Hacker News new | past | comments | ask | show | jobs | submit login
All the balls are the same color: brown (twitter.com)
90 points by vinnyglennon 3 months ago | hide | past | web | favorite | 42 comments

The smaller the image, the more effective that illusion. Looking at it on my 20+" screen at almost fullscreen I can look at individual alls and clearly see the as "brown." But with a smaller image, even looking directly at the ball, I'm seeing more of the color of the bar than the color of the bar. This makes me wonder if there's a physical aspect to the way this illusion works, maybe something to do with structure of the eye?

In the eye, cells only fire if the signal between the neighbors is great. There is also higher resolution concentrated at the center and drops off as you move away. Additionally the brain processes signals in context of the last signal it received. So if you put that all together, on a small screen the gaps are small enough that in the center of the eye, fovea, it will be processing more red than brown. You can also see the ball is actually not just brown. Its shaded which reduces the "fire near highly different neighbors" effect. The red though is solid.

There is also the fact that the eyes resolution drops off towards the edge. This makes the effect even harsher in the corner of your eye. As your eye moves from ball to ball your brain will be processing the new higher resolution versions based on the previous versions you saw in the corner of your eye. That effect fades away but your eye is actually always moving doing something called saccades, on a small screen with many prominent points it will do it a lot.

With a big screen alot of that goes away. The gaps are larger. The brown fades more slowly from light to dark so is percieved as a stronger color. Objects are larger so saccades will be less effective because they will be searching fewer objects rather than many in a tight screen. Lastly the lower resolutions at the edge of your eye will pick up more brown as well.

So there most definitely is physical aspects.

The color of the brown is also not the same: in addition to the several pixel transition area which depends on the stripe color there are slightly red, green, and blue tinted patches in the rest of the brown that differ based on the stripe color, which helps create the neon glow effect. Calling the balls all the same color seems to be entirely incorrect.

I also see an "egg" illusion where it looks like the lower left of the ball is actually flat but the rest is a dome.

I would imagine it has to do with the retina where photoreceptors exist

/e all of the physical stuff has to do with magnification of the image (or incoming light) onto the retina

Yes, if you make the image big enough, looking at each ball individually it look brown, but the ones in the periphery take on their different colours.

This poses a problem for machine learning. Optical illusions exist in real life, but when does your AI decide to go with 'brown' and when does it imagine to go with 'what humans would think'?

I look forward to a robot future where us humans can have a joke and catch out robots by enjoying these optical illusions knowing 'they' don't get it.

I believe this illusion is exploiting a quirk of human sensory vision so any machine learning will not be affected. It's the same property that most video encodings and compression algorithms exploit: Our colour and luminance perception does not have equal fidelity, i.e we dont sense as much detail in colour, where as we can sense more detail in luminance... in this case it means we can perceive the stripes, and given stripes alone we can just about discern their colour, but when mixed with a background colour it effectively blends the colour without blending the luminance.

I think this might be the same biological reason that we can't perceive much (or any) colour in low light, rods vs cones.

TL;DR our retinas are not CCDs they are a patchwork of two discrete components.

>I believe this illusion is exploiting a quirk of human sensory vision so any machine learning will not be affected.

That's the point my dude.

Humans see an optical illusion, robots do not. Meaning ML algos that care about seeing things as humans do have a conundrum.

That makes no sense. It's all about the training.

A human can tell ML that the balls in the example are either perceived red (illusion) or reality (brown). The computer will use that training model for future decisions.

Why wouldn’t the robots see it?

Unless your camera is a 3CCD camera, then CCD cameras usually have lower color resolution than luminance resolution, too!

https://en.wikipedia.org/wiki/Three-CCD_camera https://en.wikipedia.org/wiki/Bayer_filter

if you label it as such, ML algorithms will learn it as such, no ?

aha, yes i suppose, learning human fallacies, that would be quite confusing when it's perceiving something quite different to us, it would probably take more effort to learn compared to just manipulating the input image.

This kind of learning bias has already been shown to be eeked out. The 'criminal' classifier that was used in one of the US states (kentucky or tennessee i believe) for sentencing and parole decisions, was shown to carry forward the human biases from its training data. ML is only as unbaised as its training set.

iirc, james-mickens has a usenix presentation about the dangers of using ‘ai’ in this sphere of human activity.

You make a good point, though different people perceive optical illusions differently. Remember the black/blue/gold/white dress? What colour should an AI call the dress when humans can't even decide? So it's actually even more complex than you might think.

