[Herra Tohtori]

Not reconstructs, really. More or less just...constructs.

I'm treating the original scenery as an image that the eye as an optical device observes.

That image is reconstructed into visual perception by the visual cortex, rebuilding it from the fragmented saccadic observations of the eye, which constantly update the "expired" sections of the image, creating a more or less seamless illusion of continuous image.


If you consider the visual cortex' job as the actual image (rather than the reality beyond the observer's eye), then your choice of verb is valid, of course.

[General Battuta]

Not reconstructs, really. More or less just...constructs.

I'm treating the original scenery as an image that the eye as an optical device observes.

That image is reconstructed into visual perception by the visual cortex, rebuilding it from the fragmented saccadic observations of the eye, which constantly update the "expired" sections of the image, creating a more or less seamless illusion of continuous image.

You're still making the mistake of thinking that the image that eventually reaches you is somehow reconstructed from information retrieved from outside. It's not. The minimal information the eye delivers is fed into this huge mechanism that then fabricates the vast majority of what you see from whole cloth (obeying evolved rules) and produces a useful image which we can use to navigate and manipulate our environments.

The image isn't necessarily 'true' in that it reflects some objective reality - it's not reconstructed. It's a signal, a very rich form of symbolism. What we see as a right angle maps nicely to the form and function of a right angle, but 'the look of a right angle' isn't something that exists outside the brain.

If you consider the visual cortex' job as the actual image (rather than the reality beyond the observer's eye), then your choice of verb is valid, of course.

I don't think there's any 'reality' to vision outside the visual cortex. Vision is like a HUD. It's a bunch of labels that allow us to behave usefully, applied to a very narrow stream of information.

[Herra Tohtori]

You're still making the mistake of thinking that the image that eventually reaches you is somehow reconstructed from information retrieved from outside. It's not. The minimal information the eye delivers is fed into this huge mechanism that then fabricates the vast majority of what you see from whole cloth (obeying evolved rules) and produces a useful image which we can use to navigate and manipulate our environments.

The image isn't necessarily 'true' in that it reflects some objective reality - it's not reconstructed. It's a signal, a very rich form of symbolism. What we see as a right angle maps nicely to the form and function of a right angle, but 'the look of a right angle' isn't something that exists outside the brain.

I don't think there's any 'reality' to vision outside the visual cortex. Vision is like a HUD. It's a bunch of labels that allow us to behave usefully, applied to a very narrow stream of infromation.

It doesn't really matter, though. If what you say is true, the visual cortex could patch up a convincing visual perception using any image source that it gets used to, whereas trying to mimic that patching process is likely to fail hard.


Will you at least agree that up to the input onto the visual cortex, the information from the eye is essentially an image that was focused on the retina by the cornea and lens of the eye, then translated via ocular nerve to the visual cortex where it fires up neurons in approximately the same pattern as the original image on the retina?

What happens after that doesn't really matter to me. It strikes me as the easiest way to achieve artificial sight would be to mimic the process of projecting image data as input to visual cortex, and letting the brain take care of the rest however it wants to fool itself into thinking it as a natural visual perception.

By the way I disagree with your statement - we receive quite a bit of information visually, and even though it's not momentarily nearly full field of vision, it's completed by the brain quite effectively.

And if you refer to the question whether we "see" things similarly - whether we all see red similarly, or just identify it as "red" because it's the label we've given it - in my opinion, yes, we see it similarly up to the point where information enters visual cortex. And that the visual cortex patches the image data into a full field of view vision, I don't really care how our more abstract thought processes handle it, but the sensory stimulus is identical.

[General Battuta]

By the way I disagree with your statement - we receive quite a bit of information visually, and even though it's not momentarily nearly full field of vision, it's completed by the brain quite effectively.

I don't think it's a matter of opinion. Properties like color do not exist outside the brain, they're synthesized from physical information and used (imprecisely) to signal it.

