by Bor, Daniel
Now, rapidly, there will be a thinning out of possibilities—all inefficient non-life replicators will lose the race to the resources and disappear, and the superior chemical copiers will dominate. The new battle is between these thoroughbred survivors. The active fight for energy and chemicals, even at this early pre-life stage, is an evolutionary process, because the main ingredients are already present: a vibrant competition for limited resources, on a superficial level, between different forms of chemical objects—and, more essentially, between different “ideas” about how to maintain one’s shape and make copies—which the chemical details of these proto-creatures encapsulate.
For instance, it’s a “bad idea” to be a great replicator dependent on potassium abundance when there’s usually none of it nearby. There’s no point requiring sunlight to maintain your shape when your habitat is normally in pitch darkness. There’s also, more generally, no point having a chemical makeup that requires huge quantities of energy to replicate when energy is sparse, especially if rivals are around. Success as measured by a burgeoning population in the primordial soup is predicated on maintaining a chemical composition that reflects or tracks the environment most closely—what resources are readily available, what’s the best way to extract energy from the local world, what environmental changes are likely to occur that may threaten many chemical reactions and that one may need to be protected against, and so on.
(At this stage, I should stress, I’m not assuming any consciousness whatsoever in any organism except for humans—terms like “beliefs” and “ideas” are meant as a kind of shorthand to describe creatures that internally represent a certain informational perspective about the world, but without any requirements for awareness of those representations.)
In this pre-life arms race, these close analogies to micro-beliefs about the environment, as stored in the shape of a chemical self-replicating object, are critical for survival. So it’s natural to assume that those objects that somehow represent the world more accurately, with greater detail, will carry an advantage. Indeed, the key reason that life might have evolved from simpler non-life equivalents is that non-life could not have developed the complexity of physical structure, or, very closely related to this, the extent of information storage, that organic life as we know it easily can.
Let me illustrate with a schematic example. Imagine there are three primordial copying objects, Alice, Beth, and Claire, all close to an active volcanic vent. Alice has stored the information that this precise location equals resources ad infinitum (perhaps by a strong chemical bond to the rock wall). When this particular volcanic vent becomes dormant, she degrades; she doesn’t make a single copy. Beth’s chemical components, instead, hold within them the “idea” that resources can potentially be found in multiple locations (perhaps by a chemical bond to the rock wall that weakens without sufficient heat, but strengthens again when another heated rock is chanced upon). When this particular vent becomes dormant, the lack of heat means she detaches herself from the rock and floats randomly until she’s jostled against another hot rock, which allows for a chemical reaction to bond her to the rock surface. She is again close to heat and other useful resources, allowing her to make some copies of herself. But when this rock, too, becomes dormant, and there is no other vent nearby, she degrades. In a sense, Beth’s structure is molding itself more closely than Alice’s to the external data concerning where resources can be found—instead of the chemical equivalent of a belief that Alice holds that “this location is all that matters,” Beth’s concept is that any hot rock will do. Claire has a physical structure that reflects the information that heat equals resources, regardless of location (by chemically sensing and gravitating toward the nearest heat source—behavior probably too sophisticated for non-life). So Claire has a chemical form that most accurately shapes itself to the information about her requirements for heat energy, as well as how in the world to find this, and this gives her—and her similar offspring—a distinct advantage. She follows Beth to the second vent, but when this vent fails and Beth degrades, Claire directionally moves toward the next nearest heat source. Over time, in response to these dangerously intermittent vents, Claire-forms will be the only population that survives.
LIVING ON THE EDGE OF CHAOS
While Claire’s more sophisticated, accurate “idea” would have caused her to be the dominant pseudo-life creature in her world, an even more successful way of responding to a changing environment is to update your ideas about it. Making true copies of yourself is important, but with such a dynamic world, where superior rivals or new dangers might emerge at any moment, being too fixed in your representations is dangerous. In this situation, exact copies of the originally superior chemical look doomed by their antiquated inflexibility. So some mechanism that can actually inject new creative ideas—in other words, that can “learn”—could potentially be very advantageous.
At this primordial stage, on the cusp of life, changing “beliefs” simply means making nonidentical copies. In other words, a family of proto-creatures needs to maintain a healthy balance between keeping useful knowledge and accepting that their world-picture could be better; they want their offspring to be faithful copies of themselves, but not too faithful. This loosening of the fidelity of the information is potentially expensive, because by chance many offspring will be inferior, perhaps just disintegrating at birth, or in other ways missing some vital chemical detail that enhances the chances of survival or replication. But it also raises the opportunity for some of the next generation to be an improvement on the model.
