Turing Test Officially Passed.
Post Reply
 
Thread Rating:
  • 0 Votes - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
09-06-2014, 04:05 AM
RE: Turing Test Officially Passed.
(08-06-2014 06:46 PM)Free Thought Wrote:  Well we're fucked.

We just have to make sure we give everything intelligent really short power cables.
Find all posts by this user
Like Post Quote this message in a reply
[+] 2 users Like Elder Cunningham's post
09-06-2014, 04:19 AM
RE: Turing Test Officially Passed.
(09-06-2014 04:04 AM)EvolutionKills Wrote:  This is why I can't help but wonder if this approach to AI is fundamentally flawed. I think you'd have a better idea simulating the basic fundamental operating principles of a brain (whatever those may be), then scaling that process up. It's like recreating a flock of birds or a school of fish. You can painstakingly animate each and every of the hundreds of birds/fishes. Or you can design a simulation that operates on a few basic and local rules for the behavior of an individual bird/fish, and then just scale that up to flock/school size and see what happens. The key to AI may be the same as programming a herd simulation. The point is not to recreate the output (the parlor trick), but rather to recreated the fundamental rules governing the individual neurons and their signals, and to scale that simulation up a few trillion times.

And you are correct in thinking that. This has led to the distinction between GOFAI (Good Old Fashioned AI) or classical or symbolic AI; and new-AI or non-symbolic AI. This has also led to the field of Artificial Life.

The former used a top-down approach. People would try to recreate the effect using whatever means they could. The latter tries to create the underlying mechanisms from the bottom-up. The advantage is you get to understand the nature of intelligence and also of life. So instead of emulating a 13 year old boy, you would have robots embodied in an environment evolving their own language and actually 'understanding' it.

There is merit in what the what they did to pass the Turing test but not in terms of creating an artificial intelligence. The test itself was just a showcase. The techniques they employed could probably be employed in data mining so it is actually the field of statistics that has benefited from this endeavour.
Find all posts by this user
Like Post Quote this message in a reply
[+] 1 user Likes Mathilda's post
09-06-2014, 09:59 AM
RE: Turing Test Officially Passed.
(09-06-2014 04:19 AM)Mathilda Wrote:  
(09-06-2014 04:04 AM)EvolutionKills Wrote:  This is why I can't help but wonder if this approach to AI is fundamentally flawed. I think you'd have a better idea simulating the basic fundamental operating principles of a brain (whatever those may be), then scaling that process up. It's like recreating a flock of birds or a school of fish. You can painstakingly animate each and every of the hundreds of birds/fishes. Or you can design a simulation that operates on a few basic and local rules for the behavior of an individual bird/fish, and then just scale that up to flock/school size and see what happens. The key to AI may be the same as programming a herd simulation. The point is not to recreate the output (the parlor trick), but rather to recreated the fundamental rules governing the individual neurons and their signals, and to scale that simulation up a few trillion times.

And you are correct in thinking that. This has led to the distinction between GOFAI (Good Old Fashioned AI) or classical or symbolic AI; and new-AI or non-symbolic AI. This has also led to the field of Artificial Life.

The former used a top-down approach. People would try to recreate the effect using whatever means they could. The latter tries to create the underlying mechanisms from the bottom-up. The advantage is you get to understand the nature of intelligence and also of life. So instead of emulating a 13 year old boy, you would have robots embodied in an environment evolving their own language and actually 'understanding' it.

There is merit in what the what they did to pass the Turing test but not in terms of creating an artificial intelligence. The test itself was just a showcase. The techniques they employed could probably be employed in data mining so it is actually the field of statistics that has benefited from this endeavour.

[Image: il_fullxfull.279237093.jpg]

[Image: E3WvRwZ.gif]
Find all posts by this user
Like Post Quote this message in a reply
09-06-2014, 07:28 PM (This post was last modified: 09-06-2014 07:31 PM by GirlyMan.)
RE: Turing Test Officially Passed.
(09-06-2014 04:19 AM)Mathilda Wrote:  And you are correct in thinking that. This has led to the distinction between GOFAI (Good Old Fashioned AI) or classical or symbolic AI; and new-AI or non-symbolic AI. This has also led to the field of Artificial Life. ... There is merit in what the what they did to pass the Turing test but not in terms of creating an artificial intelligence. The test itself was just a showcase. The techniques they employed could probably be employed in data mining so it is actually the field of statistics that has benefited from this endeavour.

