Antelope track chain. Column in ComputerreOnline #39

←

Dmytro Shabanov

→

The Causes of Our (Im)perfection

Antelope Track Chain

Markov and Human Evolution

Column in ComputerraOnline #38

Column in Computerra Online #39

Column in Computerra Online #40

I want to start by saying that some problems that might seem simple are solved extremely complexly or not at all. Let's take the n-body problem as an example. Interesting information about this problem can be found, for instance, here. Since Newton's time, science has known how to describe the motion of a single body that does not interact with other bodies. It will move in a straight line and at a constant speed under the influence of inertia. Furthermore, we know how two bodies (which for simplicity can be imagined as two points with mass) behave: they attract each other with a force proportional to the product of their masses and inversely proportional to the square of the distance between them. If Isaac Newton solved the problem for one and two points in the 17th century, then surely science has made significant progress in understanding the interaction of bodies since then? So, a general solution to the three-body problem has not been found to this day. The possibility of solutions to this problem for some particular cases was shown at the end of the 18th century by Leonhard Euler, and soon such solutions were obtained by Joseph Lagrange. After prolonged efforts, Henri Poincaré proved at the end of the 19th century that such a problem has no direct mathematical solution: it cannot be solved through algebraic and transcendental functions of coordinates and velocities. At the beginning of the 20th century, the Finnish mathematician Karl Sundman found a solution to this problem in the form of convergent series. Unfortunately, Sundman's series do not allow for long-term and accurate prediction of the dynamics of three bodies and, moreover, require colossal computational power, difficult to access even now, for their calculation. And how can spacecraft be controlled in such a case? Solving the restricted three-body problem involves calculating the motion of a small-mass body (a spacecraft) and the gravitational field of two large bodies (e.g., Earth and Moon), which the small body does not affect. Furthermore, spacecraft control includes constant course corrections. Wait, wait. After all, real celestial bodies (which, by the way, are not material points but continuous complex structures made of many parts with varying density) do not perform any calculations to decide where to move! How do they manage? Another example. In almost every cell of our body, countless ribosomes work continuously – molecular robots that link amino acid residues according to a program reflected in the structure of RNA. Dozens of amino acids are added to the chain per second; a few seconds, and a complex protein is ready. The synthesized polymer is a flexible chain on which various radicals are regularly located – charged and uncharged, polar and nonpolar, capable of interacting with each other and indifferent. The protein chain begins to fold in a certain way even during its growth; sometimes the nature of such folding is influenced by chaperone proteins (their name comes from "usher," called "witness" in our tradition); in some cases, the already synthesized chain undergoes chemical modification, which affects the spatial arrangement of its parts. The problem of modeling the conformation (spatial arrangement) of a protein molecule is one of the most complex computational problems solved on modern computers. Why is it so difficult for modern computers and so simple for protein molecules that fold without any thought? Because the algorithms of our calculations do not correspond to the nature of interaction in our physical reality. We calculate linear chains of cause-and-effect relationships. These chains are branched, and the number of variants in them grows like the number of grains in the legend about the reward for the inventor of chess. Calculating all of them is an astronomical task; we have to calculate only some, which a certain algorithm selects as the most promising, but still, such a task turns out to be extremely complex. Perhaps quantum computers will fix the situation? The very idea of building a quantum computer is associated with an attempt to move away from using sequential algorithms. How good it would be if we didn't solve operations step by step, but launched entangled particles in a superposition state. They would immediately test all possible solutions to the problem, and then we would only have to choose the one that suits us. Are you surprised that the development of quantum computers lags behind the evolution of traditional, algorithmic ones? It's not just that they are based on more complex physics (quantum mechanics is also actively used in building traditional computers). The problem is that their logic is fundamentally different from ours. And how could it happen that our logic, with which we understand physical reality, works differently than this reality itself? If we were created in the image and likeness of the Creator of this world, we could expect to have something similar to His thinking (well, perhaps weaker, but still...). If our ability for logical reasoning were based solely on culture, one could teach a person to think and solve not like most people, but like the physical world around us "thinks" and "solves." But these are all fantasies. We are limited by our nature, and our thinking develops on a clearly defined biological basis. And this basis evolved to solve specific problems. Do you understand that a specialized microchip, "tuned" for certain operations, can perform them more efficiently than a general-purpose processor on which many different processes can be run? Our current brain is a relatively general-purpose processor, but it has evolved through various upgrades from a quite specialized device. Remember the video from the previous column. Do you think tracking is a simple