This tension between maintaining beliefs and injecting new ideas is a profound issue for any complex information-processing system, be it proto–life-forms, the neural interactions in our brains, or the scientific enterprise as a whole. Usually, though, a Goldilocks middle state, with chaos on one side and utter stability on the other, is the optimal way for any system to process information, and especially to learn useful new details about the world. This semi-chaotic activity is found whenever efficient information processing is required. It is probably the default state for networks of neurons, and it is one explanation for how complex thoughts in the human mind emerge from neuronal chatter.
A similar optimal balance between order and chaos exists in the scientific enterprise. There are cases, particularly in the softer sciences, where a prominent professor with a large ego—and a history of drawing in a large amount of grant money based on his well-established ideas—will do all he can to maintain these theories, including engaging in practices that are essentially unscientific and dogmatic. He may bend the rules to publish papers confirming his results, ignore experiments his lab carries out that contradict them, insist that his lab tow the party line, that those working under him always believe in his theory absolutely, and so on. He and his scientific progeny, his PhD students and postdoctoral assistants, may well be maintaining this viewpoint in the face of increasing evidence opposing it. For a while, due to his influence and personality, his theory may continue to flourish, but eventually it will be superseded, and his research staff will find it increasingly difficult to grab decent academic posts because of their long-standing defense of a scientific position shown to be wrong.
In a separate category are scientists who constantly generate outlandish ideas but are not particularly interested in testing them with carefully controlled experiments. Admittedly these rarely get past the PhD stage, but if they do, their careers always seem hampered by their overactive creativity.
The best scientists not only have the most respectable careers but also leave a lasting legacy of work, along with a new set of high flyers, who were former students. These renowned scientists are skilled at establishing successful theories and empirical results. But they are also quick to ditch these theories when the evidence racks up against them. They then generate new ways of perceiving the field—always with a qualified creativity.
The ability to settle on this healthy balance between stability and chaos
is probably too much to ask of pre-life creatures, except for the most advanced—those on the cusp of life—because they would lack the complexity to support it. Specifically, for effective, flexible information processing skills related to survival and replication, you first need a means of storing many solid preexisting beliefs, which DNA, as I will discuss in the next section, is supreme at doing. You then need techniques for testing new hypotheses about the environment. In life, the main method for this involves creating a host of successful offspring subtly different from yourself, with a small proportion of those differences potentially being an improvement, reflecting useful novel innovations.
Let me illustrate the relationship between complexity and adaptability with another schematic example. Imagine you have 5 different words (analogous to different kinds of atoms within a proto-life object) by which to make up a sentence 5 words in length (analogous to a replicating chemical creature made up of 5 atoms). In each case, the sentence of 5 words gives you very little information. However, there are 3,125 possible different sentences you can make. This is a reasonable number, but in the face of an incredibly dynamic world, it is still potentially very limiting. Now imagine you still have 5 different words, but you can make up a sentence 100 words long (like a replicating chemical with 100 atoms in it). Each 100-word sentence potentially carries 20 times more information than was represented by the simpler creature with sentences of 5 words. A far more striking feature, though, is that, instead of 3,125 possible different sentences, there are now 8 × 1069! Therefore, if the capacity to represent a greater variety of ideas is beneficial, the chemical object needs to be larger and more complex. Some chemical designs of equivalent size will be better than others at storing information and getting the balance of stability and flexibility correct. The specifics of the design, along with complexity itself, will provide further hooks for evolution to clasp onto.
Once a certain complexity was reached, the emergence of life itself might have been rapid, explosive, and almost inevitable. Candidate life-forms, emerging into a mode of effective learning, would have carried an overwhelming advantage over their simpler, less flexible rivals. These thoroughbred proto-life knowledge trackers would have been able to adapt, becoming ruthless at exploiting available resources and forcing all the more stable, less flexible alternatives to turn to dust.
Reaching such thresholds, and shifting into higher gear as a result of them, also happens in other contexts. For roughly 99.5 percent of the time that humans have existed, for instance, little scientific progress was made. But over the past four hundred years, with aids such as the printing press, education, and a critical mass of people seriously interested in science, actively discussing theories, and recording evidence, collective human learning—and scientific discovery—have dramatically increased.
WETWARE
At some point, in small, simple steps, basic proto-life objects probably evolved into early life-forms made up of RNA, which is a close cousin of DNA. Compared to any natural non-life alternative we know of, RNA is an exceptionally efficient and flexible information carrier.
How does RNA achieve this? Like DNA, RNA is a long string of connected components (known as bases) of four different flavors, or letters. A “triplet” sequence of three letters is an important combination—it is the way that RNA letters spell words—in DNA/RNA language, all words are three letters long. Each word represents one of the twenty or so amino acids, which are cellular building materials whose combinations form proteins. And proteins are essential for almost all functions of every cell of any organism on the planet. A whole sentence of a sequence of amino-acid-denoting words is needed to instruct the cell to make a specific protein. A whole sentence is also exactly what a gene is.7
Compared to those primitive pre-life copiers, which could represent limited information within their simple molecular structures, RNA can instantiate many times more ideas. It does this by building multiple protein molecules—potentially thousands within a cell. And each protein could be a far more complex chemical construction than would ever be possible in a simple non-life copying object.
