Adam writes this week about his difficulties reconciling some philosophical issues that are dear to my heart: free will vs. determinism and the possibility of strong AI. I must (guiltily) admit that I do so enjoy observing thinkers puzzle over these issues, even though Adam's frustration is apparent in his writing. While I doubt I will be unable to resolve his conflict, it sure is fun to try.
Both the question of free will and the question of strong AI share a common uncertainty: the lack of a comprehensive understanding of the brain. For some time, I considered this absence to suggest that such an understanding could never be reached. I mean, look around at civilization. In this country we can communicate wirelessly using cell phones, travel across the world in a day, all for the price of a few days work...if *by now* we haven't figured out the brain, an organ which sits inside the head of every living mammal on the planet, it is pretty hard to imagine waking up one morning, opening the newspaper and finding the problem solved.
As unlikely as it may seem that science is going to solve this problem, I believe it will, and this is why. When I look at the "big picture", the story seems less bleak. I look backwards and see the country that I live in is only 200 years old but recorded time stretches back thousands of years. I look and I see scientific fields that within 100 years were revolutionized with discoveries like relativity, DNA, and the transistor. Using history as a guide, I look into the future and I see neuroscience coming into its own, and the promise of technologies like fMRI being realized. I have come to believe that before declaring the brain indecipherable, I should allow the infant brain sciences (which are already yielding exciting discoveries) a chance to come into their own.
This is important because if you are like me, convinced that major advances in our understanding of brains is bound to happen, then impossibilities like the one's Adam is grappling with are themselves an impossibility. Adam believes that determinisim and conscious choice are at irreconcilable odds. If they are truely, for all time, irreconcilable, then the scientific understanding of the brain must too be impossible, because surely if we fully understood how the brain works, we'd be able to reconcile the viewpoints. The same problem applies to Adam's lament over the existence of strong AI given what we know about modern computers. Unless we are willing to give up understanding the brain, it is overly pessimistic to declare that no headway is possible on these issues.
So far, I have simply established an argument for why I believe these issues will someday be reconciled, and therefore why it is wrong to use the term "impossible". Without this argument, there would be no modern motivation to try and resolve these questions. Now I'm going to really wade into the deep end of the pool and offer my sketchy theories on how the explanations of these two connundrums are going to bear themselves out.
Determinism vs. free will. This dichotomy has serious definitional problems. One dictionary says that determinism means "The philosophical doctrine that every state of affairs, including every human event, act, and decision is the inevitable consequence of antecedent states of affairs." I have thought in the past that determinism means that the universe is predictable, and that if the universe was not deterministic, that it wasn't predictable. First of all, this definition does NOT say, or suggest, that just because a state is the consequence of an antecedent state, it is somehow easy to *predict* how that transition is going to occur. Let's detach the notion of "bound by cause and effect" from "predictable by modern means". For example, we believe protein folding to be governed by cause and effect, yet it is not efficiently predictable by modern means. The protein appears to act deterministically, yet we still have difficulties predicting how its causes lead to its effects.
What about free will? Do we really think that it means that we can make choices *despite the physical reality occurring in our brains*? That would be absurd. An explanation that satisfies me is our brains are complex systems. The complexity of complex systems comes not from what their underlying particles are doing, but from the way in which those particles are organized. Like other complex systems, the brain obeys physical laws and obeys cause and effect, but, as with the protein, it is not yet predictable by modern means.
One aspect that bothers people when contemplating determinism is the fear that we are "bound by fate" to a certain course of action. I think this is a pretty grim way of looking at a reality that can only be true in the most abstract of ways. Let's look at another complex system: the stock market. If the universe is governed by cause and effect, is it "bound by fate" as well? Theoretically, sure...but practically speaking, that fact has not helped us yet predict what is going to happen on a given day the market is open. Being bound by an unknowable fate is practically equivalent to being unpredictable.
Another worry is over a diminshed sense of freedom, which goes something like: "I feel like I'm making my own decisions...doesn't determinism mean that I am not". But just because the brain is governed by cause and effect at the particle level does not preclude humans from being able to make their own choices at the mental level. There are two key realizations to make here. One is that the part you call "you" is part of a system, therefore to think of "you" being pitted against determinism is foolish... all that you consider to be "you" is the result of deterministic energy transfers between neurons in your brain. The other key realization is that the part of the brain with which you are using when "you" make a decision is perfectly able to affect other systems inside your brain using a deliberative process, causing a choice to be made. "Your" ability to "make choices", or to cause changes in your own brain state is a function of a system, and does not impact the fact that the system is causally determined.
A different way this worry is phrased is that accepting determinism implies we are "unable to make any decision other than the one we have made". This is a nonsensical statement. What kind of non-deterministic universe are we left with if we accept the implicit definition of determinism in this statement? Taken to its logical conclusion, this means that in a non-deterministic universe, if time were rewound and replayed, you could and would be able to make a different choice given exactly the same stimuli. But does anyone actually believe this? All your molecules would be moving exactly as they did last time. Where would the new information come from that would guide your thoughts differently? Is this really the only kind of free will we are willing to accept? Since we don't get much opportunity to experiment with rewinding time, it doesn't seem like we are missing out on much by accepting a deterministic view of the universe.