And if you refer to the question whether we "see" things similarly - whether we all see red similarly, or just identify it as "red" because it's the label we've given it - in my opinion, yes, we see it similarly up to the point where information enters visual cortex. And that the visual cortex patches the image data into a full field of view vision, I don't really care how our more abstract thought processes handle it, but the sensory stimulus is identical.

I haven't commented on that question, nor on the question of where input should be introduced to create artificial visual input.

[Bobboau]

I don't think it's a matter of opinion. Properties like color do not exist outside the brain, they're synthesized from physical information and used (imprecisely) to signal it.

but I can quantify color, how can I quantify a completely fictitious property?

[General Battuta]

I don't think it's a matter of opinion. Properties like color do not exist outside the brain, they're synthesized from physical information and used (imprecisely) to signal it.

but I can quantify color, how can I quantify a completely fictitious property?

Color is a consensual hallucination based on actual, objective physical properties. The fact that we all share the same color wiring makes it easy for us to establish consensual measures, and it's even easier to quantify the physical properties color is based on.

[Bobboau]

yeah, but if there were an alien that had completely different perceptions of color than we did, we would still be able to communicate in numbers the color given off by an electron moving from the second to the first energy level in a hydrogen atom. you could make the same argument for anything we sense.

[General Battuta]

yeah, but if there were an alien that had completely different perceptions of color than we did, we would still be able to communicate in numbers the color given off by an electron moving from the second to the first energy level in a hydrogen atom. you could make the same argument for anything we sense.

No we wouldn't. You're confusing color (which is a constructed property based on multiple physical traits) with the wavelength of light.

Wavelengths do not map correctly to colors.

[Bobboau]

please elaborate.

[Mustang19]

The sun is yellow but it looks white.

[General Battuta]

please elaborate.

I was going to post this longass thing about L and S cones but basically this does a far better job.

tl;dr version, very different combinations of wavelengths can appear as the same color for reasons which are entirely resident in our eyes and brains. Color does not map 1:1 bidirectionally to wavelength.

[MP-Ryan]

Will you at least agree that up to the input onto the visual cortex, the information from the eye is essentially an image that was focused on the retina by the cornea and lens of the eye, then translated via ocular nerve to the visual cortex where it fires up neurons in approximately the same pattern as the original image on the retina?

I'll interject again here - no.  The retina performs perception/interpretation functions.  The original image pattern is lost as light strikes the photoreceptors.  What lights up on the surface of the visual cortex is the photoreceptor firing pattern, which is not a true pattern of the original image.  Due to their method of data collection, photoreceptors "perceive" rather than "observe."

Human (and other related animal) sight has biological consensus due to evolution, but sight gives us a [consensual] perception of the world, not an observation.  Battuta has been discussing this.

Try to calibrate the input signal to visual cortex to match as closely as possible the qualities of a biological eyes' retinal inputs, and let the subject's brain adapt to it and learn to process and reconstruct a visual perception from that data. The brain is so far uniquely powerful and adaptable marvel of signal processing, so I bet it'd do the job better and faster than trying to bypass the image reconstruction phase done by the visual cortex. Of course, it would not be quite the same as the "natural" vision due to differences between signals from natural eyes and artificial ocular implants, but especially if the implantation surgery was done at young age, I think it would have spectacular resources... ignoring the current practical difficulties of hooking up raw image stream to the brain at sufficient and meaningful resolution.

I think the trouble there is that the brain may inadvertently try to apply its processing rules in a way we can't anticipate.  It's a technological hurdle that we'll all just have to wait and see on =)  If I remember correctly, though (and this is a HUGE if, this class was 5 years ago), the visual cortex does perform a consolidation of visual perception before shipping that information out to the rest of the brain.

[ssmit132]

You're confusing color (which is a constructed property based on multiple physical traits) with the wavelength of light.

The fact that "colour" is a constructed property rather than a physical characteristic of an object, I found very interesting.

Now that I think about it, I learned recently that the Mantis Shrimp (which was an animal I already liked due to its powerful claws and all the funky effects that they cause) can see more "colours" than any other known animal.

