by Bor, Daniel
The main mechanism available is the proteins that genes create. Some proteins can interact with each other to follow the rules of logic in order to perform rudimentary calculations. Other proteins help sense details of the environment, while some even turn back to the DNA that created them, and turn on and off various genes, thus changing the production levels of other proteins. These new layers of information communication allow for a very complex cascade of activity, and surprisingly intelligent forms of learning and ideas.
One collective example of the generation of a complex concept are Hox genes, which control the location and number of developing limbs in animal embryos by deciding whether other genes are activated. Some of these controlled genes, one step down in the hierarchy, themselves regulate the activity of other sets of genes.8 This is highly reminiscent of many aspects of human life, such as the network of staff in a large corporation, or the many layers of categorization we mentally learn (for instance, my laptop is a kind of computer, which is a type of electronic device, which belongs to the set of machines, all of which are a form of tool, which are inanimate objects, and so on). Most complex systems benefit hugely from a hierarchy of knowledge and management, single cells included.
Far more impressive, though, is the facility for learning that microbes can demonstrate, usually via these protein-based computations. Bacteria, for instance, can communicate with each other using chemicals to indicate a lack of food, and thus each bacterium will spread out in a region to maximize consumption of what little food is available.
Protozoa and bacteria even use rudimentary forms of learning and memory when faced with different types of food or possible threats. For instance, if gut bacteria find some appropriate food, they will ready themselves to digest related food that’s likely to be nearby, as if making a kind of prediction, but will stop this behavior if they do not find it soon enough.
INTERNAL EVOLUTION
So evolution favors an accurate internal picture of the world via effective learning. But there are important limiting factors to this process. For one thing, as your internal model of the environment increases in accuracy, more energy is required to maintain this growing set of knowledge, and you become more vulnerable when food supplies fall short. And generating an increasingly large set of ideas requires an increasingly complex organism, so your reproductive rate slows down. However accurate your set of internal beliefs about the world are right now, the world can change catastrophically and instantaneously, and if you are sluggish at making copies, there’s little chance the critical DNA component of your ideas (if you have others, such as mental memories) will be able to update fast enough to track the changes, making extinction far more likely. Finally, if you have to become a larger, more complex organism to store all these extra ideas, then your bulkier biological machinery is also more likely to break down.
Bacteria hit that sweet spot of just enough complexity, but without it being an undue burden on survival. Consequently, they are capable of surprisingly clever information processing, but they are otherwise simple and small enough to replicate quickly and efficiently. They are the most successful type of creature on the planet by any yardstick you’d care to use: by numbers, because there are a staggering 1030 of them; by diversity, because they live not just on all continents and in all climates of the world, but also in acid, in radioactive waste, and deep in the earth’s crust; and even by longevity, as bacteria have been known to spring back to life after lying dormant for tens of thousands or even millions of years. Bacteria existed in vast quantities across the earth billions of years before animals turned up, and it’s very likely that they will still be around long after humans have perished. Based on this evidence, it seems highly plausible that there was an active trend, via evolution, from the origins of life onward, to favor those creatures that could process information most dynamically and accurately—but only up to the complexity of bacteria.
So why do animals exist in the first place? Part of the explanation is that they arose and succeeded by chance: Given sufficient evolutionary probing over sufficient time, with the right conditions, ever-increasing possible niches of survival will be explored, or, in other words, ever-increasing sets of biological ideas will be entertained. Animals are just one random set of strategies for survival. Of course, humans are a fascinating, wondrous example of what an organism can become, with our rich consciousness and deep intellect, but evolutionary success is a different matter. Having a brain such as ours, for instance, seems to lead to runaway processes that endanger our own existence—excessive CO2 emissions being one catastrophic example of a set of damaging products of our great collective consciousness.
Leaving these caveats aside, I now want to explore the details of this niche that animals exploit. Modern bacteria can form and adapt ideas immediately by encoding information not just in DNA, but further afield within the cell, mainly by using protein to represent additional ideas, or, even more powerfully, by building many computational links between DNA and proteins. Although ingenious, this system is also terribly limited, since only incredibly rudimentary information can be learned, moment to moment. So what else can be co-opted to manage even more information, if better computational power is a potential evolutionary niche that is to be exploited? With bacteria combining to represent ideas about food as one primitive example, the next logical step is to move beyond the confines of the cell wall.
We are now in the realm of multicellular organisms, with cells specialized for specific functions within the organism. Bundles of nerve cells making a brain are one route nature took, with the evolutionary “hypothesis” that learning and storing even more information on the fly would compensate somewhat for the greater investment of time and resources required to maintain this organ.
Some basic change in the world may take non-animals—including the cleverest bacteria—generations to encode via natural selection and DNA. But even the simplest of animals can, strikingly, learn a wide range of lessons from the environment over just a few seconds. Other more complex features of the world, assimilated easily by animals, may never be captured by DNA alone. In this way, a threat that would have destroyed a non-animal organism, or even a whole non-animal species, because of its limited capacity to process information, might not even harm an animal.
