Sunday, June 26, 2005

Science and Teleology, part V

Before I begin, I want to thank those who've had the patience to follow along. I am aware that this essay seems to have been going all over the place without appearing to approach anything like a conclusion. I am happy to say that we're almost there.

In the previous installment, I talked about learning as well as introducing the idea of certain biological constraints, both in terms of protein and neural space, which I called "budgets". I also argued that nature "prefers" to give creatures as much instinctual knowledge as possible and noted that learning, although common, tends to simply be filler for instinctual knowledge. This left of with the question of why some animals are born with large gaps in their knowledge of the world, which left us with the question of why this was the case. I ended with a question that suggested that the reason some animals more lack instinctual knowledge is because they aren't able to.

In order to understand this it is necessary to take one more (apparent) detour. In order to understand why some animals seem incapable of retaining more instinctual knowledge, we need to understand something about genetics.

Many people have the idea that our genes are like blueprints. It is supposed that if you could carefully examine a given person's genes, for instance, that you would, in principle, be able to precisely reconstruct her. This is a view that is reinforced by media articles that report the discovery of a gene for this or that (e.g., a "fat" gene) as well as popular fiction which has clones popping out of vats who not only look like the original but, in many cases, even has the memories of the individual! Even discounting pop media and pop fiction, one only has to look at a pair of identical twins in order to suppose that genes do, in fact, encode the exact biological details of one's body.

Looks can be deceiving, however. Identical twins do have a lot of similarities (the precise degree of similarity being the subject of some controversy) but a close examination will reveal differences. Twins do not, for instance, have matching fingerprints, retinal patterns or birthmarks. Likewise, if one could microscopically examine a pair of twin's circulatory system, lungs, etc, one would be able to catalog innumerable deviations from one another.

Rather than thinking of genes as blueprints a better analogy is to think of them as recipes.

Let us take a recipe for something like a cake. If you examine the recipe, you can probably get a good idea of what's going to be produced if you combine all of the ingredients and follow the instructions. You'll know that you'll have something sweet, fluffy and bread-like with a layer of even sweeter confection. If it's a fairly complicated cake you might have layers of food with different textures and tastes. However, if you were to extract a cubic centimeter of an actual cake and if you tried to map it back to a precise line of instruction, in most cases you'll find yourself frustrated.

When scientists say that a given gene is for a given trait, what they really mean is that the lack of the gene inhibits the trait from developing. It's like trying to reverse engineer a meal by selectively removing lines from a recipe book and seeing what the final product looks and tastes like.

If you've done any cooking, you know that the ingredients and the instructions are only part of the processes. There are any number of external factors that influence how a given dish will develop. Two different burners will likely produce different temperatures even when put on the same setting. Likewise, a frying pan with a copper bottom will distribute heat differently than a cast iron one. Whether I'm cooking at altitude or at sea-level will influence the final product.

The advantage of using recipes rather than blueprints is that recipes can withstand a fair amount of variation in these conditions. In fact, recipes are robust enough that they can often survive the omission of certain ingredients or the substitution or addition of other ingredients. True, the final product will taste different but, in many cases, it will be edible and, perhaps, even more enjoyable than the initial recipe. In fact, I suspect that one of the reasons that people have an intuitively difficult time accepting the idea that mutation is a factor in the Theory of Natural Selection is that they think of mutations as being changes to a blueprint. If I substitute one gear for another in a precision Swiss watch, it is unlikely that the watch will continue to function and almost wholly implausible that it will function better. On the other hand, if I substitute Swiss cheese for cheddar in my Denver omelet the end result will almost certainly be palatable and may well taste better.

Even when we cook a given recipe using the same equipment and conditions, we often find that cooking the same meal twice does not produce exactly the same result. Infinitesimal variations in cooking time, ingredient quality and quantities can make "identical" meals taste slightly different. In like manner, even if you do everything precisely the same, down to the millimeter, the microsecond, and the millionth of a degree centigrade, meals would still come out subtly different because the process of combining them is statistical. I can say, in very exact terms, stir the pot of sauce fifteen times counter-clockwise and such-and-such a speed with such-and-such a radius, but that will not produce an identical distribution of ingredients.

Genetics is a very, very indirect process. Enzymes (which are a kind of protein) interact with DNA to produce proteins (some of which are enzymes) which are used to build cells which (in certain cases) interact to form tissues which are "folded" into organs which a grouped into systems which, ultimately, comprise a body. It is not a bad analogy to think of DNA as being you body's "recipe", the proteins in your body as your "ingredients", your mother's wombs as being the oven that baked you, and your body as the meal that was produced (except that you continue to "cook" over the entire span of your life, a bit like an omelet still cooks for a few minutes after it has been removed from the stove).

Actually, DNA is more like a book of recipes than a single recipe. There are recipes for building lungs, for building skin and (foreshadowing alert!) building brains. Just as with real cookbooks, some recipes are more precise than others.

