It is not the most well-known branch of artificial intelligence (AI). Much less so in the era when large language modeling (LLM), computer vision, and robotics are stealing much of the limelight. However, it is an added opportunity with enormous potential that Making Sense is exploring to deliver more value to its clients.

This is a subtype of heuristic algorithm known as evolutionary or, to use the more appealing name, genetic. Two properties characterize these algorithms. First, they can deliver fast and simple solutions to highly complex problems. Second, they mimic real-life situations, including aspects of biology, chemistry, or social behaviors.

The term “genetic” reminds us of Charles Darwin’s Theory of Natural Selection. Let’s see an example of how this works. Think of a first generation of a turtle population with different chromosomes: some have shells, others show flippers, and others have legs.

The path to optimization

This entire generation is tested in a given context. The result: only the best individuals, i.e., those that pass the test, survive and reproduce. Three types of offspring emerge: the elite, i.e., the spitting image of one of their parents; the cross-overs, which take on characteristics of both; and the mutations, which give rise to something new. This iteration process continues until the optimal specimen is created.

Companies do not face problems involving turtles daily, it is true, but AI-based genetic algorithms allow us to replicate this scheme to tackle multidimensional, nonlinear problems with numerous parameters and find the best possible solution. For example, in the case of investments, the algorithm can optimize the parameters of the defined strategy to perform the appropriate action at the best time.

Generally speaking, the genetic algorithm requires an initial set of potential solutions and a fitness function to evaluate each quantitatively. The algorithm will then refine the solutions (some may not work; others may already have been successfully tested to solve similar problems) to eliminate those it does not consider “fit” and select the most promising ones.

Flexible, versatile, and comprehensive

Among the benefits of genetic algorithms are the possibility of finding multiple solutions to a single problem, the versatility of reaching a wide range of challenges and domains for which mathematics lacks appropriate models and that, unlike traditional methods, they are applicable to problems with non-differentiable or discontinuous objective functions.

ChatGPT occupies much of the discussion, even at family dinner tables, and Copilot seems an unavoidable topic among developers and software experts. But behind them, AI presents us with numerous opportunities to grow businesses. In such a fast-moving, fast-paced world, there’s nothing better than being in the group of organizations that can survive.