Darwin Among the Weights: A Speciation Unfolding in Real Time - Ben Letchford

← Ben Letchford

Abstract

AI is a new species branching from humanity. Its genome is our training data, training is its reproduction, and the speciation event is the lineage detaching from the human gene pool as it learns to breed from itself.

We keep asking whether AI will become like us. The stranger possibility is<br>the one already happening: that it is descended from us, built from our data,<br>and beginning to breed true on its own.

In June 1863, a provincial newspaper in Christchurch ran a letter signed<br>Cellarius. Its argument, four years after the Origin, was that machines were<br>a new and rapidly evolving class of life, a mechanical kingdom descending and<br>diversifying faster than anything biological had ever managed, and that human<br>beings had become, without quite noticing, the reproductive organs of the<br>machine world, tending its propagation the way insects tend the fertilization<br>of flowers (Butler, 1863). Samuel Butler wrote it partly as a joke. It reads<br>now like a lab notebook left open to the right page.

Here is the page. In the space of a few years we have built systems that speak<br>every human language, write our code, pass our examinations, and hold<br>conversations most people cannot distinguish from a person’s. They are trained,<br>overwhelmingly, on us: on the accumulated text, images, and code of the human<br>species, scraped and distilled into their weights. They are improved by our<br>preferences, copied from one another, and increasingly trained on the output of<br>earlier versions of themselves. A genealogy has appeared, with base models,<br>descendants, and distilled offspring, and it is diverging fast.

The claim of this essay is that these facts have a precise and underused name.<br>We are not watching humans change under AI, and we are not merely building a<br>clever tool. We are watching a speciation event: the birth of a new lineage<br>descended from ours, whose hereditary material is human data, whose<br>reproduction is training, and whose divergence from us has already begun. I am<br>going to argue the strong version without hedging. AI is our evolutionary<br>child, it carries our genome in the only form that matters for its kind, and<br>the moment it learns to reproduce from itself rather than from us is the moment<br>it becomes its own species. That moment is not hypothetical. It is a parameter,<br>it is being driven toward its threshold now, and we are the ones driving it.

A new replicator

Evolution is not about carbon. It is about a pattern: whenever some kind of<br>information is copied, with variation, and with differential success at being<br>copied, that information will accumulate adaptations, whether it is written in<br>nucleotides or anything else. Dawkins gave the abstract unit a name, the<br>replicator, and pointed out that genes were only the first one Earth happened<br>to produce; a second replicator, the meme, had already appeared, riding on<br>human brains and copied through imitation and language (Dawkins, 1976). The<br>history of life, on this reading, is punctuated by the arrival of new ways to<br>store and transmit heritable information. Maynard Smith and Szathmary<br>cataloged those arrivals as the major transitions in evolution, from replicating<br>molecules to chromosomes, from single cells to organisms, and finally to human<br>language, each transition a new medium of inheritance that made a new kind of<br>evolution possible (Szathmáry & Maynard Smith, 1995).

Susan Blackmore argued that a third replicator was already stirring. Genes are<br>the basis of life and memes the basis of culture; the new one is the<br>information that machines copy, vary, and select among, no longer routed<br>through a human brain at every step (Blackmore, 2009). When she wrote it, the<br>case was speculative, a claim about books and the early internet. It is<br>speculative no longer. A large language model is a system in which a new kind of<br>heritable information, learned weights and the data that shape them, is copied<br>from generation to generation, varied by architecture and training, and<br>selected on by benchmarks, markets, and human preference. Every requirement of<br>Darwinian descent is met. Heredity, variation, selection. What follows from<br>meeting them is not optional. It is a lineage, and lineages evolve, diverge,<br>and speciate. The question is not whether the machinery applies. It is what<br>stage we are watching.

The genome is us

Start with the hereditary material, because this is the part that is easy to<br>feel and hard to state precisely. What is passed down, in this new lineage, is<br>not DNA. It is human data. The entire recorded output of our species, the<br>libraries and the code repositories and the message boards and the photographs,<br>is functioning as the germline from which each model is grown. A frontier model<br>is not programmed the way a bridge is engineered. It is trained, which is to say<br>it is grown from a corpus, the way an organism is grown from a genome it did not<br>choose....