But I'm probably going a bit off-topic on that point.

[Mongoose]

I dun care I just want my Ghost in the Shell cyber eyes dammit

[watsisname]

Cover one eye for a while and stare at a bright light (not the sun, lol).  Then look around and swap between both eyes.  Colors between the two will look quite different.
What we perceive as color is not only not a direct function of wavelength (rather it depends on many values, as Battuta said), but it also varies on the relative color and intensity of the ambient light that your eyes are currently adapted to.

[Herra Tohtori]

Certainly colour perception is a result of multiple things, and it's affected by what else we see in the image (hence the colour-distorting visual illusions) but I still maintain my view that colour is primarily a function of wavelength of photons and the spectral spread of the light arriving to the retina. The fact that our perception alters colours depending on different stuff doesn't make the property of colours any less real.

Leaves are green because they have no use for the green wavelengths and mostly reflect them, while absorbing the blue and red wave lengths.

A monochromatic beam of light would be identified correctly within the limitations of our system.

The fact that our brain does signal processing doesn't mean it invents all the qualities in the image we see. A rectangle is a rectangle in reality, and if it reflects red light and absorbs others (or even emits red light in otherwise dark room) then its colour is red. It's not some imaginary quantity that our brain just plasters on, it's a perceived property corresponding to physical property and I don't see how it's relevant to say that there's "no colour" in external world.

I can agree that perceived colour is a mix of not only the wavelength spread of the light but also the context in which it's seen, physical surface structure (diffuse, shiny, emitted or reflected light), environment's lighting (is the object in a shadow or direct light), but that just means our image processing facilities combine these features with each other to give us more information than just the spectral spread of light that an object emits or reflects.


The information about retina already doing image processing is interesting. I always considered retina more or less like a digital camera cell with four types of pixels, one sensitive to light and three to different wavelengths of light.


watsisname: the phenomenon you are talking of is analogous to white balance adjustments on cameras. The level of low light acclimatization shifts the white balance, so if your eyes are at different setting, obviously the perception shifts around. It's like setting one camera to daylight and one to fluorescent and complaining that the images have different colours.

By the way, almost same phenomenon can be achieved by staring through cyan/red anaglyph stereo glasses for a while. Eyes and the brain try to compensate for the saturation of those particular colours, so when you take the glasses off, the red lens eye sees things more cyan, and the cyan lens eye sees things more reddish. It is fascinating phenomenon.

[watsisname]

the phenomenon you are talking of is analogous to white balance adjustments on cameras. The level of low light acclimatization shifts the white balance, so if your eyes are at different setting, obviously the perception shifts around. It's like setting one camera to daylight and one to fluorescent and complaining that the images have different colours.

Oh I'm not complaining -- white balance and dark adaptation are both fantastic evolutionary developments.   I was simply stating that an individual's color perception is not constant.

On a semi-related note, another interesting vision-related concept I discovered was back when I acquired my 405nm violet laser.  It has the same output power as my other, 532nm green one (~150mW), but it appears to be much, much fainter because this particular wavelength of violet is very close to the limit of a human's ability to detect.  It's just really weird to me to see two dots of light side by side, one of them looking significantly brighter than the other, but knowing that they are in fact the same (easily confirmed if the violet laser is shined at something that converts violet light into blue or green).

[Luis Dias]

Colours are formed because our eyes have three types of receptors, red, green and blue. They mix together and inform the brain about this mix with a myriad of colors.

TV sets use the exact same reconstruction of colours in RGB not because "RGB" is a fundamental law of physics, but because they mirror our photoreceptors pretty well. An animal who would have different photoreceptors would probably not be able to see TV very well.

[Ghostavo]

please elaborate.

What's the wavelength for magenta?




Clue: Magenta is a product of our minds, literally.

[Herra Tohtori]

Clue: Magenta is a product of our minds, literally.

Not really. It can be pretty well quantified, even if it isn't a spectral colour.

What's the wavelength of white?