Universal statistical signatures of evolution in artificial intelligence architectures
For evolutionary biologists and AI researchers, this work demonstrates that the statistical structure of evolution is substrate-independent, governed by fitness landscape topology rather than selection mechanism.
This study shows that the distribution of fitness effects of architectural modifications in AI follows a heavy-tailed Student's t-distribution, with proportions (68% deleterious, 19% neutral, 13% beneficial) placing AI between compact viral genomes and simple eukaryotes. The DFE shape matches D. melanogaster and S. cerevisiae, and architectural origination follows logistic dynamics with punctuated equilibria.
We test whether artificial intelligence architectural evolution obeys the same statistical laws as biological evolution. Compiling 935 ablation experiments from 161 publications, we show that the distribution of fitness effects (DFE) of architectural modifications follows a heavy-tailed Student's t-distribution with proportions (68% deleterious, 19% neutral, 13% beneficial for major ablations, n=568) that place AI between compact viral genomes and simple eukaryotes. The DFE shape matches D. melanogaster (normalized KS=0.07) and S. cerevisiae (KS=0.09); the elevated beneficial fraction (13% vs. 1-6% in biology) quantifies the advantage of directed over blind search while preserving the distributional form. Architectural origination follows logistic dynamics (R^2=0.994) with punctuated equilibria and adaptive radiation into domain niches. Fourteen architectural traits were independently invented 3-5 times, paralleling biological convergences. These results demonstrate that the statistical structure of evolution is substrate-independent, determined by fitness landscape topology rather than the mechanism of selection.