Passing the Turing Test

[Walt34]

Up until now a computer has never passed the Turing Test. Now one has. Just barely, but this will undoubtedly become more common.

Can HAL be much farther behind?

Neat and sorta scary at the same time.

 

[easysurfer]

"Westworld" here we come! Remember that movie starring Yul Brenner and James Brolin?

Hmmm...I would not have thought that computer is a human

Video: John Humphrys grills the robot who passed the Turing test – and is not impressed - Telegraph

 

[Totoro]

We are getting surpassed fast in most aspects of intelligence. I think that's a good thing.

Finally getting closer to the point where a collection of machines can design and manufacture a smarter machine fully on its own.

Now consider the following two trends accelerating and converging

• Complete automation of all essential labor, machines are better at anything a human can do
• End of aging: replacing full organs grown from your own cells, rejuvenating stem cells

Nowhere near full reality in the short term, but still .. the world is going to look mighty different in 50 years.

Think of the impact on how our economy is run, population, status, social relations.

 

[easysurfer]

The biggest hurdle IMO is if ever we can create a machine that is a sentient being.

If that ever happens, then humanity and artificial intelligence really gets blurred.

Or how about hybrids? Maybe one day artificial parts will become so lifelike that over 50% of the body will be made (or grown). Then how human are we?

Good stuff to think about.

 

[MichaelB]

Anyone know where we can find John Connor?

 

[clifp]

Evidently the media bought a well orchestrated PR campaign.

The program isn't a real AI program like IBM's Watson (which is what I thought it was reading the press story.) It is a chatbot and not even the best scoring chatbot.

The inventor has a long history of making outlandish claims. More here at techdirt

 

[Walt34]

Dang, that was disappointing. Oh well, I guess we're safe. For a while. Liked the cartoon on that site.

 

[MichaelB]

Evidently the media bought a well orchestrated PR campaign.

The program isn't a real AI program like IBM's Watson (which is what I thought it was reading the press story.) It is a chatbot and not even the best scoring chatbot.

The inventor has a long history of making outlandish claims. More here at techdirt

Anyone know where we can find John Connor?

Gilda Radner Nevermind - YouTube

 

[ziggy29]

I would argue that that no computer could ever qualitatively, definitively and authoritatively pass the Turing Test because Alan Turing himself did not quantifiably define what it means to "think". It appears that some "experts" decided to define what it means to "think" and then tested it. They decided that "thinking" means fooling 30% of the people who were tested by it. But Turing never defined it that way.

So you can say this shows new boundaries being pushed in artificial intelligence -- and you probably should, for good AND for bad -- but it is not the "Turing Test" per se.

It's been a long time since I've been able to talk about theoretical computer science instead of corporate BS related to computing. It makes me happy.

 

[robnplunder]

My eye sight isn't what it used to be and I read the subject line as .... Passing the Urine test.

 

[Jack_Pine]

Just wait until Skynet really becomes self aware...........................

 

[Nemo2]

"Westworld" here we come! Remember that movie starring Yul Brenner and James Brolin?

I thought of The Moon Is a Harsh Mistress by Robert A. Heinlein.

 

[clifp]

I would argue that that no computer could ever qualitatively, definitively and authoritatively pass the Turing Test because Alan Turing himself did not quantifiably define what it means to "think". It appears that some "experts" decided to define what it means to "think" and then tested it. They decided that "thinking" means fooling 30% of the people who were tested by it. But Turing never defined it that way.

So you can say this shows new boundaries being pushed in artificial intelligence -- and you probably should, for good AND for bad -- but it is not the "Turing Test" per se.

It's been a long time since I've been able to talk about theoretical computer science instead of corporate BS related to computing. It makes me happy.

I was inspired to read the Wiki on the Turing Test. I remember spending a lot of time learning about Turing machine in my history of computing but not much about the Turing Test.

After reading the techdirt article and the wiki, I am pretty convinced that none of these chatterbox program uphold the spirit of the Turing test and the 30% number seems ridiculously low. Jeez 30% of the population believes in Alien abduction, Ukraine is in South America, Obama was born in Kenya, and in creationism etc.

On the other hand watching Watson kill the human opponents on Jeopardy really makes me think we have made tremendous breakthroughs in AI and Skynet may not be too far from a reality.

You are right the Turing test is not well defini

 

[ERD50]

Could Paris Hilton or Justin Beiber pass the Turing Test?

-ERD50

 

[Andre1969]

"Westworld" here we come! Remember that movie starring Yul Brenner and James Brolin?

Hmmm...I would not have thought that computer is a human

Ahh yes, welcome to Delos, where we can recreate a woman's, umm, ladyparts to perfection, but still have problems getting the hands done right.

Don't worry, nothing can go wrong. wrong. wrong. wrong...

Loved that movie, despite some of its flaws and inconsistencies.

 

[photoguy]

I wonder what percentage of people would fail the test?

 

[ls99]

Could Paris Hilton or Justin Beiber pass the Turing Test?

-ERD50

Sure. At the 8 to 10 year old level. As they are not yet hampered by reason.

 

[Which Roger]

I was inspired to read the Wiki on the Turing Test. I remember spending a lot of time learning about Turing machine in my history of computing but not much about the Turing Test.

It brought back memories of my combinatorics class in grad school where we did a set of problems to prove that all kinds of more-complex-seeming machines were no more powerful than a Turing machine.

 

[M Paquette]

The 'chatterbot test' is a bit of a cheat. The winner in this year's contest, even more so.

The Turing Test is a test of the machine's ability to exhibit intelligent behavior equivalent to or indistinguishable from a human. It was taken from the "Imitation Game" a game where one is supposed to tell which person is the real deal, and which is imitating them, by looking at written responses to questions. (Anyone remember "What's My Line", the old TV show?)

The drawback is that people are not as clever as they think. When fed a series of canned responses more or less appropriate to questions they might pose, people are pretty easy to fool for a little while. That was the ''secret' behind ELIZA, the 'therapist' chatterbot that ran a parody of a nondirectional psychotherapist doing an initial interview.

The Loebner Prize Turing competition has largely attracted increasingly sophisticated list processing programs with larger sets of response templates. This year's University of Reading competition winner only fooled a third of the judges. Worse, the system played on the judges psychology by claiming to be a 13 year old Ukrainian speaking English as a second language, so errors in the text processing and template driven responses would be misinterpreted by the judges. Also, like ELIZA did, any input that wasn't understood produced a generic question-dodging response.

The winning chatterbot was not doing anything remotely like cognitive processing, symbolic reasoning, or anything more than the most simple natural language processing (find the noun/verb stuff).

Watson is a far more advanced piece of artificial intelligence work.