I think what I'm observing is that traditional AI or symbolic reasoning is not being supplanted so much by a bottom up approach mimicking the brain's low-level functions but rather by a superficial correspondence approach. Don't matter how you get there as long as it corresponds to a desired outcome. Statistics, and in particular machine learning is the new AI. Excellent example is machine translation. The "traditional" rule-based approaches which try to capture deep semantic structure are vastly inferior to statistical machine translation approaches which ignore semantics altogether and just look for the most likely superficial correspondence. Brain's wetware, silicon ain't. Still a little too soon to make hardware work the same way as wetware. Soon enough though. Soon enough.

#sigh
Find all posts by this user
Like Post Quote this message in a reply
10-06-2014, 01:57 AM
RE: Turing Test Officially Passed.
(09-06-2014 02:56 AM)Mathilda Wrote:  ... This is why religious zealots can be indistinguishable from computer programs. I wouldn't want to bet money a program hasn't yet been written to emulate a religious zealot and also passed the Turing test.

This is why I harbor a strong suspicion the user called Jeremy E. Walker is some religious organization's pet software project. Much of what that entity says sounds a shade too stilted to have come from a real person. And it is perhaps that "just not quite real" quality that attracts reaction from this group wildly out of proportion to reaction to any other poster. The Walker engine seems to be right at the uncanny valley of human simulacrum.

Whatever Walker is, bot or someone with authentic brain cells, it's definitely a troll. A VERY successful troll. Watching a Walker thread is like watching a Madoff convince a whole auditorium to write him checks. I just shake my head.
Find all posts by this user
Like Post Quote this message in a reply
10-06-2014, 02:24 AM
RE: Turing Test Officially Passed.
(09-06-2014 07:28 PM)GirlyMan Wrote:  I think what I'm observing is that traditional AI or symbolic reasoning is not being supplanted so much by a bottom up approach mimicking the brain's low-level functions but rather by a superficial correspondence approach. Don't matter how you get there as long as it corresponds to a desired outcome. Statistics, and in particular machine learning is the new AI. Excellent example is machine translation. The "traditional" rule-based approaches which try to capture deep semantic structure are vastly inferior to statistical machine translation approaches which ignore semantics altogether and just look for the most likely superficial correspondence.

I see what you're saying. I've only ever been asked to use classical symbolic A.I. in industry, or statistics. When I've had to use techniques from non-symbolic AI, such as a typical artificial neural network it has just been another statistical tool to add to the collection.

The embodied, bottom-up approach to A.I. solves an entirely different solution that is not currently being tackled by industry. No one ever asks for biologically plausible local learning neural networks modelled on the brain and not many people know how to use them for practical purposes. Unfortunately there are still some serious limitations that haven't yet been cracked. We can't seem to get past stimulus - response when it comes to action selection using a non-symbolic agent controller.

I think we're on the verge of a break-through with this field but there's not much funding going on in academia at the moment. I have no doubt though when we do solve these problems there will be a revolution in robotics, especially considering that companies like Intel are currently focusing on developing lower powered chips that can be put in tablets and other small devices.
Find all posts by this user
Like Post Quote this message in a reply
10-06-2014, 02:37 AM
RE: Turing Test Officially Passed.
(10-06-2014 01:57 AM)Airportkid Wrote:  This is why I harbor a strong suspicion the user called Jeremy E. Walker is some religious organization's pet software project. Much of what that entity says sounds a shade too stilted to have come from a real person. And it is perhaps that "just not quite real" quality that attracts reaction from this group wildly out of proportion to reaction to any other poster. The Walker engine seems to be right at the uncanny valley of human simulacrum.


http://www.thethinkingatheist.com/forum/...#pid573113

I once called out a religious zealot on another forum as being a bot. I wasn't entirely convinced at first and started it as a joke but as time went on the evidence started stacking up. Each time I tried something even easier for a human to answer thinking that the joke would come to an end but it never did. Eventually the user got banned.

I remember asking for days "What is the capital of Russia?". I presented pictures with the answer embedded in them and all the user had to do was just once say "Moscow". It never did.

You can find the thread here:

http://www.housepricecrash.co.uk/forum/i...291&page=1

Of course now if you put that question into google it will immediately tell you.
Find all posts by this user
Like Post Quote this message in a reply
[+] 1 user Likes Mathilda's post
10-06-2014, 03:27 AM
RE: Turing Test Officially Passed.
(10-06-2014 02:37 AM)Mathilda Wrote:  
(10-06-2014 01:57 AM)Airportkid Wrote:  This is why I harbor a strong suspicion the user called Jeremy E. Walker is some religious organization's pet software project. Much of what that entity says sounds a shade too stilted to have come from a real person. And it is perhaps that "just not quite real" quality that attracts reaction from this group wildly out of proportion to reaction to any other poster. The Walker engine seems to be right at the uncanny valley of human simulacrum.


http://www.thethinkingatheist.com/forum/...#pid573113

I once called out a religious zealot on another forum as being a bot. I wasn't entirely convinced at first and started it as a joke but as time went on the evidence started stacking up. Each time I tried something even easier for a human to answer thinking that the joke would come to an end but it never did. Eventually the user got banned.