If you view evolution essentially as the competition between ideas, with the best ones eventually claiming victory, then animals are in a sense clamping on an additional, internalized version of evolution in order to enhance their chances for survival.
Thus all life undergoes genetic hypothesis-testing via evolution. The feedback about whether your concepts are right or wrong usually comes from the environment directly, which selects those concepts for persistence across the generations, and the bad concepts for death. If you happen to be a sophisticated type of bacteria, then a small but vital component of the feedback you receive can come from the intermediate steps of proteins, which help sense and adapt to very crude features about the world on the fly.
But for animals there is an additional buffer to process an important subset of beliefs that really matter for survival and reproduction. Feedback still comes from the environment, but much of that feedback need not affect DNA at all, since it can merely change the ideas stored in brain cells. And, in combination with movement, animals can now actually interact in pointed ways with the world in order to test beliefs very actively. The number of possible ideas an animal can entertain in a lifetime is effectively infinite, especially since wrong ideas no longer risk death.
Moreover, the more mentally complex the animal, the more elaborate its internal model of the world is. Thus, much of the environmental feedback that used to be required to change a belief, whether genetically or neurally stored, can now occur entirely within the complex, structured, internal environment of the animal’s brain.
Animals with particularly complex brains could even test many competing ideas without moving a muscle. For instance, in the middle of the night, unable t
o turn off my consciousness sufficiently to fall asleep, because I’m obsessively thinking about consciousness science, I feel a sharp hunger pang and conclude that the best course of action is to obtain a very large bag of cashew nuts. I initially decide to visit the kitchen, but then recall that a now rather irritating spring-cleaning the previous day cleared out most of the food. I then imagine the usual situation of going to the supermarket, but realize that my standard one is closed after 9 p.m. So I either could go to the 24-hour supermarket, which is a 15-minute drive away, or a gas station a kilometer away. I can work out the optimal way to obtain a much-needed, intensely fattening snack, potentially from miles away, without ever leaving my bed, which is in some ways incredible.
This illustrates that evolution has begotten a form of internal evolution, and this internal evolution becomes ever more apparent the more intelligent the animals are, to the extent that we humans have brains that very much behave like internal evolutionary worlds.9 We represent the world so fully, so accurately, that we can play out scenarios in our heads and explore a large range of options—all while hardly expending any physical effort. Such experimentation is now as safe as it’s possible to be—we don’t risk anything whatsoever by searching through the options in the mental realm—not survival by genetically betting on a loser, not even physical damage by learning painfully from our mistakes. This seems a universe away from the proto-life “ideas” with which we began this chapter, but it’s not. It is merely a sequence of connected evolutionary steps, all based on the theory that effective information processing naturally confers an advantage.
THE COMPUTATIONAL LANDSCAPE OF A BRAIN
Given the last few paragraphs, I should emphasize again that animals are not necessarily superior to other organisms in terms of evolutionary success. An oak tree, for instance, with its working hypotheses that physical toughness is highly protective and that the sun is a plentiful source of energy, may be just as long-lived and populous a species as a mouse. It’s just that animals have a fascinating, powerfully pointed set of advantages converging on complex information processing, and the overall genetic assumption of these organisms is that these few profoundly superior traits outweigh the many limitations.
The simplest benefit an animal gains from a nervous system is the regulation of its basic states (homeostasis). This computer in the head is able, in almost every animal, to help control important biological features. By monitoring the animal’s internal temperature, for example, and initiating any needed response when the animal becomes too hot or too cold, it can keep this temperature within the optimal range, acting just like a thermostat. Even in cold-blooded animals, where temperature is regulated entirely by the ambient heat, this can be achieved by directing the animal to move into a hotter, less shaded region if it is too cold. This single regulatory process provides a powerful advantage for many animals over non-animals—namely, that they can exist in a wide variety of locations, and may not need to shut down for winter. When you have a computer around, you can also fine-tune many other features, such as how much energy (glucose) or water there is in the blood, the concentration of salt, and so on. In each case, there is a monitoring system and, if necessary, a chemical messenger (a hormone, usually) that cause a change to correct any form of imbalance.
But this internal regulation is only a tiny portion of what an animal brain does. The main purpose of a brain is to sense the outside world and move around based on this data. Retrieving accurate external information confers a significant potential survival advantage. Although some bacteria, with incredibly crude sensory skills, can detect via protein switches when food is scarce, which is indeed very clever, such an organism would look remarkably stupid if food was actually at the center of its world, but undetectable by it just because it was hidden by a chemical barrier. An animal with multiple senses might be able to look for food, immediately see exactly where it was, what it was, whether other animals were feasting on it, smell how energy rich it was, and hear if there were any predators nearby waiting to pounce as soon as the animal tucked in.
We take our senses so much for granted, and rarely perceive them for what they really are. Our senses are nothing more than conduits to pick up physical information about the environment: a small portion of the electromagnetic spectrum for vision, for instance, which almost everything around us reflects or even emits; the compression waves of air or water for sound, which many moving things generate; and chemical offshoots of interesting objects for smell. Our different senses feel utterly distinct to us, but there is an important bottom line here: It’s all just information.
