The Metaphysical Movie Projector

Perhaps it could be said that the philosopher has only two tricks: seeing one where others see many, and seeing many where others see one. Anaxagoras is credited as the first thinker in the West to understand the moon, not as a god, but as a lump of rock that is illuminated by the sun’s light. His brilliance is thusly attributed to his cold realism: his conception of the moon as a phenomenon of matter and energy without the need for any divine externalities; but I wonder if there is another aspect of genius in this theory of the moon. His idea that the moon appears to us due to the sun’s light might be of more philosophical significance than is typically acknowledged. Was it obvious to the ancient Greeks that “appearance” was mediated by light? Was it obvious that “appearance” was mediated by any conduit at all? Let me elaborate. An ancient people, ignorant to philosophy and science, might naively believe that an object “appears” to them simply because it sits in front of them. To clarify my point, I will use the example of a tree. As a man walks up to a tree, it appears in his vision. Without delving into philosophical subtleties, the man might just assume that the appearance of the tree is a natural consequence of his proximity to it. The tree appears to him because it is “right there”. It is not immediately obvious that the appearance is contingent upon external factors: the presence of light and the proper functioning of the eyes. It is not immediately obvious that interference with these factors (fog, blurry eyes, etc) will warp, if not remove, the image of the tree. What one learns from philosophical contemplation is that the appearance of the tree is something distinct from the tree-in-itself. The appearance of the tree is therefore a process. It requires the tree-in-itself to not only exist, but to have its spatial data transmitted to the eyes and transduced into the strange language of the brain. Therefore, we have a conceptual bifurcation of the tree into object and image. Anaxagoras must have understood this bifurcation in order to construct his theory. Whether or not he is the originator of this concept of mediated appearance, I do not know; but in any case, it seems that he was extremely prescient for understanding it and incorporating it into cosmological discourse. In this essay, I would like to introduce a second bifurcation into this process; a bifurcation at the concept of the image.

The concept of appearance tell us that perception has two components: the thing-in-itself and the image of that thing. In a more information-theoretical sense, we could call these two components: the data and the interpretation; but if we return to the example of the tree, we notice that this two-pronged theory is not sufficient - for what exactly is the “data”? What happens when we see a tree is that light energy reflects off of the physical object and transfers data to our eyes. As such, there are really two types of data here: the physical space occupied by the tree and the light energy which contains a description of that physical space. We can understand this relationship as a data transduction - that is data being transferred from one format to another. The tree, in its occupation of physical space, can’t actually get to us. There is no way of sensing that a tree is occupying some arrangement of physical space directly if I am standing 10 feet away from it. For my eyes to actually sense the tree, the physical-space data must somehow be transmitted to them. This is the purpose of light. By means of light, physical-space data is tranduced into a moving format (photons) which can actually reach and interact with my eyes. Therefore, we need to add an additional component to the theory of appearance. We now have physical-data, transmitted-data, and interpretation.

But what exactly is interpretation? What I mean when I use the word in this context, is the visual experience of seeing a tree-image in front of me. There is a trap we can fall into here - similar to the trap that the ancients fell into when they assumed that appearance simply happened because of proximity to the object. We tend to assume that the human brain is what performs the interpretation of data into visual experience. Therefore, when considering the problem of the tree, one might try to end the conversation at the eyes. Once the light-data interacts with the eyes, then the data has been brought into the brain and the visual experience occurs; thus completing the process of appearance. Just as the ancients assumed that visual experience just happens when one has a direct line of sight to object, so we assume that the visual experience just happens when data reaches the brain. It is here, at the information entrance, into the brain, that I would like to suggest my bifurcation.

Consider a movie camera recording an image of a tree. This camera is performing an operation similar to what the eye is doing. The camera receives light data and interprets it into electronic data which describe the image. The problem here is that, unless one adopts a bizarre system of metaphysics, he does not believe that the camera actually has a visual experience of the tree in the same way that a human does. In fact, all the camera does is perform another data transduction. The camera receives data in the format of light and transduces it into the format of its underlying hardware: physical discoloration if the hardware is film, or magnetism on a disk if the format is a hard-drive.We can conceive of eyes as a camera using different underlying hardware. Our eyes receive data in the format of light, and transduce it into the format of neurological activity. The question then is: how does visual experience arise. Visual experience does not arise when the underlying hardware is film or hard-drive. Why are neurons special in this sense?

Computational neuroscience posits answer to this problem based on a theory of layered information processing. Allow me to provide a cursory overview of the theory for sake of argument: Let us first acknowledge that the neurons immediately behind the eye perceive color. A neuron (or small cluster of neurons) will exhibit a firing rate proportional to the color of a pixel of visual input. These neurons form a layer whose function is to mediate raw color data. This layer then projects its data into a deeper layer which mediates curves and edges. The neurons in this second layer have firing rates proportional to the curvature or edginess of a chunk of the visual stimuli. This layer projects into yet another deeper layer which mediates things like perception of something like full objects. The progressive increase in sophistication continues a few layers deeper. It is hypothesized to terminate in a region related to something like concepts. Interestingly, this hierarchical layering of data organization is broadly the same structure which underlies contemporary machine learning algorithms - which have recently demonstrated excellent capabilities in image recognition and interpretation. This congruence between functional neuroanatomy and information theory is adduced to prove the underlying as the basis of interpretation.

As Wittgenstein taught us, many philosophical debates stem from a misuse of language. The problem with this theory from computational neuroscience can be attributed to a misuse of the word interpretation. In their context, interpretation means something like conceptual understanding. Interpretation means that the visual data is processed with enough sophistication that complex behaviors can be coordinated. Driving is an illustrative example. The cacophonous mess of road signs, pedestrians, and cars must be, not only recognized, but parsed into a logical space such that gas-pedal and steering wheel behaviors can be coordinated.

But this is not at all the kind of interpretation that I am discussing. I am discussing interpretation in the sense of a perceived visual space. Understanding and perception are different things. To see this, let us take a counterexample to that of the driver - the blind man using a tool to walk. Imaging a blind man using a highly-sophisticated light-to-sound transduction system. This system ingests photon-data into a computer interface which translates it into air-vibration-data which is then emitted through headphones. The blind man can now parse light data by “hearing it”. The blind man perceives that a building is on his right by hearing a rumble. He avoids an accident with a cyclist because he perceived the man as a noise increasing in volume. The salient point here is that this man is now able to use light-data to navigate physical space - just as the driver of the car was in the original example; however, this man, by construction of the example, does not have any visual experience. The light-data, while still being utilized for navigation, is now realized as auditory experience instead of visual experience. This example demonstrates interpretation of light-data - that is the understanding of light data so that complex behaviors can be coordinated; but the example does not demonstrate Interpretation - the translation of light-data into a perceived visual experience. As such, this example serves to prove a distinction between the understanding of light-data (as encoding by the brain) and the perception of light-data transduced into a visual experience. This distinction is the bifurcation that I wish to introduce into the image.

At this point, we can make two assertions. 1). light-data is represented in the brain (transduced into neurological hardware). 2). visual experience is not an inherent property of the brain’s representation of light-data (because the brain can represent light-data in a way that is not visual - as shown above). The union of these two assertions leads to a fascinating consequence. The eyes and the visual cortex function to transduce light-data into neurological encoding. Yet our typical experience of the world is to perceive light-data as a visual experience. Therefore, there must be some mechanism performing a final transduction in which neurological representation is transduced into visual experience. I call this mechanism the metaphysical movie projector. Like a typical movie project, it ingests a representation of light-data and realizes it; but, in this case, we have a very special form of realization. Perhaps it is the only form of realization that is actually real.