- cross-posted to:
- opticalillusions@feddit.nl
- science_memes@mander.xyz
- cross-posted to:
- opticalillusions@feddit.nl
- science_memes@mander.xyz
You must log in or register to comment.
deleted by creator
I can easily see it’s black and white when the image is expanded, but with the smaller non-expanded version of the image, my brain just refuses to interpret it as anything other then red.
I see gold and white
It’s clearly saying “Yanny”.
listen again it says brainstorm
My brain is a bad employee, always taking lazy shortcuts. No one wants to work anymore smdh
my brain
That is, you.
Is it arranged in such a way that gives the illusion of red, or do we mentally fill in the blanks because we already know what colour a coke can is supposed to be?
it’s the pixels.
Yes.
So yes is answering the first question or the second question?
Yes.
But more seriously, I don’t know. At first, I thought it was the first one, if only because of the choice of magenta as a “counter color.” But then, when I partially cover the can, I no longer perceive the uncovered part as red. So, I don’t know.
That’s pink
Finally a good idea for a captcha ai could not guess … 😉👍🏻
well, it’s rather the opposite - AI is vulnerable to these kinds of tricks, you can inject some patterns into images to make the AI perceive the image in a completely different way, humans are in fact much less susceptible to this, at least for now
So how does ai exactly knows which color is the picture… I always thought that by analysing the binary it is possible to recognize previous learned sequences for predetermined colors and as such answer the question… In this example there’s no sequence for red for example…
AI basically recognizes patterns and returns a likely output based on the patterns it found. It doesn’t check “this is a red pixel, so this is red”, it checks “this pattern is usually red, so this is red”.
And what about shapes that doesn’t have a default colour? For example an apple ? Red green yellow ? Or a bold number
It will come up with an answer at random or have bias based on the training data or one based on random chance.
Zoomed in, some of the pixels do appear pinkish. Has anyone checked the rgb values to verify that there’s no red shade at all?
