Evolutionary Approach to AI
Hitoshi Iba · 2025
In chapter 2 , we explain recent advanced studies on the emergent approach to AI. For example, sexual selection will be described in detail. This selection method is widely applied to various engineering fields as “novelty search.” We show the effectiveness of sexual selection with well-known benchmark tasks, i.e., the knapsack problem and N-queen problem. As we have discussed in chapter 1 , the key idea of generative AI is “creativity.” In this context, IEC (Interactive Evolutionary Computation) framework is explained in relation to recent prosperous researches in design engineering. IEC has useful applications integrated into deep learning. Next, we explain the concept of “co-evolution.” As a result of co-evolution, one of the following is established between organisms: (1) competition (both harming each other), (2) parasite (only one benefits while the other does not), and (3) cooperation (both benefit from each other). The transition from competition to cooperation is believed to occur in the course of evolution. In other words, a host that initially excludes the parasite learns to cooperate with it at a certain point. This mechanism has been used in the applications of evolutionary computation. An example of this is illustrated in evolving sorting networks, i.e., constructing an efficient sorting algorithm with as small a number of comparators as possible.