Artificial intelligence learns like evolving populations across generations
The evolution of sex for artificial intelligence: a population-genetic framework for multigenerational model populations
Machine LearningNeural and Evolutionary Computing
Summary
Artificial intelligence models develop through generations like living organisms do, using processes similar to biological reproduction. The authors compare AI practices such as retraining models on outputs of other models to sexual and asexual reproduction in biology, creating a mathematical framework to describe how AI models inherit information. They test this framework on various AI models and find parallels including effects like genetic drift and blending inheritance. This helps explain how AI models evolve and sometimes fail, and how combining models can lead to better results or reproductive isolation.
What this means in practice
- •For machine learning engineers: Design training regimes that optimize AI model combinations to improve performance and avoid collapse by using population-genetic inspired strategies.
- •For ai system integrators: Develop better AI model ensembles and merger approaches based on genetic inheritance analogies to exceed individual model performance.
Authors
Giorgio F. Gilestro
Abstract
Some aspects of AI development resemble a population process in which models are specialised, retrained on the output of peers, or combined by averaging weights. These practices lead to generations of models, in the biological sense studied by population genetics. Here, I develop this parallelism and interpret multigenerational model populations in terms of sexual and asexual reproduction, formally recombining the two fields. I test these analogies in an exact inheritance model, in trained networks (recurrent, feedforward and variational autoencoder generators) and in large language models, and show that they hold generally, with some measurable architecture-specific biases. Training recursively on model output is known to lead to model collapse, a process previously described as akin to genetic drift; I develop all that follows. A minimal model of a learner retrained on its parent's output reproduces the Wright-Fisher process exactly; verified real data added to each generation play the role of immigration, with the surprising finding that the absolute number of real data samples matters, not their share, exactly as in population genetics. Training a child on the average of its parents' outputs cancels the benefit of having several parents, matching blending inheritance (and reviving Jenkin's objection to Darwin), whereas combining parents so that each keeps its strongest contribution preserves it; merged language-model specialists exceeded every parent across seeds (the Fisher-Muller effect); and lineages become reproductively isolated, losing the ability to merge at all, when they have learned conflicting conventions and not when they have merely drifted apart. As AI societies become societies in time as well as in space, a mathematical framework for their inheritance acquires predictive power. Remarkably, that framework can be adapted almost wholesale from biology.