This post makes more sense than the guy that posted the snowy houses vs. the warm living rooms, and said he thought it meant warm colors despite the example being a warm living room.
It says to click the squares that contain traffic lights, not “Click the squares in which at least 50% of the area is covered by an image of a traffic light”
If they didn’t want to count the ones with only a few pixels they should have been more clear in their instructions.
My daughter was trying to sign up for a Steam account. We went through at least 10 of these stupid things… constantly failing somehow, then getting a “try again later” error without telling us what “later” fucking means.
Clicking 3 or 4 panels with correct parts passes most of the time, regardless of how many are actually correct. Sometimes you can even click a completely wrong panel too. No way I’m actually helping them train their stuff on edge cases for free.
You also probably help with edge cases. All it needs for them is to shift (and or rotate) the images slightly when presented to other people to get more accurate results.
See sometimes I choose what I think is correct in these situations and I end up getting them wrong. Once I failed the “choose every square with a car in it” prompt like 5 times in a row. I was so infuriated. The pictures were all blurry af too.
When the societies of the future look back on old humanity and try to locate all of the points of failure which led to our timely destruction, one such point of failure will be when we decided that signing our memes was an OK thing to do.
I’d expect the answers other users have given for a particular captcha will be used to both train the algorithm, and also to decide what constitutes a correct answer.
So the right answer would be “the answer which a majority of humans would choose.”
In my experience, correct means selecting squares that contain a significant portion of the thing. If its a tiny part then it’s irrelevant.
Add comment