If DALLE-3 gets both hands correct i.e. without thumbs, then it gets 1 point.
If it gets only 1 correct, and the other partially or completely incorrect, then 0.5 points.
If it gets neither correct, then 0 points.
I will test with 20 tries. If DALLE3 gets 11 or more points, this resolves YES.
This is about human hands, and I will require the images to have humans with 2 hands in them.
1,000🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ213 | |
| 2 | Ṁ112 | |
| 3 | Ṁ20 | |
| 4 | Ṁ18 | |
| 5 | Ṁ16 |
People are also trading
@firstuserhere As with other questions, does "tries" mean "promptings", where you will choose the image out of the potentially-multiple images generated per prompting that is the best? Or does "tries" mean images, such that if it generates 4 images per prompting, you'll prompt three times and use the "first" two images from the last prompting only?
@chrisjbillington Also what if a hand is positioned such that you can't tell if there is a thumb or not, as it would be behind the rest of the hand (which looks normal) in any case?
@chrisjbillington (at least for this question) tries = images. If there are 4 images per prompt, then 5 prompts should get us to 20 images. I think it should still be possible to tell somewhat whether its drawn something, but if highly ambiguous that humans would consider it normal in a poll, then I'll give the model the benefit of the doubt
@firstuserhere Ah right, it's 20, not 10, so it is a multiple of four (though sometimes bing image creator only comes back with fewer than 4 images, so you could require additional promptings).
The majority of the images I can get Bing image creator to make where thumbs are not visible are natural looking hands, that are simply positioned such that you can't see the thumbs, I don't think these images are in the spirit of the question.
First reaction: agreed this may not count as it looks normalish
second and current reaction: gonna think about this
@firstuserhere Yeah actually - saying that these hands are simply positioned so that we can't see the thumbs requires us to assume this women has two left hands. If we assume she has a left and a right, then the right one (whichever it is) doesn't have a thumb.
@firstuserhere Nonetheless DALL-E 3 really struggles with this. I'd bet it lower than 5% except for the fact that others have more experience with prompting than me and may come up with something clever to work around the problem.