Consider a sandwich. Although I've made a point of distinguishing recipes from blueprints, making a sandwich is a lot more like following a blueprint than, say, blending a margarita. You could actually make a blueprint of a sandwich, rather than a recipe, and not lose anything essential (so long as certain components such as the mayonnaise or peanut butter had been prepared in advance). Some parts of our bodies are like that from a genetic standpoint.

When a body part needs to have a precise structure, our genes must contain enough information to specify its structure to the necessary degree of precision. That comes at a cost and that cost is informational. Genes are information. The more information you need to describe a given structure in a body, the longer the section of DNA you need to use to describe it. This is a finite resource.

You might wonder why an organism can't simply make longer DNA. The amount of DNA that an organism can have is limited by the size of its cells. Cellular sized among animals whose cells have a nucleus is limited by physical constraints. Each of your cells is essentially an non-independent organism. It eats, it metabolizes, and it excretes. It needs to do all of these operations through membranes (an inner and an outer set). Membranes constitute areas while the bulk of a cell's machinery occupies a volume. As you know, areas increase as a square while volumes increase as a cube. This if I have a cell with an area of 2 and a volume of two and I want to double its size, I now have a cell with an area of 4 and a volume of 8. This means that I have twice the area for doing such critical tasks as absorbing nutrients but that I am now supporting four times the volume of cellular machinery that needs to be fed. It is for this reason that all animals, from whales down to fleas, have approximately the same sized cells.

Much of the complexity of the body can be reduced to simple repetitive algorithms. Lungs, for instance, have millions of branching tubes. One might naïvely suppose that this means that millions of genetic instructions are required to "construct" lungs; however, the structure of lungs can be viewed as a simple set of instructions that says, "build a tube, then split the end of it in half and build two smaller tube; repeat until done."

There are different sort of cooks in the world. Some cooks slavishly follow a recipe exactly while others consider a recipe to be a general theme which is supposed to be improvised upon. The "recipes" in the DNA cookbook tell you (meaning the cellular machinery of your body) what sort of cook you need to be. Perhaps a bit confusingly, a single recipe can require different types of cooks for different parts of the recipe. Some elements of the circulatory system, such as the jugular vein, are not open to a great deal of improvisation while others, such as the capillaries, require extensive improvising.

Again, the determining factor is how much variation can be tolerated in a given part of the system. If you don't have an aorta, you're probably dead. As such, your genes need to provide enough information to ensure that there is going to be an aorta in your body. On the other hand, the precise layout of the capillaries in your big toe can withstand a lot of variation, so long as it gets enough blood, so your genes can get away with a fairly compressed set of "branch as appropriate" instructions.

Now let us consider the brain. As I've stated previously, many organisms don't have brains or have such miniscule brains that they might as well not have them (insects, for example, can live for extended periods after being decapitated, dying ultimately of starvation rather than for lack of their brains). In some respects brains really are like computers: if you don't have one and have gotten along fine, you probably don't need one, but once you get one and start to really use it, you quickly find that it becomes a necessity that you can't do with out.

Like computers, the more brainpower you have, the more advantage you get out of it. This is especially true if you find yourself in any sort of "arms race" with brainy competitor animals (including, potentially, members of your own species). However, like computers, the more powerful your brain, the more expensive it costs. In some cases, the expense is worth spending a sizable fraction of your protein budget on brain mass. However, also like a computer, a brain is as only good as its software and firmware. It doesn't do any good to have a beefy brain that doesn't have the necessary programming to cope with the problems of survival — you might as well be a tree!

As previously noted, from a survival perspective, instincts are good. The problem with learning is that if you have to learn something that is critical for your survival, you run the risk of either getting killed before you learn it or, just as bad, learning the wrong thing and getting killed by false knowledge. In this view, you want to have to learn as absolutely little as possible. It is better to be born wise than to have to attain enlightenment at a later date.

The problem with instincts is that the only way to pass instinctual information is through the genes. That means that every bit of knowledge you need to have must be encoded in your genes. That translates into physical space on the genes. The more complex your knowledge is, the more space is required. It is necessary to understand that when we are talking about instinctual knowledge, we are actually talking about physical structures in the brain. An instinct to be wary of predators is actually a three-dimensional labyrinth of neurons connected in a lattice that is structured to evoke a particular response (perhaps a release of adrenaline which, in turn, activates other neural clusters) when presented with certain stimuli. This means that in order to have that structure, the genes that control its development have to contain enough information to reliably "tell" the body how to create it. Some of that can be compressed into abstract recipe-like instructions that allow for compression but a fair amount will require more blueprint-like specifications.

Brains get complicated very fast. The human brain (which is a rather extraordinary example) has something like 10 billion neurons and, more importantly, 100 billion connections between those neurons. The human genome, by contrast, has about 100 thousand distinct genes.

An animal's genes can only specify so much information to allocate to a given brain. Once your brain grows past a certain size and complexity, your genes have no choice but to allow the body to improvise the actual structure. What this means is that your genes can't "anticipate" what brain you're actually going to end up with. They can pre-program in a certain amount of information but are helpless to program the rest.

In my last essay, a person with the handle of "Murky Thoughts" challenged my assertion that few animals need to learn how to walk. He noted that colts, by way of example, do walk on their first day but that they don't walk well. He also suggested that relatively few vertebrates are born knowing how to see (this is actually mainly true of avians and mammals, but regardless) and that they need to learn how do so. He's right; however, this is mainly due to the complexity of an animals overall nervous system rather than the complexity of its brain. The animals in question have sufficiently complex nervous systems that their genes can't "know" how, precisely, the nerves (from say their retinas) will connect to their brains. This inability to anticipate that requires a certain amount of learning, much as a rat needs to learn the maze of tunnels that it's born into. However, colts are born with the knowledge of how to walk once they figure out how their legs are connected to their brains. That information is inherited.

Human babies, by contrast, don't have the knowledge of how to walk buried in their neurology, waiting in anticipation of the time when their brains figure out which nerves connect to which muscles (and, or course, in anticipation of having muscles that are strong enough to do the job). Even after they work out their neural schemas, they still have to go through the tedium of actually learning how to walk. A fascinating contrast to this is the fact that infants are born knowing how to swim. If you drop a newborn into a tank of water (and under no circumstances am I suggestion that you do so!) she will not only know how to swim but will also know how to avoid breathing in water when she's submerged. Most curiously, they lose this ability within three days. In order to swim after that point, the child has to learn how.

In most animals, the genetic limitations of instinct translate into physical limits to the complexity of their brains. Even if they have the protein to spare towards the construction of bigger brains, doing so would be a waste of resources since they wouldn't have the necessary programming to take advantage of them and they lack the luxury of time and effort that would be required to develop programming for them after the fact. It takes a rather special set of circumstances to justify the complexity of a brain that doesn't come preinstalled with the majority of the software necessary to survive.

A handful of animals do have that luxury. This is the central feature that unifies lion cubs, elephant calves and human babies. Each of these creatures is born into a protective environment that allows them to have brains that are more complex than their genes. Not only do they have the safety of being protected at birth but they have the luxury to take their time in growing up.

So let us suppose we have that luxury. We have a turbo-charged brain which (by necessity) has blocks of computational space that aren't programmed from birth. How do our genes tell us to take advantage of it? After all, a surplus a free space doesn't do us any good if it just gets filled up with junk and noise. Even though our genes can't directly program it, it needs to be able to tell us how we ought to utilize it. This is accomplished through the insertion of meta-instincts.

Instead being born with the gift of knowledge, we are bequeathed with the knowledge of how to get knowledge. A kitten doesn't know how to catch mice, but it does have an instinctual desire to play with small, moving objects. Because the kitten can rely on its mother to feed it in the meanwhile, it can spend time developing those skills in a non-survival context. By the time the kitten reaches maturity, when it is critical for it to be able to feed itself, it has (we hope) learned everything that it needs to know to do so.

This is play. This is also the essential function of play.

There is, of course, more to the story. Let us consider the difference between calculators and computers. A calculator is a tool for performing mathematical manipulations. So, in essence, is a computer. The primary difference is that all of the functions of a calculator are preprogrammed. If your goal is simply to have a tool to perform those calculations, you really don't need anything more sophisticated and, indeed, for quite a long time people got along perfectly well with calculators. A computer, by contrast, isn't limited by its built in operations. You can build on those basic operations in a very flexible and open manner. The need for that flexibility is debatable but once you start to use it, it quickly becomes apparent that there are quite a few advantages to having that flexibility.

Although nature doesn't "like" having to rely on learning, once the requirement of flexible learning arises (and the necessary circumstances to support it) a vast amount of potential is unlocked. Being able to learn doesn't simply mean that you are forced to learn how to catch your dinner; it means that you have the potential of figuring out a better way to catch your dinner.

Of course that comes with its own set of risks. One of the ways to minimize that risk is to have a non-genetic channel for transmitting information. If you have the ability to compare notes with others (including peers as well as parents) you can minimize the risk of working from bad information. This is the root of culture, the rudiments of which we can see in such animals as chimps and (perhaps) dolphins.

Again, play is a useful meta-instinct; however, the instruction transforms from "figure out how to do X" to "figure out how those around you do X". An even more sophisticated version is when X is actually left as a free variable. Instead of a desire to interact with a specific purpose, our instinct becomes an abstract desire to simply interact with our peers and our parents with a generalized desire to learn from them without any particular goal.

It is instructive to watch human children at play. Some games seem very strongly connected to necessary skills. All children, everywhere, have games that involve the tossing and catching of objects. What is more interesting is that, past a certain age, children start to become obsessed with rules. In a certain sense, the rules become the game with a great deal of effort being spent on deciding what is "fair" and what is "cheating". My own feeling is that when you see children arguing about rules you are witnessing the development of our moral sense. I believe that it is on the playground, more than from our parents or from such institutions as church and school, that we learn (or fail to learn) how to be moral creatures.

At this point, we have our teleology. All that is left is for me to wrap this up into a conclusion. Since I've already been tardy twice in writing this, and since I expect the conclusion to be relatively short, I won't make you suffer another two week delay. The next installment will be next week.

No comments:

what is this?

Tell me when this blog is updated. . .