  

gevmage: Ah-ha!  That's exactly what Skynet WANTS you to think!  Wake up Sheeple!


imgs.xkcd.com

I think the xkcd test would have a higher bar.
 
2014-06-09 01:34:42 PM  
Dragonflew:  Internet for morons.

Isn't that redundant?
 
2014-06-09 01:34:45 PM  
Computers can't pass the Turing test, people can only fail it.
 
2014-06-09 01:35:58 PM  
Oops, I should have investigated further.  It is a paired (man vs machine) experiment.

From the  University of Reading's press release:

       Simultaneous tests as specified by Alan Turing
-          Each judge was involved in five parallel tests - so 10 conversation
-          Total of 30 judges took part
-          In total 300 conversations - based on 150, five minute parallel paired tests.
-          In each five minutes a judge was communicating with both a human and a machine.
-          Each of the five machines took part in 30 tests.
-          To ensure accuracy of results, Test was independently adjudicated by Professor John Barnden, University of Birmingham, formerly head of British AI Society.

Ok, so I guess the point is that there is no official definition of what a Turing test pass is because the man himself merely proposed a thought experiment but didn't actually give criteria he felt was a pass (since he rightly anticipated that the rate of fooling people would improve with time and computing power).  Obviously there is an upper limit.  50%.  Flip a coin.  Unless we make computers so clever they're, to quote White Zombie, more human than human.

/Anyone done a power-law analysis on this?
 
2014-06-09 01:38:18 PM  

farkingismybusiness: Wake me up when one passes a Voight-Kampff test.


At least then a five-minute time limit would make sense, as reaction time is a factor.
 
2014-06-09 01:40:33 PM  
I was skeptical when I read that they had Eugene be a 13 year old boy because they knew that if he were to be an adult, there was no way they could pass the test. The stacked deck doesn't seem good enough.
 
2014-06-09 01:42:29 PM  

farkingismybusiness: Wake me up when one passes a Voight-Kampff test.


The problem with getting a computer to pass the Voight-Kampff test is, first you would need a programmer who could pass it.
 
2014-06-09 01:48:18 PM  
I'm glad they included #7, because I have said that since high school. And people act like I have committed a science heresy. Like I countermanded heliocentrism or something.

The turing test is a weak, vague and highly arbitrary "test". And for more reasons than stated in the article. My biggest problem with it is that it is ridiculously anthropomorphic by assuming something has to think exactly like a human to be considered intelligent.
 
2014-06-09 01:48:59 PM  
So we don't have to castrate the computer because its openly gay and time how long it takes to commit suicide? Or is there a different Turing test?
 
2014-06-09 01:51:23 PM  

GameSprocket: I've know plenty of people who could fail the Turing test.


This is actually the Turing test's biggest problem -- most judges have a harder time telling a person pretending to be a computer from a computer pretending to be a person.
 
2014-06-09 01:52:54 PM  

ThrobblefootSpectre: My biggest problem with it is that it is ridiculously anthropomorphic by assuming something has to think exactly like a human to be considered intelligent.


Untrue.  The applicant simply has to convince the observer that it is self-aware.
 
2014-06-09 01:53:31 PM  

ThrobblefootSpectre: The turing test is a weak, vague and highly arbitrary "test". And for more reasons than stated in the article. My biggest problem with it is that it is ridiculously anthropomorphic by assuming something has to think exactly like a human to be considered intelligent.


Fine.  Give us something better.
 
2014-06-09 01:53:55 PM  

Mad_Radhu: I also think that impersonating a 13 boy is a kind of low bar, since they are basically just seething cauldrons of testosteronone-fueled anger and lust. Really not hard to simulate.


Yeah, I'm pretty sure all they had to do was make a robot that screamed 'farkshiatDICK' in a girly voice on XBox live to pass.
 
2014-06-09 01:55:39 PM  

fluffy2097: Where is the Obvious tag?

You can teach a computer to say "how does that make you feel" but you can't make it care about how you feel.


I've met plenty of people like that. What are you getting at?
 
2014-06-09 01:58:48 PM  

uttertosh: I've met plenty of people like that. What are you getting at?


The people have a spark inside them that is capable of caring, even if they do not.

Computers are incapable of caring. empathy cannot be put into a database.
 
2014-06-09 02:03:45 PM  
But, but I was promised SkyNET...
 
2014-06-09 02:03:49 PM  

ciberido: farkingismybusiness: Wake me up when one passes a Voight-Kampff test.

At least then a five-minute time limit would make sense, as reaction time is a factor.


That, and the lack of lesbians.
 
2014-06-09 02:05:31 PM  
Kryten was a judge.  Holly conspicuously absent.
 
2014-06-09 02:08:29 PM  

fluffy2097: The people have a spark inside them that is capable of caring, even if they do not.


Psychopaths don't. I've met a fair few of them. (worked in sales)
 
2014-06-09 02:08:31 PM  

GameSprocket: I've know plenty of people who could fail the Turing test.



Why do you say they couldn't pass the Turing test?
>
 
2014-06-09 02:11:03 PM  

fluffy2097: uttertosh: I've met plenty of people like that. What are you getting at?