We are now dealing with a system capable of enormous complexity and flexibility, even if any change in implicit ideas can largely only arise from the random changes of the RNA code in future generations. Before, it might have appeared a stretch to discuss simple replicating non-life chemicals as representing ideas about the environment, because the information would be so minimal and so closely locked into the shape and chemical properties of the object (although this immature information-carrying capacity was still the critical feature that evolution acted upon to move from non-life to life). But now it should become clear that an RNA-based life-form, with its special code of letters, like the 0’s and 1’s on a desktop computer, and its software programs for making proteins, is carefully shaped by evolution largely as an information-storage device. There is also a vast potential for adaptation across the generations as evolution tweaks the sequence of letters in order to update the successful traits recorded in RNA—killing off those creatures with letters that do not capture the world well, and nurturing those with letters that reveal the best ideas. In this way, the genes are not only storing information, but, if viewed over many generations, also blindly learning about how best to live in the world.
But while RNA is a mammoth step toward life compared with simple replicating chemical objects, it has various drawbacks. As a molecule, RNA is unstable and tends to degrade relatively easily. This is no problem for a short piece of information, which can be replicated quickly, but for anything longer, with many thousands of letters of information, it simply isn’t practical. The longer sequence of information would deteriorate so quickly that the organism would have little chance of passing on to the next generation those useful qualities that natural selection bred into it.
In other words, if you want to increase your information capacity, RNA is not your molecule of choice. It simply doesn’t scale up well: The more information it stores, the less information it can successfully pass on to the next generation. Any useful balance between stability (maintaining a belief) and chaos (creatively exploring new ideas—some good, some bad) will slide disastrously toward the chaotic side, and all beneficial concepts accumulated in that family of RNA life-forms will eventually be lost, inevitably along with the life-forms themselves.
DNA solves this problem. Bacteria, arguably the first real life-form, may seem to us exceedingly simple. However, even the smallest, most basic bacterium requires a DNA string of more than 100,000 letters of code in order to form the recipe of its biological makeup. DNA, despite requiring considerably more energy to copy, is vastly more stable than RNA, which means that far fewer mistakes appear when it is duplicated. For these reasons, there may well have been a strong evolutionary pressure for life to start using DNA as the primary storage molecule for information (with RNA now playing an intermediary role between DNA and protein). So DNA may have arisen relatively easily and early, especially since it is extremely similar in structure to RNA—the main difference being that DNA is made up of pairs of letters in a double strand, rather than the single strand of RNA.
I can now return to the issue of the extent of complexity and adaptability in a concrete way, asking these questions for life rather than for some simpler non-life alternative. When compared to the 3 billion letters of code in the human genome (the entire complement of genes in an organism), the 100,000 letters in a bacterium is tiny. Nevertheless, it is sufficient in principle to generate vastly more possibilities of different types of proteins than there are atoms in the universe. In fact, to exceed the number of atoms in the universe, 1080, you only need a few hundred DNA letters. So bacteria can, in principle, be reprogrammed to do almost anything you can conceive of. For instance, biotechnology engineers are currently creating novel forms of bacteria to make diesel fuel as a waste product.
A UNIVERSAL RECIPE AND A UNIVERSAL LANGUAGE
We have now arrived at the common blueprin
t for all life on the planet: DNA stores the instructions for the organism’s structure and function, and RNA, having made a temporary copy of sections of this DNA code, in turn converts this information into many different types of proteins—and proteins are the key molecular tools of all organisms. This model must have been a wild success when it first occurred, by its dominance wiping out all the alternatives. Scientists believe this is the case because virtually all life that we know of, from the simplest bacteria to all the animals, including ourselves, shares exactly this process, and exactly the same language.
This universality of DNA is completely staggering. I recently stayed in a hotel on the east coast of Italy, in a village called Vasto, where I was forced to communicate with the local staff in a combination of incredibly meager Italian and awkward hand-gestures. It wasn’t so bad, as I know a smattering of Spanish and French, and we all fumbled our way through conversations. All these languages are relatively similar to each other anyway. After all, we live on the same continent and are—of course—the same species. And yet, it’s entirely understandable that the differences in our histories and culture, stretching back a couple of millennia, would have created changes sufficiently deep that most of the words in Italian sound and look at least somewhat different from English.
But the contrast between European languages and the language of genes couldn’t be more extreme. Our evolutionary path diverged from bacteria a billion years ago. Nevertheless, the meanings of virtually every single one of the sixty-four possible words (these triplets of bases that code for amino acids) are identical between us. So throughout the biological world, it’s not just that there is only one universal language, but that there is a single dialect of a single accent within this single language!
