News for hackers. Stuff that matters.
Re:A.I. Experiments by jedberg on Tuesday November 15, 2016 @07:13PM (#12963541)
It asked me to draw a tree. I drew a palm tree. It said "palm tree" on the bottom, but then said it failed.

I drew the palm tree because I've studied AI and that's a classic AI mistake.

If you go to Hawaii and ask students to draw a tree, almost all of them will draw a palm tree. Ask them to draw a bird and it looks like a parrot (instead of the robin you see typically in the "lower 48").

It's interesting that this seems to suffer from the same selection bias.

Re:A.I. Experiments by ythn on Tuesday November 15, 2016 @08:09PM (#12963772)
It seems biased. When it said draw the moon, I drew a circle with a smaller crater shaped circle inside it, and it immediately guessed "the moon". Later when it asked me to draw a cookie, I drew the same exact shape (circle with smaller circle) and it immediately guessed cookie, not moon. What's going on?
Re:A.I. Experiments by halfdanj on Tuesday November 15, 2016 @10:36PM (#12964365)
The classifier is returning a set of confidence scores, so if the shape you drew probably had a high confidence of both cookie and moon. (in fact if you draw it, and click on it in the endgame screen you can see some details what it saw in the image). We are not hinting the classifier what your task is, so we try not to bias the classifier. Jonas (developer behind quick draw)
Re:A.I. Experiments by Florin_Andrei on Tuesday November 15, 2016 @08:12PM (#12963788)
Sounds like that drawing satisfies both criteria. (shrug)
Re:A.I. Experiments by ythn on Tuesday November 15, 2016 @08:16PM (#12963807)
So it's not really "guessing", it's just seeing if my drawing matches a predetermined bank of answers? Makes sense, since one time it asked me to draw a "police car" and before I had even finished drawing the chassis of a normal car it had already guessed "police car" and moved on.
Re:A.I. Experiments by Meegul on Tuesday November 15, 2016 @08:25PM (#12963845)
It's probably assigning a number of possibilities with a confidence rating, and as soon as the confidence rating is > than some percentage of what it's asking you to draw, it says it.
Re:A.I. Experiments by eyqs on Tuesday November 15, 2016 @09:14PM (#12964014)
I'd say it's "guessing" from a predetermined (rather small) bank of answers, since the same guesses appear over and over again.
Re:A.I. Experiments by halfdanj on Tuesday November 15, 2016 @10:33PM (#12964348)
it's correct that we have "only" trained it on a couple of hundreds of classes, so it will only guess from that dataset.
Re:A.I. Experiments by yummyfajitas on Wednesday November 16, 2016 @06:26AM (#12966104)
As an amusing side note, whether or not a coconut palm (and presumably regular palms also) qualifies as a tree is a pretty controversial topic in Goa.

http://timesofindia.indiatimes.com/city/goa/Coconut-tree-los...

Re:A.I. Experiments by JaumeGreen on Wednesday November 16, 2016 @03:50AM (#12965422)
It also fails when people have been giving it wrong data.

I had to draw an aircraft transport, I draw a reasonable one. What it was expecting, instead, was a plane, As most people just drew that.

Re:A.I. Experiments by Symmetry on Wednesday November 16, 2016 @08:19AM (#12966594)
And of course humans suffer from that too, in a much more sophisticated way.

http://lesswrong.com/lw/nk/typicality_and_asymmetrical_simil...

Re:A.I. Experiments by nl on Wednesday November 16, 2016 @01:37AM (#12964991)
Yes, and this is an interesting way to get more labelled data.