The people have a spark inside them that is capable of caring, even if they do not.

Computers are incapable of caring. empathy cannot be put into a database.


Not all people are capable of caring.  But they can *FAKE* it.

So why can't we teach a computer to fake it?
 
2014-06-09 02:11:46 PM  
I see no cause for alarm. Now if the bot were named "David" and carried a teddy bear, I'd be worried.

/Nowhere near obscure on fark.
 
2014-06-09 02:11:56 PM  
I have been an artificial neural network expert for more than 10 years now. Not bragging, but I can write a script that passes the Turing test any day by simulating the output of a 1 year old child.
 
2014-06-09 02:14:58 PM  
33% is not a passing grade.
 
2014-06-09 02:14:59 PM  

MindStalker: But, but I was promised SkyNET...


What's left out of the story is that it began to learn at a geometric rate.
 
2014-06-09 02:19:18 PM  
dittybopper:
So why can't we teach a computer to fake it?

We did. It was called Eliza. It was one of the first "AI" ever created, and it simulated a conversation with a therapist.

The professors were quite disturbed when they found that every person in the department was talking to Eliza, and talking about shiat they should be seeing a real therapist for.
 
2014-06-09 02:21:22 PM  

AloysiusSnuffleupagus: Oops, I should have investigated further.  It is a paired (man vs machine) experiment.

From the  University of Reading's press release:

       Simultaneous tests as specified by Alan Turing
-          Each judge was involved in five parallel tests - so 10 conversation
-          Total of 30 judges took part
-          In total 300 conversations - based on 150, five minute parallel paired tests.
-          In each five minutes a judge was communicating with both a human and a machine.
-          Each of the five machines took part in 30 tests.
-          To ensure accuracy of results, Test was independently adjudicated by Professor John Barnden, University of Birmingham, formerly head of British AI Society.

Ok, so I guess the point is that there is no official definition of what a Turing test pass is because the man himself merely proposed a thought experiment but didn't actually give criteria he felt was a pass (since he rightly anticipated that the rate of fooling people would improve with time and computing power).  Obviously there is an upper limit.  50%.  Flip a coin.  Unless we make computers so clever they're, to quote White Zombie, more human than human.

/Anyone done a power-law analysis on this?


I'll suggest a slight modification to your study design - why tell the judges that one of each pair is a computer and the other is a human? Just see how well they can figure out which ones are human without any such clues.

Assuming the ability of a human to pass as human is 90%, and that each judge has a tendency to judge one way or the other in such a way that their data tends to have a correlation of 0.1, you'd have roughly 80% power to detect the difference (with an alpha of 0.05) between computers passing as human 78% of the time or lower. In other words, if computers could pass as human > 78% of the time, your study design would not be able to tell the difference between them and humans.

/Or did you not mean statistical power?
 
2014-06-09 02:21:29 PM  

Pelvic Splanchnic Ganglion: Dragonflew:  Internet for morons.

Isn't that redundant?


Which reminds me of Tchaikovsky versus Pachelbel...

img.fark.net
 
2014-06-09 02:22:42 PM  

traylor: I have been an artificial neural network expert for more than 10 years now. Not bragging, but I can write a script that passes the Turing test any day by simulating the output of a 1 year old child.


You can make a computer poop and throw up?
 
2014-06-09 02:31:59 PM  

draypresct: traylor: I have been an artificial neural network expert for more than 10 years now. Not bragging, but I can write a script that passes the Turing test any day by simulating the output of a 1 year old child.

You can make a computer poop and throw up?


Have you been to the Politics tab?
 
2014-06-09 02:32:38 PM  

draypresct: You can make a computer poop?


You're apparently not familiar with Windows 8.
 
2014-06-09 02:33:02 PM  
So there's no need to start welcoming our robot overlords (yet).

The screaming banshee buzz that beast makes is enough to wake the dead.

http://www.youtube.com/watch?v=W1czBcnX1Ww&feature=kp
 
2014-06-09 02:49:48 PM  
Who are the judges? I would assume they're well qualified, but then I think about the college football playoff committee and how *that* got picked and wonder if that 1/3 that got fooled are former politicians that grew up before the internet and just need a fancy board to sit on.
 
2014-06-09 02:51:18 PM  

fluffy2097: uttertosh: I've met plenty of people like that. What are you getting at?

The people have a spark inside them that is capable of caring, even if they do not.

Computers are incapable of caring. empathy cannot be put into a database.


Any reason why not? Your brain is just a biological computer of sorts and it handles empathy just fine.
 
2014-06-09 02:54:22 PM  

Ned Stark: Any reason why not? Your brain is just a biological computer of sorts and it handles empathy just fine.


For the same reason why thermonuclear fusion is easy for the sun, but hard for humans to make power plants with.

Scale.
 
Displayed 50 of 85 comments

First | « | 1 | 2 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is archived, and closed to new comments.

Continue Farking
Submit a Link »
Advertisement
On Twitter






In Other Media


  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.

Report