This question also applies to audio illusions, such as the (in)famous Laurel/Yanny clip.

"Brown" is already going with "what humans would think", from the sensors that try to capture the image according to the peculiarities of human vision, to whatever representation the system has of the social construct of "brown".

If you squint your eyes, the balls look red, blue, and green, but that's to be expected if there are colored bars over them. And if you don't squint your eyes, the "balls" look about the same color... So what's the illusion?

Perhaps different people experience this image differently. For me, viewing the image on an iPhone, with no squinting at all, the balls look strikingly red, green and blue. It's only after zooming in on the image, that I convinced myself that the balls were brown with colored bars over them. As others have suggested, perhaps a small image size is what makes this illusion work.

Zoom out or move a little further from the screen and try again.

You could even say that actually, the balls aren’t brown. They are made up of pure red, green and blue subpixels on your screen.

It’s really not a particularly interesting illusion.

You're suggesting that this wouldn't work if it were printed or a painting?

No, I’m saying it’s the same type of illusion. Small areas of color blend together and look like a single color to our eyes. (CMYK printing of course works on the same principle)

That's not so clear. The LCD subpixels and CMYK dots may well be below the optical/retina resolution of the eye, in which case it's not really an illusion at all as it's already in the signal data the brain receives. This is quite likely so, because you cannot see those dots individually even with concentration. The same likely goes for squinting your eyes as they become optically unfocused.

In contrast, here the issue is undoubtedly in how the brain interprets the signal data, i.e., an illusion.

Subpixels of modern high resolution screens are below what a normal eye can resolve, but that really isn’t necessary to perceive apparent shades of color. You get the same experience even with a very low resolution RGB screen, where you can distinguish the subpixel elements easily.

It’s not right to say that the balls are “brown”, when they are in fact striped brown and red/green/blue. It’s like saying that a zebra is black if you just ignore the white.

Even then the paint / toner would be a mix of a small number of base pigments in most cases.

The following blew my mind. I asked my four year old twins and both told me "brown, brown, brown. All balls are brown".

Can it be that such an illusion is learned?

I wonder if that says more about their focal attention. If I really fixate on a ball, it becomes the same brown or tan color. I can scan from one ball to the next and mentally confirm that they all have the same color. While doing this, I actually feel my mind's "white balance" compensate as I shift my focus from one ball to another.

But, by default, I do not fixate on any ball. I take in the gestalt composition, and in this mode I perceive a field with three classes of balls with different tints. This happens as soon as I reframe my attention, even right after I have just sequentially confirmed their color as above.

There was also one interesting comment in the Twitter thread, posting a version that desaturated only the color stripe that overlaps the ball (at a bounding box rather than right at the circular edge of the ball). In this version, the gestalt composition shows me the same tint on all balls, and instead I see three classes of tinted tile in the background behind the balls.

You may want to look into whether it's a learned phenomenon and also consider getting them tested for color blindness.

If you look at each ball one at a time, it's obvious that each is brown. The effect only works if you look at the balls with your peripheral vision and then make a summary judgement.

I'm zooming in on one pair until each is about one hand-width in diameter. The effect is still present for me.

That's very interesting. But I think it probably is the fact they have fresher more youth receptors. If you think of when kids touch warm things and say it's too hot, it's a similar concept. Their eye receptor will fire more strongly reducing the main effects of the illusion. Although, if you are on a bigger screen, it may be that it's your eyes that are failing. Bigger screens or zoomed in this illusion breaks.

All perception is learned. Our sensory machinery just gives us noise. We learn through experience to differentiate the signal.

For example, paralyzed babies will also be blind, even if their bodies are moved by someone else.

Yup, my 5 year old saw them all as brown.

The point of this illusion is that it goes against what we expect for colour discrimiation. We would expect the balls where the red bars going over the top to look less red not more red.

It's also worth checking out his research: https://twitter.com/chazfirestone/status/1109105677006487552

On mobile this is a perfect illusion, unless you physicslly hide all but one ball.

I think it is the same principal as simultaneous contrast.


You can actually see the real color if you focus on one and pay attention to your peripheral vision

Though the shade of brown is different, right?

Doesn't seem to be. I edited out all the lines with Photoshop, and what remained looked identical.

Yes, that's how RGB pixelization works.

Applications are open for YC Winter 2020

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact