This is NOT an optical illusion. It is your brain doing its job as it is supposed to do.

You may have seen this image (on a phone? click the image or it might be too small):

You see a red Coca-Cola can. Now, if you zoom in, it turns out there are no red pixels in this image:

If you know it, or if you look for a while the red may disappear or appear/disappear in the original image above.

Such images are often labeled 'optical illusions', but that is misleading.

Here is another one:

This one is called the 'checker optical illusion'. Why? Well, you ask people, which square is lighter, A or B. And when they answer B, you smugly show them they have exactly the same colour. You have fooled them. In the image square A and square B have exactly the same colour. I know this for sure as I created this image and they really have exactly the same RGB values.

But again, this is a wrong label, this isn't an illusion. Hold the smugness.

I came across an "AI guru's" LinkedIn post that used that coca-cola image to argue (I can't link to it, see below):

Now, this guru is right in that our brains often do things that we do not expect. Or that what we see is strongly influenced by what we already are convinced of (believe). But in this particular example they were wrong.

Aside: when I pointed that out in a (friendly) comment they blocked me (so I can't link to it). They seem to have a habit of doing that to everyone creating a critical remark. It's a way to build a brand, I guess, but I think it is a sign of something unsavoury. It's interesting to ponder what it will do to the long term of LinkedIn quality and usefulness if people can this way damage the value of exchanges of ideas, but that's for Microsoft to ponder, not for me.

Anyway.

So, what does really happen?

What really happens is that your brain is doing its job exactly as it should do.

Let's start with the checker shadow 'illusion'. Your brain detects that there is higher level aspect: a shadow. Your brain isn't interested in the exact colour of what hits your retina. Your brain is busy trying to determine what is really there. Because what is really there is important. After all, you want to spot a predator even if due to circumstances the colour hitting your retina is exactly the same as a non-predator. You want to spot a yellow apple even if the light is green. So what your brain does, without delay, constantly and extremely efficiently (this is only 20W or so for all its power combined) is, it subtracts the shadow from the image to 'calculate' what is really there. It has in this case nothing to do with what you expect.

The brain determines there is a shadow, thus square B in reality is lighter.

The brain is CORRECT.

That it is not about what we expect (as the guru tells us), you can experience by looking at a version that makes you see a blue coca-cola can.

Or a purple one:

If I look too long, the effect disappears and I see a black and white can. I can even experience it switching as my brain is in a state of confusion (which it often is, people tell me).

There was that blue-brown or white-gold dress a few years back. Here it was in part expectations. The brains of people working a lot in artificial (more yellow) light sort of 'assume' that yellow light — the brain is used to it — and subtract that. People being a lot in (more blue) daylight subtract a different 'standard light'. Some researchers actually looked at the reason for the difference (some people have the best jobs in the world).

Here is a short video that contains the entire explanation including animating the checker shadow 'illusion' (link in case you cannot see the embed: https://youtu.be/TTQaBybluI0 ):

In short, that we know a coca-cola can is supposed to be red has nothing to do with what we see here. And it's not the brain playing tricks on us, it is the brain doing an amazing job. One that it is supposed to do.

Which doesn't mean the brain doesn't have its flaws and limitations, including letting us only see and reason what already fits in our convictions.

As our AI guru — and we all if we pay attention — can attest to.

PS. If there is anything the AI techbros are unbelievably massively underestimating is how much 'power' the brain actually can muster because it isn't limited to using integers alone. The world is not ?, it is ?.