Compiling to Reality

I haven't blogged much this week about Philosophical Foundations. Partly this is a function of outside time commitments, but it's also got something to do with a general lack of interest in some of the points being debated in class this week.

This week's readings had mostly to do with Functionalism and responses to it. On the whole, I find Functionalism convincing, and the responses to it not so much- which is probably why it's hard for me to get very worked up about the in-class debates.

But let me say a couple of things about the one on Monday. Near the end of class, things kind of degenerated because one person basically refused to be convinced that there is any need to study things at a level that abstracts away from the physical.

For what it's worth, I understand some of his concerns. There is something kind of chessy about Functionalism. One does get the feeling that Functionalists are refusing to fill in some important blanks. And indeed, this has mostly to do with the fact that they never seem to do much more than identify the necessity of some kind of "functional" level of explanation. So, in these terms, something like "belief" isn't readily explicable in terms of quarks. It would seem to be the kind of thing that can be instantiated in many possible ways. We don't expect everyone's brain to contain exactly the same neurons to exactly the same dimensions with exactly the same patterns of connections, and yet we can speak meaningfully about all these different brains having "beliefs." Whatever the common denominator is, it doesn't seem to be (entirely) physical; a better way to explain it might be in terms of what functions are performed. In other words, "belief" is a kind of program that can run on many different hardware configurations. And for the functionalists, these different hardware systems don't even have to be subtly different. They can be radically different: provided they all implement the same algorithm, we're good to talk about "belief."

Well fine, but what's said to be left out here is anything beyond this observation. Having decided that brain state algorithms are multiply realizable is all well and good, but it's still not clear what, exactly, it is that accounts for the similarities across individuals. Do we really want to commit to a view of the world that says that what is important about things like intelligence and so on are ghostly algorithmic laws that are, one presumes, irreducable and cannot be subjected to scientific experiments? What force is it that drives these laws? And, perhaps more importantly, how do we identify universals in these laws? How do we study them? Does it do us any scientific good to adopt this worldview?

As for the question of "what" the laws are and "what drives them," I don't really personally consider that to be very interesting. I am simply noting it here to demonstrate a certain amount of sympathy with people who don't buy into Functionalism. I don't honestly see why this should be any more of a puzzle than questions of why the physical laws of the universe are the way they are. No one really knows why it is the case, for example, that motion is transfered from one object to the next. We can spin theories about objects composed of atoms and these atoms emitting force fields that move the atoms in the other objects and so on, but at some level the answer still comes down to postulates. At some level the universe simply is constructed in such-and-such a way, and there is nothing more to be said about it. From a purely historical view, it would seem unlikely that the nascent Science of Mind has actually hit this kind of level - but I see no reason in principle to rule it out. It may be that nothing more can be said about functions as such except that the universe happens to be so constructed that, e.g., the Lambda Calculus accurately describes computation, or whatever.

As for the rest of it, I am a bit frustrated that the Functionalists seem satisfied saying that there is a functional level but don't want to probe much deeper as to what the nature of the functional level is, or if this knowledge can ever be put to good (maybe psychological) use, or what have you. Nevertheless, I do think the basic idea is sound.

Now, the debate in class had to do with whether the Functionalists ultimately want their theories to be physically realizable or not. Of course 98% of Functionalists are probably also some kind of materialist or another - and they do indeed believe that the physical component is indispensable for an object's existence in reality. We can talk all day about multiple realizations of chairs, for example, but there still has to be a physical chair present in order for anyone to sit in it, or see it, or clean it, etc. The concern from the student in question, though, was whether Functionalists were giving up too soon. At some level, if everything has to be physical, then shouldn't we just keep plugging ahead with science until we are indeed able to describe everything in terms of quarks?

And this is where I don't really have much to say in response. It seems completely obvious to me that we will never be able to describe everything in terms of quarks, even though quite probably everything in the universe ultimately is realized in terms of them (or whatever the new sub-sub-atomic particle of the week happens to be).

This can be seen from a number of perspectives:

• Informational - there is simply no way to represent such knowledge to ourselves. If we were to attempt to study, say, someone's brain on the molecular level and thus give a completely scientifically-deterministic account of that person's future actions and so on, we would need to specify a number of variables potentially greater than the number of extant particles in the universe. Not, of course, to list what's going on in the person's brain (for that we would only need the particles that actually are in the person's brain ... though of course delimiting the "brain" on the level of quarks is certainly an unsolvalble problem, I should think. This quark is part of the "skull," this one the "brain," this one the fluid near the skull, etc.??? Can't be done.) - but to list which arrangements of quark correspond to which functions, and also to know how these different states affect each other in regular ways, etc.


• Explanatory - even assuming we were capable of somehow storing all this information, it still wouldn't qualify as "science" I don't think. Science as I understand it is a process of generalization. Simply presenting the brain as is doesn't count. If the "explanation" is ultimately nothing more than noting the position of every quark, well then we've already got our explanation. We can simply point to the already-existing brain and be done. Naturally this isn't a very satisfying conclusion, though, because it's clear that it doesn't explain anything to just point to things as they are. We're interested in capturing general laws, regularities, etc. It's not clear what predictive power some kind of massive matrix representation of all the quarks that make up a person's brain would have. But it is clear that understanding things on a functional level has predicitve power (at least as regards the mind) - and that it is considerably more efficient than trying to store the entire quark matrix somewhere and perform operations on it.


• Feasibility - The idea of explaining everything at the physical level doesn't seem intuitively feasible anyway. Partly this is for the reason above - that it isn't always clear, say, which quarks are involved in "brain" and which are in "skull" and which are in "love" and which are in "cat" and so on. In fact, it would probably never be clear. Most of these quarks would be simultaneously participating in many such operations. Which sort of makes one wonder whether it would ever be feasible to write out all the "rules." Would we need rules that specified situations when the surrounding quarks were involved in "cat" or "breathe" while we're trying to determine whether a person is in love? Or maybe the person loves their cat. Etc. etc. Of course there's no way to know until we are at that level whether this problem can be dealt with (or if it even is a problem) - but until we get there, the idea of having to keep track not only of where all the quarks were and what they were doing, but whether they weren't also involved in other things and what that meant for what we're trying to determine? The data explosion is immense.

I like to think of it as compiling, in a sense. We program according to general principles. No one (or, rather, very very few people) programs at the level of bits. There was a time when people did literally flip switches and populate registers and call it a program, but with the levels of computation we expect from our machines today, it simple isn't possible. So what we do instead is have a program called a compiler that automates the process of transforming our higher-level principled programming languages into machine instructions, and those instructions tell the machine what to do with the bits. Opening up a computer and trying to fathom what it was doing by noting, at each processor cycle, which bits were set and which were off, would be a herculean task of epic proportions. It's not clear that it would even be possible. Hence the need for the compiler. Brain functions could easily be analogous. No doubt there are some important differences, but in some sense the functions that make up the thought process are compiled into thought primitives which are compiled into neurons which are compiled into chemicals which are compiled into atoms into particles into quarks STOP. Naturally we're stretching the term "compile" here (because there is no actually process - the instantiation is concurrent with the program - the same thing, ultimately) - but it's a better way of understanding it, I think, than trying to know the brain by looking at the quarks. We assume that the brain is ultimately realized in terms of quarks, but it cannot be understood on that level.

And that's really all I would have to say about that. Probably I'm missing endless amounts of subtlety, but it's one of those questions that seems ultimately useless. We are, in any case, not in a position to study the brain as a function of quarks, nor does it seem likely that we will ever be. And so I can't really get that excited about this debate.