I find it difficult to understand how our sensory apparatus can have any knowledge (or even "knowledge" if that is different from just knowledge) about anything at all. I would have thought it was people who know this or that, and it is with their sensory apparatus that people know whatever they know. The view that it is our sensory apparatus that knows (or "knows" if that is different) is something I find incomprehensible. Can you explain what you mean?
I found it difficult to accept myself, so I sympathize.
The thing that finally convinced me was ironically perhaps a computer program rather than a person.
Much of this description will be over simplified, as a full discussion would be too long to post here. I'm also going to forgo references for now, if you express interest after reading this I'll go to the effort to dig them up for you. I hope you'll forgive my laziness.
The problem that our visual system solves is extraordinarily complex. The primary input to the system is the eye of course, but we should set aside most of the eye's complexity for the moment as most of that organ's function involves focusing a clear image on the retina, along with some tracking information. Let us just say that vision begins with a projected image of light that has reflected from the environment.
This image is registered by structures in the eye, that react to the light by stimulating nerves at the back of the eye. The nerve bundle caries this information to the visual cortex where it is transformed into perceptions. Ultimately our conciseness receives several different types of information. We see colors, light and shadow of course. We see objects and entities. We receive information about trajectories, and rates of speed of objects, and creatures. and much more.
The information received in this way, which most intrigues me is the three dimensional "model" of the space we see. You know for instance how to navigate through a room, in part because you can plot a course using this information from your visual perception.
Think for a moment about how complex the process is to reach out and pick up a cup. Think consciously about how your arm must move, how each angle of your joints and the flexing and relaxing of the mussels which set these angles must occur, each at the right time with the right level of exertion.
It is very difficult to move this way intentionally, you will likely catch yourself cheating. The point is not only to illustrate the complication of the task, but also to show the effortlessness which which you do even more complex tasks all the time. Your brain is doing a tremendous number of very clever things behind your back
In order for your brain to accurately plan this motion it requires a model of the environment, as well as a model of itself (your sense of body awareness.) The question is how do we get such a model?
Consider the form of data your vision collects for a moment. The image projected on the retina contains what is essentially a pattern of colored splotches. It is not exactly an image in the usual sense because of peculiarities of the eye's function, but for out purpose here let us pretend it is like an image on the computer. That is, let us assume that your eye receives a pixelated image that it passes along to the brain.
From this image your brain must construct a model of reality that could have produced it. I don't want to get into issues of edge finding algorithms so you'll have to take my word for it that it is possible to calculate where lines and regions are in the image. This it turns out is the easy part
(And brain research has detected these sorts of basic processes happening within us.)
Transforming this information into a 3d structure turns out to be harder than we think. One major issue is in deciding what shape in the world corresponds to a particular arangement of lines. There are patterns of lines which are innately ambiguous you may have a set of lines which could be an indent or a bulge for example with equal evidence of both views. Mathematically it is impossible to determine which option is correct.
So how do we do it? Well it turns out that if you make certain assumptions you can get the right answer with high probability. The key is that these assumptions must have a very high level of correspondence with the way the world actually is.
One example is the assumption that a particular scene is evenly lit, this lets you use differences in shading on either side of a line to determine which way the fold goes. Another is that objects are continuous in texture and color, this can increase the accuracy in a similar way. there are many other rules like this.
A computer vision researcher (apologies for not remembering the name) discovered this while working to make a program that could decode images into 3d models. (this is not particularly new research) He faced with exactly the problems we've been discussing here. So he cheated
He equipped the system with rules based on valid assumptions about the world and was able to achieve a high level of accuracy. (note that in the research I'm referring here the images were 3d looking shaded line drawings, not photos)
This in itself could be dismissed as irrelevant to how our visual systems structure except for one very interesting discovery. When he fed his program images of optical illusions that we humans perceive as ambiguous, the program saw what we see! The state of it's model alternated between the two possible solutions just as ours perception does.
The key is not that the program was imperfect, but rather that it was imperfect in the same way that we are imperfect! To me this is profound evidence that we work in a similar way.
If so, then our visual apparatus has knowledge of the world in the form of rules about the behavior of light's interaction with mater. This knowledge is of course implicit most likely rather than explicit, and clearly we don't have conscious access to these calculations.
I apologize if this is disorganized, I'm not going to edit it.