What does comparatively matter in 2015?

This is based on my experience of what people are generally focused on. And what really, comparatively, matters.

First, a lot of people seem to place all their eggs in the basket of AI. Specifically, AI based on current computer architectures. Ethically, in terms of effective altruism, and in terms of raw efficiency. Even in terms of consciousness, for most people who care about consciousness.

This would be absolutely rational if you had a reason to think that a von Neumann computer architecture could solve the phenomenal binding problem. But since there is no reason to think it would, the effort seems fundamentally misguided. Not in intention, or in terms of correct effort. Understanding intelligence and consciousness are of prime importance. Indeed, far more important than what most human thoughts are directed at.

That said, this misguided effort will still be of tremendous value. By having people focus on computing in a utilitarian and philosophically motivated way, we will do a lot of progress in the area, faster. Which is absolutely great, in general. Being highly competent, as a civilization, in computing technologies is a great start! But it is not the end. At some critical point, people will realize that phenomenal binding itself has to be addressed. What is it? Why do brains solve it? Do brains solve it on their own?

A sufficient number of people are working on that already, that focusing on what will be the focus in the future is nonetheless what has the comparative importance. What is it? Well, consciousness studies themselves. Without assuming that your particular procedure of choice (which currently happens to be statistical learning by serial computers for most people who think about this) will produce consciousness. you can go ahead and study consciousness itself. But what do you study here? You can’t, so far, put consciousness in a petri dish and analyze its chemical structure.

What you can do, however, is to study the computational capabilities of consciousness. Thus, infer its computational role in the survival of species. By describing what consciousness is capable of in rigorous mathematical ways, you can reverse-engineer the reasons why natural selection recruited it as the medium for representing world-simulations.

More so, I predict that by reverse engineering consciousness’ computational capabilities, a wide variety of new applications will be discovered. We will find how to solve a variety of computationally challenging problems using conscious experience. The computational comparative advantage of consciousness, as a computing substrate, will be revealed. Such comparative advantage must exist to begin with, or natural selection would not have recruited consciousness to solve problems related to representing the environment.

How do we study this comparative advantage? 

Look, I don’t know. There could be many obvious ways of doing so that simply escape me right now. After all, not many persons are thinking along these lines (and thus, again, why starting to work on this is comparatively highly efficient). But I will start by making sense of visual experience. I think that by characterizing our visual pattern recognition abilities in computational ways we can start to make a headway in describing qualia’s computational capabilities. Of course visual experience is a very specific kind of qualia. It also happens to be one where a lot of functional properties can be mathematically captured. We represent a 3D Euclidian space. Not any space. And thus many geometrically grounded questions can be formulated. For example:

How fast do the interpretation about the visual scenery propagate?

To what extend do the local *interpretations* matter? And to what extend do global interpretations of a scene propagate downwards to the specifics? How fast are these transitions?

How do images that represent semantically distinguishable forms inhibit the existence of contradictory representations with semantically opposed qualities? At what speed do these inhibitions propagate in the visual field? (again, the question focuses on the experience rather than on the neural substrate or any other specific preferred or assumed substrate besides qualia).

All of these questions can be formulated in very rigorous ways. And then tested empirically with psychophysics. We can start to ask these questions and to find answers. And thus, to characterize the computational characteristics of consciousness as such.

This work will certainly be synergistic with the current approaches. No doubt. But to optimize the balance of minds on each part to have the most synergistic mix, more persons could shift to the study of consciousness as such. That’s why this is the current comparatively most important matter.

Why should we care about the computational characteristics of consciousness? 

Because there is a comparative advantage to consciousness as a computing substrate, there must be applications in which consciousness is the single most efficient computing medium! That should motivate any practical-minded person. But, what should be even unsaid since how obvious it is: Consciousness is also the very source of what is valuable! By understanding and becoming capable of engineering it, we could finally produce value itself directly. Any utilitarian should be extremely excited about this possibility.

Leave a Reply