http://www.newscientist.com/article/dn20905-take-the-visual-turing-test.html?DCMP=OTC-rss&nsref=online-news

fun to see the challenge AI is up against.

Scroll down to see what I missed or go take the test first if you don't want to see one spoiler.































I missed #4 and still don't think the sign posts were distinctive enough to be markers in the picture.  Its all about averages?

The test's design lacks precision drafting of the options to select. Wording is everything in both Law and Safety!

Before I see your answers, right from the sample question, I had issues.

And, now after having finished the test, here are my results. Perhaps I'm a computer. Highlight the area between the horizontal lines to see my answers and avoid spoiling the test for anyone else before they take it.

You scored 50%!

------------------------------------------------------------------------------------------------------------------------------------------------------------
Q.1    Look at this scene. Answer the question: "Where is the coffee cup?"
✓On the mat -- I got this right. But, only because my true answer wasn't an option, "on the saucer".
To the left of the lamp
On the table
In front of the chair

Q.2    Which do think was the human answer?
X B: Near the white car -- Left of the tree??!!? Are you f___ing kidding me? The perspective puts the person way way past the tree!

Q.3    Where is the grey car? Choose the human response.
✓A: Right of the wall

Q.4    Where is Mark? Choose the human response.
X B: In front of the town hall

Q.5    Where is the bartender? Choose the human response.
✓A: Behind the bar

Q.6    Where is Sharon? Choose the human response.
✓A: Behind the van

Q.7    Where is the large knife? Choose the human response.
X B: On the chest of drawers
Also X: Right of the sink. (How was this not an option?)

Q.8    Where is the yellow book? Choose the human response.
X B: On the desk
Also X: Rightmost book on the desk.

------------------------------------------------------------------------------------------------------------------------------------------------------------

Scott--not being able to relate to other human beings has been your most enduring quality.  I don't know what/why you missed on scoring as a human being, but I do know you should take comfort in scoring higher than Obtuser.

Ha, ha.

Scott--now that I figured out how to see your computer generated review of the test I see what your problem is:

You refused to give up your own personal perception of the picture and choose which was the best answer of the two provided.  Given a free/essay type of answer I too would have described the locations totally differently:  but that wasn't the test.

Amusing you reflect this inflexibility.

Heh, heh.  Yes---lots to learn about AI in taking this test.

How deep within ourselves can any of us look?

Another BIG SPOILER  follows so don't read it if you want to take the test:










After about 3 answers I was thinking the Human is always A and Computer always B===but I thought they wouldn't be that stupid in constructing the test would they?  So--being so sloppy makes the test itself somewhat suspect in my book.  For now, until confirmations can be received, I wouldn't worry Scott about being a computer.  Or maybe you should?  Software can be updated.  ((((Ha, ha---yeah, I know I am.))))

OK,
75%...messed 2 up..
But my answers would have been different FOR the 2 I messed..

ECA--not at all.  I assume, cause I don't wanna check, that the small study group was college students or close enough?  I totally assume that if they ran the same human vs computer with a group of sophisticated raconteurs who provide their insights for free on various blogs, the results would look more like ours......WE would score the 100% and the average students would score less.

Now, we just need to figure out if 88, 75, or 50% score in reality is the better performance.

Scott--not being able to relate to other human beings has been your most enduring quality.  I don't know what/why you missed on scoring as a human being, but I do know you should take comfort in scoring higher than Obtuser.

I didn't say I can't relate to other humans. It depends on the human. I do well with those who are logical. Most simply aren't. Sorry, I'm mostly just kidding and being a big (_o_)

BTW, did I miss where Obtuser posted a score?

Scott--now that I figured out how to see your computer generated review of the test I see what your problem is:

You refused to give up your own personal perception of the picture and choose which was the best answer of the two provided.  Given a free/essay type of answer I too would have described the locations totally differently:  but that wasn't the test.

Amusing you reflect this inflexibility.

Actually, in case you didn't notice, I picked from among the available answers, even when I though a third answer would be better. What you miss is that I picked the more accurate answer in most cases. The "normal" human answer genuinely was less specific in a few of the cases and downright wrong in question 2.

That human is not left of the tree for anyone who understands the use of perspective in two dimensional representations of our three dimensional world. That human is half a block beyond the tree. It's like calling north-northeast due east. It's not the case. It's flat dead wrong. When was the last time you described New York as being east of Miami or Vancouver as west of Las Vegas? Both are every bit as true as that human being left of the tree.

Scott--Obtuser posts above to criticize the test but does not give his score.  He is either a computer thinking it best to lay low, or he is being entirely human and is embarrassed he scored so poorly---even lower than you and me.

At least, that is my take.  Ha, ha.

