[Art]

Forget the Turing Test - Try this man's idea instead...

http://www.newscientist.com/article/mg22429992.900-forget-turing--i-want-to-test-computer-creativity.html#.VJFWbntdDEU

[squarebear]

Creating original poems and jokes has already been beaten. Mitsuku has twice won the funniest ever computer contest by doing this:
http://funniestcomputer.neurogrid.com/2012-transcripts/mitsuku

http://funniestcomputer.neurogrid.com/2013-transcripts/mitsuku

[ivan.moony]

something similar can be produced already today: writing an algorithm by description. I think it is a kind of intelligence also.

something like: write us a program that sums 10 numbers. if the result is greater that 50, divide it by two while the result is even.

Yet, there are still problems with natural language processing, but in OpenCog they claim they have achieved some progress in that direction. Right now they are talking there about inputting Wikipedia articles into knowledge base. I'm thrilled by this possibility.

[Art]

Anyone actually read this part?

How is your Lovelace 2.0 test different?
In my test, we have a human judge sitting at a computer. They know they're interacting with an AI, and they give it a task with two components. First, they ask for a creative artefact such as a story, poem or picture. And secondly, they provide a criterion. For example: "Tell me a story about a cat that saves the day"; or "Draw me a picture of a man holding a penguin."

I just think it might show strengths or weaknesses of different bots, those scripted with canned material vs those able to draw upon various data sources and able to formulate complete, perhaps semi-complex, thoughts.

There is obviously more than one way to indicate intelligence, but are they just looking for a witty conversationalist or a creative chimp with a typewriter?

[squarebear]

First, they ask for a creative artefact such as a story, poem or picture.

Tell me a poem...

And secondly, they provide a criterion.

...about a cat/pencil/T-shirt/river etc etc

Already done.

[Don Patrick]

Lady Lovelace spinning in her grave.

First off: It's not his idea, it's someone else's idea only slightly refined, and even then it's still a pretty pointless exercise. The best this addresses is a century-old objection that robots can't write poetry or turn a canvas into a masterpiece, which went hand in hand with the arguments that they can't do math and can't play chess. Is anybody going to create a human-level A.I. to pass this test, or train their human-level A.I.-in-the-works in the arts? No, they'll just build a better keyword-oriented painting robot, the sole use of which is to pass this arbitrary test.

[squarebear]

No, they'll just build a better keyword-oriented painting robot, the sole use of which is to pass this arbitrary test.

Spot on! Which is exactly what I did to win the Funniest Computer Ever contest.

[Don Patrick]

I did not follow that contest very closely, but I thought it was very clever how you did something similar towards the X-Prize for artificial intelligence (another proposed alternative to the Turing Test), where the A.I. was to make a speech about a given subject.

Of course, my real motive for opposing this test is that I have a lot of artist friends who might lose their jobs when people start building artistic robots.

[8pla.net]

Is the http://www.funniestcomputerever.com contest still going?

[squarebear]

There isn't a contest this year 8pla. The organisers are venturing into new ground and didn't have time to run it this year.

Don - Thanks, glad you liked it. I guess it goes to show that whatever proof of intelligence people expect to see, there's usually a way to code it.