Solutions That Evolve: Building Self-Improving Systems with Genetic Algorithms - Barry Stahl

A conference talk covering genetic algorithms through two practical use cases. The first applies genetic algorithms to find optimal strategies for a variant of Snakes and Ladders, explaining core concepts like chromosomes, genes, alleles, fitness functions, mutation rates, crossover, and selection strategies. The second, more advanced use case explores using genetic algorithms to find words or phrases that are maximally distant in embedding space, touching on how text embeddings encode semantic meaning via cosine distance. The talk also covers deploying self-improving systems in production using continuously running exploratory services that update operational strategies as better solutions are found.