Genetic Algorithms

In The You On AI Encyclopedia

The algorithm's operational logic is adaptation itself — the mechanism by which any complex adaptive system discovers solutions to novel problems. Biological evolution uses the same logic with DNA as the building block substrate. The immune system uses it with antibody segments. Economic markets use it with business strategies, technologies, and institutional arrangements. In each case, the system does not design solutions. It evolves them.

The creative process in You On AI — the writing of a book through human-AI collaboration — follows this logic with striking fidelity. The process is iterative. The author generates multiple versions of an argument. Some are produced by the author alone, some by Claude, most through back-and-forth in which question generates response, response provokes revision, revision generates new response, and the cycle continues until something emerges that neither the original question nor the initial response contained. This is a genetic algorithm operating on ideas rather than bit strings. Building blocks are concepts, metaphors, argumentative structures, rhetorical moves. Variation is generated by the interaction between human question and machine pattern space. Selection is provided by human judgment. Accumulation occurs across the project's duration.

Holland's framework predicts that such systems require balance between variation and selection. Too much variation relative to selection and the population drifts — the genetic algorithm with no fitness function does not converge. Too much selection relative to variation and the population converges prematurely — often on a local optimum rather than the global one. The optimal balance is narrow and sensitive to perturbation. AI tools have dramatically increased variation available to creators. The machine generates building block recombinations at speeds dwarfing any previous source. But selection capacity — judgment, domain knowledge, aesthetic discernment — has not increased proportionally. The result is a system where the variation-selection balance shifts rapidly toward variation, and Holland's framework predicts with mathematical precision what happens: the population drifts. The prescription is not to reduce variation but to invest at least as heavily in selection as in generation.

Origin

Holland developed the genetic algorithm at Michigan in the 1960s, publishing its definitive formulation in Adaptation in Natural and Artificial Systems (1975). The framework took nearly two decades to achieve wide recognition, becoming foundational to evolutionary computation through David Goldberg's 1989 textbook and the growth of the field in the 1990s.

The algorithm's conceptual debt to biological evolution was explicit from the start. Holland's insight was that evolution was not merely a biological phenomenon but an abstract computational procedure that could be implemented on silicon and applied to problems — engineering design, scheduling, optimization, machine learning — where conventional methods failed.

Key Ideas

No model of the solution required. The algorithm discovers solutions without understanding them, through the feedback of fitness evaluation.

Variation-selection balance. Neither alone produces adaptation; the balance between them determines the system's trajectory.

Cross-population recombination. When populations with different building block repertoires interact, the combined space of possible solutions expands exponentially.

The human as fitness function. In AI collaboration, the human provides the selection pressure that distinguishes genuine emergence from plausible noise.

Premature convergence is the risk. Systems that lose variation collapse onto local optima, losing the capacity to discover global ones.