Does Evolution Always Mean Progress: Genetic Algorithms and Natural Selection

The debate about whether evolution means progress is a long-standing one, and the example of genetic algorithms explores whether evolution can lead to progress. Genetic algorithms optimize problems by mimicking evolutionary principles, but natural evolution is a complex process of adapting to different environmental changes.

 

Is evolution progress? This is a topic that has long been debated, with many evolutionists sharing their opinions. Some argue that evolution, starting from simple self-replicating organisms, ultimately converges on a final form of life, while others argue that evolution is simply an adaptation to the environment and cannot be viewed as a one-dimensional “progression”. As you can see, there are many different opinions and perspectives on evolutionary development. But whether the debate is organized or not, there are methods that actively harness evolution to achieve optimal results, and they’re working very well. It’s called a “genetic algorithm,” and it’s a methodology for optimizing mathematical models that adopts the concept of evolution from biology.
First, let’s explain what a genetic algorithm is. A genetic algorithm is an algorithm that represents the possible solutions to a given problem in some form of data structure, and then uses incremental variations to get closer to the optimal solution. This is accomplished by allowing different solutions to solve the problem, and then iterating and testing them until a solution that is close to optimal on the quantified data is found. In this model, the data structure of solutions can be analogized to genes, and the testing of the problem and the evolution of solutions based on the testing can be analogized to natural selection. As these iterations and tests are repeated, only solutions that are increasingly close to the optimum remain and continue to combine with each other, eventually converging to the optimum. A recent example of this algorithm in action is the famous “A Study on the Possibility of Improving the AI of a Strategy Card Game Using the Principle of Natural Selection” by a high school student. This study used the computer card game “Hearthstone” to record the evolution of the winning rate as a limited deck of cards was continuously evolved by a genetic algorithm. As the deck’s generation increases, you can see that the deck is increasingly composed of cards that perform well in testing, and the average win rate gradually increases. This doesn’t have to be a game, though; genetic algorithms can be used to find the optimal antenna shape for high radio reception, for example, wherever the results can be quantified.
In summary, a genetic algorithm is a problem-solving method that mimics genetics to produce an optimized model for a given environment. There is a similar phenomenon and term in evolutionary theory. It’s called “natural selection. Natural selection is the theory that within a given environment, species with traits that are best suited to that environment will survive longer and leave more offspring (i.e., genes) than those that are not. Genetic algorithms are similar to this. The model with the best properties for the problem performs well in testing, and more models with that solution exist in the next generation. This is why the methodology is called a “genetic” algorithm. It mimics genetics almost perfectly. However, the difference between the two is the variability of the environment. In the case of genetic algorithms, the problem, or environment, doesn’t change. No, it doesn’t try to change. This is because the model can be noisy, and it’s expensive to converge to a model that solves multiple problems. Engineers want to draw one conclusion for one problem situation, but nature is different. Nature is a complex system that is constantly changing, where variables affect the environment, which in turn changes the environment, which in turn changes the variables. Even if an individual emerges that is optimized for the challenges nature imposes, it may not perform well in a changed environment.
Combining the idea that evolution is a progressive phenomenon with the idea of survival of the fittest, which is the basic goal of natural selection, we can say that evolution is a process of adaptation. In a genetic algorithm, the model that is finally adopted, or the optimal model, is the one that produces the best results for a single test over generations. The model that solves the problem and gives the shortest and most accurate answer is the “optimal” one. But nature is constantly changing. How do we define “optimal” in nature? Nature does not consist of a single problem situation. Many problems are randomly imposed on individuals, and it’s very difficult to determine what form is optimal. An individual that is optimal at one moment may rapidly become less so as the environment changes, and an individual that performs well on average across a range of problem factors may lose out to an individual that specializes in extreme situations where a single factor plays a large role. While humans now occupy a significant area of the planet, no one can easily say that we are the “optimal solution” that has been derived over hundreds of millions of years in the laboratory that is Earth.
A laboratory is essentially a controlled environment. Nothing changes except what needs to change, and nothing is different except what needs to be different. In this way, we can pick and choose one characteristic of the natural environment to explore. But nature outside the lab is the opposite. It is a world where everything changes, where many things are different. Individuals that are optimal at the time may become obsolete and extinct over time. And nature doesn’t do controlled evolution. In a genetic algorithm, only the traits of the best-scoring models survive and are passed on to the next generation. But nature doesn’t selectively leave offspring, it reproduces randomly and multiplies the population. The resulting individuals may not score higher than the best ones in a natural test environment, but if they “survive,” their genes become “conserved.” This is the ultimate difference between the lab and nature. This is the ultimate difference between the lab and nature. The lab is a controlled environment that is designed to “evolve” on a problem, while nature is an environment that “might” evolve. A genetic algorithm is clearly an algorithm that produces a model that has evolved through evolution. In this sense, we could even say that evolution has a progressive direction. However, nature is not a selector that only recognizes the optimal individuals, and many sub-optimal individuals live alongside it in manageable populations. For example, there are insects that are frequently eaten by top predators, but can be found almost everywhere. From this perspective, it’s hard to see evolution as directional.
The debate about the direction of evolution is a huge scientific debate that cannot be concluded by any one person, but it can provide ideas and food for thought. Genetic algorithms created using hereditary phenomena certainly provide “advanced” answers. Whether this progress is the same progress in nature is another debate.