'Eugene Goostman' - the first program to pass the Turing test.

[Cookie Monster]

http://www.independent.co.uk/life-style/gadgets-and-tech/computer-becomes-first-to-pass-turing-test-in-artificial-intelligence-milestone-but-academics-warn-of-dangerous-future-9508370.html

 

 

A programme that convinced humans that it was a 13-year-old boy has become the first computer ever to pass the Turing Test. The test — which requires that computers are indistinguishable from humans 

 

 

​This is a major breakthrough in AI. Being able to convince humans a computer program is another human has massive implications for society.

 

By being able to pretend to be a 13 year old the police could potentially use this level of ai to help catch paedophiles; however when the complexity reaches the level of an adult the risks of online fraud become huge.

 

On the bright side game developers should be able to eventually exploit this to advance ai beyond its current rudimentary levels.

[Assassin]

Inb4 skynet jokes. 

[Faceman]

Man... this is some scary $%&^.

[Faceman]

How do we know that you're not a computer?

[Jobbe03]

Really... They make a super smart AI and this is how they make it look... 

[bbqsauce]

Still, being able to convince people it is not a computer isn't the same as having true intelligence?

To me it seems like an over elaborate program with a huge database of responses.

 

plus, it only convinces 30 percent of the judges, i'm no expert on the turning test but why is the passing rate so low?

[JAKEBAB]

How do we know it isnt just designed to beat the turing test and thats it like that episode of numb3rs? *its all a conspiracy*

[bbqsauce]

How do we know it isnt just designed to beat the turing test and thats it like that episode of numb3rs? *its all a conspiracy*

conspiracy sure but to me being able to only convince 30 percent of the judges still isn't all that. I mean they used the fact that it's designed to be a 13 year old kid to fool people.

 

again, i'm no turning test expert. Just my opinion. lol

[LePawel]

-nom-

they should have went with this instead:

[Askew]

How intelligent is this human though? There's plenty of morons that I could convince I was a computer if I wanted to.

[ManWithBeard1990]

This is true. Just being able to convince someone it isn't a pc is not the same as having any decision-making or problem-solving skills whatsoever. The implications for npcs are promising though.

[Mooshi]

Why 13 year old specifically?

 

inb4tocatchapredator

[YellowDragon]

conspiracy sure but to me being able to only convince 30 percent of the judges still isn't all that. I mean they used the fact that it's designed to be a 13 year old kid to fool people.

 

again, i'm no turning test expert. Just my opinion. lol

Just getting to that 30% took this long, sure the program fooled 30% which is impressive. However I think if you were having a conversation with a machine there are certain uncanny "flags" that people would notice. For instance perfect spelling/ grammar, also the vocabulary. Ass humans were are pretty good at being able to pick out stuff that resides in the uncanny valley.

[Sauron]

I don't think we should be worried (yet), this technology is just in early stages and remember that almost 70% of the judges spotted the fact that it was a computer. Besides, the programme used has a programmed character, so every reaction it can have is predictable. If it had the potential to become skynet we would know.

[Homicidium]

Really... They make a super smart AI and this is how they make it look... 

 

Stereotype much?

[darkspeedy26]

(somehow i cant remove this quote above)

 

 

 

 

It convinced 33 per cent of the judges

 

 

my best guess is that there were exactly 3 judges, and one of the 3 isn't as smart...

 

33% isn't a coincidence, its exactly a third

[Sauron]

my best guess is that there were exactly 3 judges, and one of the 3 isn't as smart...

 

33% isn't a coincidence, its exactly a third

could be, or maybe it's 2 on 6. I doubt they would make this kind of experiment with only 3 judges.

[mr moose]

How do we know it isnt just designed to beat the turing test and thats it like that episode of numb3rs? *its all a conspiracy*

The Turing test is a little ambiguous because there are two ways it can be administered, The model they used with unrestricted questioning (interrogator) it is very hard (read impossible) to program for this specific test.

 

http://www.academia.edu/474617/Turings_misunderstood_imitation_game_and_IBMs_Watson_success

 

How intelligent is this human though? There's plenty of morons that I could convince I was a computer if I wanted to.

 

with intelligence being the ability to use knowledge, I would say they only have minor hurdles left as they have the knowledge side all wrapped up.

 

I don't think we should be worried (yet), this technology is just in early stages and remember that almost 70% of the judges spotted the fact that it was a computer. Besides, the programme used has a programmed character, so every reaction it can have is predictable. If it had the potential to become skynet we would know.

 

True,  Watson almost passed Turing test quite a while ago although I believe they used the other method (results compared to a human responses) in the above article.

[mr moose]

(somehow i cant remove this quote above)

 

 

 

 

It convinced 33 per cent of the judges

 

 

my best guess is that there were exactly 3 judges, and one of the 3 isn't as smart...

 

33% isn't a coincidence, its exactly a third

 

The royal society wouldn't sully their reputation by giving a pass with such a small number of judges,   I'd say it would have been sufficient number for the task, it could have even been an odd number but that the media rounds down the results for ease of reading.

[Sharif]

coolz

[Murdoch]

Given the transcripts I read from the same program 2 years ago. They've either improved the algorithm exponentially, or the judges have been severely generous or retarded.

 

Not to mention convincing 'one third' of judges that a computer is a '13 year old child from a foreign country' is hardly a giant breakthrough. That accounts for 'lack of knowledge' and 'misunderstanding language' which if you're not going to test properly, what's the point of the test at all?

[Wauthar]

And so it begins...

[Osmium]

-snip-

 

The Turing test has been around for a pretty long time and to be honest I don't really think it holds a lot of validity(an opinion shared by many academics much much wiser than me). The way I see it is if an AI passed this test that was designed for a totally different purpose than passing the test then it would actually hold meaning in my eyes. But when a program is designed to excel in the test and not designed to really do much else it to me at least does not hold much weight. Turing said that a computer can be considered intelligent when it cannot be distinguished from a human but the Turing test is only a single test in a single circumstance while there are tons and tons of different interactions that the program designed to pass the Turing test cannot do and it would become quite obvious that the AI was not in fact human in those scenarios. So I view his own test somewhat flawed.

 

I am more prone to agree with Roger Penrose's outlook on the matter in all honesty. He believes that actual intelligence is non algorithmic and as such cannot be truly obtained by strictly algorithmic machines.

[techswede]

If they changed it to mimic a cod playing 13 year old they would probably be able to fool me.

Every response could just be "F*** You" and "I F*** your mother". 

 

This is cool though.

[Fred Castellum]

If they changed it to mimic a cod playing 13 year old they would probably be able to fool me.

Every response could just be "F*** You" and "I F*** your mother". 

 

This is cool though.

My thoughts exactly.