task? Readers reproached me for excessively considering human tracking ability superior to, say, a wolf's. Of course, in the ability to catch the scent of prey with a low nose (or sense its approach with a high one), we are no match for a wolf. Let's trace how the relative importance of different sensory organs changed during our formation. Our fish-like and semi-reptilian ancestors seem to have had good vision. But the class of mammals became a group of nocturnal animals. Vision simplified, and smell and touch came to the forefront. For most mammals, these senses are the leading ones. But one of their groups mastered the trees. Neither smell nor touch will help assess the properties of a branch to jump onto – vision is needed here. And weakened vision began to improve again. In the course of evolution, different groups of primates restored the lost pigments necessary for quality color vision in different ways (see Dawkins' "The Selfish Gene"). And then one of the primates became a hunter capable of tracking prey by its tracks. Did he restore his sense of smell, so useful for tracking prey? No. Comparative genomics shows that after the divergence of our ancestors and chimpanzee ancestors, we continued to rapidly lose genes responsible for the olfactory receptor function. The fact is that we interact with tracks in a fundamentally different way than a wolf. A wolf sniffs the track; we, like our ancestors, read it. A wolf doesn't even look at the track it's following. A person examines it, noting not only the shape of the track but also the nature of the arrangement of individual prints, the peculiarities of their indentation into the substrate, and much more. While tracking prey, a person can draw conclusions like: "Here it heard a bird, whose track is visible to the left, and moved to the right; and here it saw the shadow of a hiding place ahead and turned back to it." This is our privilege. And are you surprised after this that for long-term information storage, representatives of our species chose signs applied to a plane? And that they initially pressed characteristic imprints into soft clay and only later began to use a clean surface for paint? We are excellent at reconstructing chains of cause and effect. Not every effect necessarily corresponds to one cause; we are quite capable of analyzing the interaction of two or three factors that interest us. But still, our thinking remains algorithmic, chain-like. "A influenced B such that C should have happened, but due to D, E occurred." The consequence of our chain logic is that in any cognition, we try to impose it on the object of our study. It doesn't matter how it functions – we think this way and build models of external phenomena on this basis. As a result, we have a very poor understanding of the dynamics of processes based on the simultaneous action of many internal interconnections. I will give a couple of examples: one simpler, the other more complex. Are we capable of adequately perceiving complex self-organizing processes? Shortly after the "Orange" Revolution in Ukraine, I communicated with many intelligent Russians. The idea that mass citizen protests could be caused not by CIA intrigues but by self-organization was rejected as absurd. Recently, many of these people were at Bolotnaya Square. The version that Hillary Clinton led them there is rejected by them with the same certainty – it was quite natural for them to act this way. And the Soviet Union, of course, was dissolved by the CIA (extremely surprised by such success, completely unexpected for it). And do you think that those who want to control political processes have not tried to learn to calculate complex socio-political phenomena based on the interaction of large human masses? They tried, they tried. Only, apparently, they did not achieve much success – innate logical schemes interfere. And the same logical schemes are responsible for the fact that when observing such processes from the outside, the simplest explanation is the version of a conspiracy or backstage provocation. To hell with politics. Let's consider modern ideas about the development and evolution of organisms instead. How do we tend to perceive it? A culture of bacteria with a broken enzyme responsible for lactase synthesis (an enzyme that breaks down the sugar lactose) grows in a medium containing lactose. Bad, hungry. Suddenly, in one of the cells, the lactase gene mutates; its new version synthesizes a normally functioning protein. The cell acquires the necessary enzyme, it begins to break down lactose, the cell gets additional nutrition. As a result, it begins to grow and divide better than its competitors. After some time, the new trait becomes characteristic of most cells in the studied Petri dish. Look, the chain is clearly visible here: gene → trait → effect. We love such chains; they are understandable to us. Remember when the Human Genome Project was being carried out, we were told that as soon as we decipher all the genes, we would know how the development of representatives of our species is controlled? They deciphered it. They obtained a lot of valuable information, learned about many individual gene → trait → effect chains. But the main task, the solution of which was proclaimed as inevitable by the PR people, was not solved and will not be solved in the foreseeable future. For many traits that are truly interesting to us, everything turns out to be more complex. How? Approximately like this: (interaction of non-genetic inheritance mechanisms) → genes → (interaction of genes, traits, and environment) → traits → (interaction of traits and multifactorial interaction of organisms with the environment) → effects. And, believe me, the complexity of these interactions significantly exceeds the complexity of the interactions of three mutually attracting bodies. So, what to do? Stop trusting our minds? Of course not. But do not forget about their limitations and look for ways to reason about complexity that go beyond the simple schemes of our innate logic.

←