One demonstration of this comes from experiments on ferrets. If you rewire the ferret visual pathway from birth, so that instead of going to the visual cortex it ends up in the auditory cortex, then the auditory cortex, which should be specialized for hearing, ends up doing a pretty good job of processing vision, allowing the ferret to see. This auditory region will even take on characteristics (such as having neurons specialized for representing the angle of an object) that the visual cortex would otherwise process. It isn’t quite as good as if the visual cortex were doing the job, but it is quite functional. Another striking example comes from people blind from birth, who, when reading Braille, process the words mainly in the visual regions of their brain. Given that the visual regions have no sight-based input with which to work, this portion of cortex has adapted to take on the Braille-reading task instead. These sorts of examples show that all sensory processing is just information to the brain, and that almost any brain region can process any type of information, even if it was originally earmarked in development for a specific type of data.
Importantly, especially in more complex nervous systems, our perception of the world is far from being a mere copy of the physical information hitting our senses. Instead, an active, ever-changing, yet unconscious statistical machine is in force, transforming the basic information we receive into a detailed model of the immediate world, including how it is likely to change in the near future, and what in the environment is particularly relevant to us.
But there’s no point just filling up your sensory bank with information, however well processed, if you’re just going to hoard it, never using that hard-earned knowledge for useful purchases. What’s needed is a way to tie information to behavior—in other words, to move. In simple animals, this is all too clear, as the connection between what they perceive and how they act is usually direct and almost immediate: A worm, sensing food, will go toward it, or sensing some threat, will move away. But the more sophisticated an animal, the more processing takes place in the gap between senses and movement.
For humans, the brain’s main evolutionary purpose—moving the body—is rather obscured by just how much happens inside; the link between what we sense and how we behave is a long, fragile tangle of thought. The processing and the calculation of what action is best out of the millions available to our finely tuned bodies and remarkably accurate world picture has taken center stage. Nevertheless, it’s useful to entertain the interesting perspective that, essentially, even a human brain is there primarily to move the body around in the most useful ways.
The first mechanism for movement is instinctive behavior, a tool that all animals possess. An instinct is a genetically determined brain program to marry some sensory input with some prescribed response—designed, as always, to maximize the survival and reproduction of the animal. Here the genes are basically exploiting the ability of the brain to store and act upon information too complex to be executed only in a direct genetic form. For instance, if I unwittingly touch a delicious pie that has just been removed from the oven, before I know what has happened, I find my hand darting away from the potentially burning source. It takes a moment or two for my consciousness to catch up and realize that I nearly burned myself. The primitive regions of my brain sensed the heat and programmed the response before my higher cortical regions were even informed, either of the heat or the arm twitch. It all looks very simple a
nd natural, especially when consciousness appears to be merely a spectator at the event, but you still need some surprisingly complex neuronal processing to make this important reflex occur—you need to know in which direction to send the arm, which muscles to trigger, by how much, and so on.
Matching responses to fixed sensory input is all very well, but on its own, it doesn’t get you all that far. In one episode of The Simpsons, Lisa’s original school science fair exhibit is destroyed by her mischievous brother Bart. In revenge, she decides that Bart himself will be an unwitting component of her new exhibit, so she carries out a series of tests to determine whether Bart or a hamster is smarter. In one experiment, the hamster starts nibbling at a piece of food, but receives a mild shock. It immediately learns never to touch the food again. Cut to Bart, who discovers a cupcake with electrodes attached and a sign in front boldly warning, “DO NOT TOUCH.” Chucking the sign behind him nonchalantly, he reaches for the cupcake, and his hand darts away as he receives a sharp electric shock. But this just makes him angry, so he reaches again. Surprise surprise, he’s shocked again and his hand darts back. He continues repeatedly to reach for the food and repeatedly shocks himself. In the real world, I don’t think Bart would have made it to the age of ten. But what he—and the hamster—clearly illustrate is that it’s all very well having instincts, but without even the most basic kind of learning, they don’t get you very far.
Even the simplest form of learning is usually surprisingly powerful. If an animal tries to eat a food source that is toxic, it will quickly not only link this input with the result, but also avoid anything similar in the future. This is an incredible feat. The animal has managed to link in memory a useful, abstract copy of the object with its effect (more intelligent animals will add to this mix the relative intensity of the effect—for instance, chocolate is more tasty than spinach). What is more amazing is how few brain cells you need to learn in relatively sophisticated ways like this. Returning to the humble nematode worm (specifically C. elegans): This tiny worm, all of 1 mm long, has exactly 302 neurons. Nevertheless, it can learn to connect an arbitrary, neutral smell with a nearby food source and approach the smell whenever it is presented, presumably in the assumption that food is soon to follow. It can learn to stop moving away if some initially potentially dangerous stimulus repeatedly shows itself to be harmless, thus scrubbing out a previously firm, important belief. The nematode worm even shows crude seeds of socialization, with some strains only stopping to eat when in a group.