Strong AI vs. the apparent lifelessness of computer code. Adam's argument again seems to focus on his discomfort equating the determinism of computers with the liveliness of humans. I think that most of what I have already said about determinism applies here. I would add here though, that I am still holding out the possibility that Turing-machine based computers are not sufficient or at least very poorly designed to be able to handle the kind of informational transformations that the brain performs. When we end up building a model of the brain in a computer, the hardware may have to be significantly different. Basically we are going to need hardware that is capable of modelling complex systems really well. Wolfram alludes to this shift in modelling in his book, A New Kind Of Science. Given that the problem appears to require a more distributed notion of computation than we can support even with our modern parallel architectures, it seems like we still have some ways to go before we get there. If you allow for these different kinds of computers to be invented before Strong AI to be accomplished, it seems easier (for me) to accept that the goal can be accomplished.
Whew. If you are still reading, thanks! This was as concise as I could be on this issue!
Complex Systems are a topic on my mind rather frequently these days. In particular, I worry that we do not have very good ways of understanding them yet. In response, I propose the following methodology for understanding complex systems:
One way to understand an unfamiliar system is to use reductionism. When trying to understand a tree, we might examine it and identify its constituent parts. We may describe these parts (say leaves) in terms of functions that they perform (perform photosynthesis). Consequently, we may build up an understanding of a tree from a description of how its parts contribute to the whole (the leaves provide the tree with energy). From there we can try to understand how it manages to survive, grow, etc.
What I am now going to propose is an alternative to this methodology which I believe is better suited for the understanding of complex systems, which, by their nature, oppose understanding by simple reductionism.
Let us look for a moment at a methodology employed by political scientists when trying to understand the complex system that is a political landscape. Frequently, the political landscape is analyzed by looking at many different "aspects" and trying to find correlations between them. For example, in order to understand the actions of a president, political scientists may look at the state of the economy, the party affiliations of Congress, the president's own agenda, and the pressures from international forces. Each aspect is examined in its own right, but also with relation to the whole. Political scientists try to connect the dots between these various aspects in order to model the way in which the President will react under different circumstances.
This kind of work is more appropriately labelled "synthesis" rather than "analysis", as it requires the integration of disparate sources of information, rather than the finer division of a single source. It is also important to note that at no time do political scientists believe that any one of these aspects by themselves represent "the whole truth".
The methodology that I propose is inspired by this form of understanding through synthesis. What I attempt to do below is to describe this methodology more abstractly and formally.
Imagine that we have a small classroom of people who have circled their desks in a room. In the center of the room, we place the complex system. The task is to develop a computational model of the system which predicts how it changes over time. At first, the complex system is hidden from view of the students, obscured by a covering. A set of tinted, transparent panels are handed out to each of the students. Each panel has a stand which allows it to be placed upright on each desk. The panels are large enough so that the system can only be viewed through them.
When the covering is removed from the system, each of the panels displays a different view of the system. However, the different views are not simply based on orientation relative to the system. What the students see in each panel is only partially understandable. As they are allowed to compare different panels, they discover that some panels show the system in ways that appear as if they could not be coming from the same system. Stranger still, when students attempt to look at the system directly, without a panel, they discover that they cannot see anything at all.
In order to solve their task, the students discover that they must sit and study the panels both individually and as a whole. Due the vast differences in the panels, each student is forced to come up with their own description of what kinds of things occur in their panel, but must simultaneously ensure that their descriptions are flexible enough to enable collaboration and gain a sense of the system as a whole.
By now, the relationship of this thought experiment to the way political scientists work is clear. Each panel represents the different aspects of the political landscape which must be united to form a coherent picture of the whole. What I have outlined here is a general approach to studying complex systems by way of multiple viewpoints. What I hope to prove is that this methodology is crucial to understanding complex systems in the future.
I would go one step further, however, to say that I believe that this methodology is also a way of understanding how our brains look out at the world. If we turn our thought experiment inside out and put the students in the center of the circle, and the system on the outside, we could look at the panels as the different senses. What our brains must do is both understand what information a single sense is providing, but also understand what information it is conveying about the world as a whole. Looking at the brain this way, from the outside in, we might describe its structures as enabling the most beneficial cross-talk between the different senses.
Hopefully I'll be able to come up with an even more rigorous way of defining this methodology, which is beginning to look like a system in itself.
Man, are intellectual property rights ever broken in America today.
I bought a CD by the new band Velvet Revolver called Contraband.
I listened to it on the car stereo just fine. Then today at work I tried to listen to it on my computer.
Turns out that I couldn't listen to it without installing some software and accepting a license agreement! And of course, at work I am prohibited from installing any software on my local computer myself. So I couldn't listen to it at work :(
Fortunately I found this. Now my album really is Contraband :)
The most annoying part is that there was actually a program running on my computer, which is apparently part of Microsoft's Digital Rights Management plan, whose only purpose was to garble the music on my CD until I accepted whatever agreement I was supposed to accept! Why would I want to have a program like that on my computer?! Fortunately it is relatively easy to disable.. but damn!
Supposedly this was indicated on the cover of the album, but I sure didn't notice. Apparently this is the most major copy protected album to date.
Next time you go buy a CD, look carefully to make sure you aren't falling into the same trap.