CONCEPT
Model Genealogy
The recognition that AI systems do not exist as standalone artifacts but as nodes in a tree of descent—inheriting from foundation-model ancestors, diverging through fine-tuning, and merging across lineages in ways that make the AI ecosystem a genealogy in Darwin's literal sense.
Darwin's single illustration in On the Origin of Species was a branching tree. The tree was his picture of how life diversifies: a lineage splits, the branches split again, most twigs die back, and a few extend forward so that all living forms are the surviving tips of a vast ramifying structure. This geometry is also the right geometry for the space of AI systems. A foundation model is a common ancestor. Its fine-tuned variants are branches. Distilled compressions are offspring. The relationship between any two models is genealogical—a matter of shared ancestry in training history—exactly as Darwin said the relationship between any two organisms is genealogical. To understand a model is to ask what it descended from, what it inherited, and what selection pressure modified it along the way. A flaw in the foundation—a bias absorbed from the training corpus, a blind spot, a vulnerability—propagates into every descendant unless specifically bred out,
Keep reading with YOU ON AI
Unlock the full book, 10,000+ field-guide entries, and a 1000+ thinker library. If you have a book code, register now — it takes a minute.