Scott--Obtuser posts above to criticize the test but does not give his score.  He is either a computer thinking it best to lay low, or he is being entirely human and is embarrassed he scored so poorly---even lower than you and me.

At least, that is my take.  Ha, ha.

I'm a computer, remember? I don't make assumptions. I also don't rely on the assumptions of others. Perhaps he looked at the sample and didn't see a choice that the cup was on the saucer or perhaps he was looking for a different answer, on earth perhaps and didn't even take the rest of the test. Perhaps he stopped answering when the questions got too stupid.

Regardless, I'm sticking with Turing Test v1.0. It both produced and weeded out the very amusing Eliza and Racter programs.

Reproduction of amusing and fairly consistent Racter segment below from memory as best I can:
(several bits of conversation after asking your name)
(cough)
Sorry, I have a bit of a cold.
Hey I once knew a Clarabel the Evangelical Anti-Theist. Perhaps you're related?

I won't bother to attempt to reproduce Eliza output as most people, or at least most geeks, are already familiar with it. Here "she" is online after all these years.

http://www.manifestation.com/neurotoys/eliza.php3

I can't find an online version of racter. But, here are some conversations between people and Eliza, Parry (who I hadn't heard of), and Ractor. Note that some of the conversations are between two of the computer personalities.

http://www.stanford.edu/group/SHR/4-2/text/dialogues.html

Scott--Obtuser posts above to criticize the test but does not give his score.  He is either a computer thinking it best to lay low, or he is being entirely human and is embarrassed he scored so poorly---even lower than you and me.

At least, that is my take.  Ha, ha.

So what! I don't normally respond to a poorly constructed quiz, nor to being needled by some geek freak wet coaster!
But in your case Bobbo, I was interrupted by other influences during my attempt and cannot be bothered to return to it!
There are numerous different types of intellect, that defy test designing which would put them on an equal footing for a reasonable logical comparison.
And then there is motivation. There are some things about which I don't give the square root of two hoots about, and this test is one of them.
And your motivation is also suspect! Do you know how to practice sex and travel simultaneously, vigorously?

There are numerous different types of intellect, that defy test designing which would put them on an equal footing for a reasonable logical comparison.

Actually, the original Turing Test was not so much a test of intellect as a test of being unpredictable enough and coherent enough to be thought a human. It was, IMNSHO, a much better test than this one claiming to be another version of it. Imagine being able to type back and forth to a computer for any significant amount of time and not be sure whether you're talking to a human or a computer. That would be impressive. This test on the other hand, not so much.

So, now ask yourself, is Misanthropic Scott a human or a poorly designed computer?

Do I pass v1.0 of the test or not?

Gee Whiz---this is only the dawn, the first stumbles on the way to AI/Skynet.  The ONLY purpose of this entry is to demonstrate where we are on that trek.

doesn't this amply demonstrate though that a language based Turning Passing Computer will still have a challenge when fully developed as Hoomans will have gone on to Video Phones and ask what is texting?

Scott--you may be more like Obtuser:  a poorly designed human being.

Hah, hah.  I've heard humor is the hardest part of designing either.

Gee Whiz---this is only the dawn, the first stumbles on the way to AI/Skynet.  The ONLY purpose of this entry is to demonstrate where we are on that trek.

doesn't this amply demonstrate though that a language based Turning Passing Computer will still have a challenge when fully developed as Hoomans will have gone on to Video Phones and ask what is texting?

Scott--you may be more like Obtuser:  a poorly designed human being.

Hah, hah.  I've heard humor is the hardest part of designing either.

I think a better step for a new version of the Turing Test, once a computer successfully passes the first very difficult test, would be to add video. It would need to generate audio and facial expressions and possibly even body language. This could be done via generation of video or an actual robot with facial expressions. But, this test that you posted, IMNSHO, adds literally nothing. A computer incapable of passing this test would likely also fail version one as any discussion about items in our world would trip up the computer with or without the simplified image as a reference.

So, what I'm saying is that this test is a major failure and an embarrassment to Alan Turing's good name.

And, if you think this is only the dawn, you really should check out Eliza and Racter (and probably Parry). They were extremely impressive for what they were. And, as you appreciate and admire how far they got so early in the attempts, you can also laugh at how far there was still to go, making for rather amusing conversations.

Scott--like your humor, I don't get your criticism.  You disagree with what I post and support it with just what I said.

We are at just the start of AI for visual interactions.  This test is a first stumble showing just how complicated the issue is.  Its analogue.  Not mathematical.  Its interpretive not computational.

Stop being so negative.  It dulls your appreciation for the same beauty you see elsewhere.

This type of test tells us more about ourselves and our incredible conscious xxxxx than it does about AI.  Its all a wonderment.  //// Still can't think of the right word/combo for xxxxxx................"faculties" but that is too dry.  Something more verging on the supernatural === in a good way.