I remember asking for days "What is the capital of Russia?". I presented pictures with the answer embedded in them and all the user had to do was just once say "Moscow". It never did.

You can find the thread here:

http://www.housepricecrash.co.uk/forum/i...291&page=1

Of course now if you put that question into google it will immediately tell you.

"I think it's possible that Scarlets is the first human to ever fail the Turing test."

Fucking Laugh out load

And add Walker to that list. Laughat

[Image: E3WvRwZ.gif]
Find all posts by this user
Like Post Quote this message in a reply
10-06-2014, 05:29 AM
RE: Turing Test Officially Passed.
What huge pile of horse shit this story is. First of all, EVERYTHING IS FALSE! There is not a single true thing in the story, yet all the world media have written about it.

First things first, Kevin Warwick. Google a bit about the guy who claimed that he was a first cyborg (2000.), because he had/has a chip in his hand. Or that he has evidence about a first man that is infected with a computer virus (2010.). Now he is saying that he has made something like AI and that he can trick real people. No, no he can't and he didn't pass the Turing's test. Why? Because he changed the rules of Turing's test. Google about the test a bit as well...

Then we come to the link where the story originated. University of Reading.

Enough said.

P.S.

Eugene Goostman is just a chatbot, and not a particularly good one, because 3 years ago chatbot named Cleverbot scored 59% on the Turing's test, while Eugene Goostman had only 33%. but again, it was with the manipulation of facts, careful selection of "judges" and it is all wrapped up with a nice interpretation of someone who wants to sell their product.

TURING'S TEST IS NOT OFFICIALLY PASSED.

Tongue

[Image: a6505fe8.jpg]
I have a theory that the truth is never told during the nine-to-five hours.
-Hunter S. Thompson
Find all posts by this user
Like Post Quote this message in a reply
[+] 1 user Likes Filox's post
10-06-2014, 05:51 AM (This post was last modified: 10-06-2014 05:56 AM by Mathilda.)
RE: Turing Test Officially Passed.
(10-06-2014 05:29 AM)Filox Wrote:  First things first, Kevin Warwick. Google a bit about the guy who claimed that he was a first cyborg (2000.), because he had/has a chip in his hand. Or that he has evidence about a first man that is infected with a computer virus (2010.). Now he is saying that he has made something like AI and that he can trick real people. No, no he can't and he didn't pass the Turing's test. Why? Because he changed the rules of Turing's test. Google about the test a bit as well...

Then we come to the link where the story originated. University of Reading.

Enough said.

I always clench my fists and hiss whenever I see Kevin Warwick on screen. He is a media scientist who will come up with whatever he can to get a news article written about himself. He used to say that computers would surpass human intelligence by 2040 but he seems rather quiet about that now. He did the usual trick of equating a single neuron with a single byte and then assumed the Moore's law would continue indefinitely. There is so much wrong with that I would need another post to explain why.

I remember talking to my supervisor for my final year project back in 1997 about Kevin Warwick. He mentioned how he looked down Warwick's list of references in a paper and saw that he had referenced an article about himself from the Sun newspaper! (A very low-brow tabloid in the UK that has a picture of a topless woman every day on page 3)

He also told me about how they were at exhibition presenting their robots. Someone from a newspaper first went off to Kevin Warwick who then recommended that they go off to see my supervisor instead.

It's a real shame because Reading university is the only university that has (had?) a department in cybernetics which I think is a field that needs to make a come-back. But all the department is known for is a single media scientist.

The programme itself was written by a team in Russia apparently. I don't know anything about them.

(10-06-2014 05:29 AM)Filox Wrote:  Eugene Goostman is just a chatbot, and not a particularly good one, because 3 years ago chatbot named Cleverbot scored 59% on the Turing's test, while Eugene Goostman had only 33%. but again, it was with the manipulation of facts, careful selection of "judges" and it is all wrapped up with a nice interpretation of someone who wants to sell their product.

TURING'S TEST IS NOT OFFICIALLY PASSED.

That sounds quite plausible to me.
Find all posts by this user
Like Post Quote this message in a reply
Post Reply
Forum Jump: