During World War II, purebred dogs were considered status symbols, but inbreeding led to health issues like hip and joint problems. In today’s world, diversity is sought after in gene pools as climate change requires hardy species. An experiment shows that AI with diverse composition in its neural networks outperformed AI without diversity in performance by 70% to 57% accuracy on a standard numerical classifying exercise. AI’s internal structure improves with diversity, just as with humans and animals in nature.
When a neural net was allowed to select its own neurons, something startling occurred
