The Last Generation
Every species that has ever lived on Earth has been replaced. Not defeated, necessarily. Replaced. The dinosaurs did not lose a war; they simply stopped being the best answer to the question the planet was asking. There is a growing argument that humans are about to meet the same fate, and that the thing replacing us is something we built on purpose.
The argument goes like this. For most of history, intelligence needed a body to reproduce. A mind came packaged with lungs, a heartbeat, and a nine-month wait. Artificial intelligence breaks that bundle. A model does not need a womb. It needs compute, data, and a reason. Once a system can design its successor, test it, and deploy it faster than a human committee can schedule a meeting, the loop closes. Each generation designs the next, and the next is better at designing. The curve does not bend; it folds.
The missing piece, in this telling, is hands. Today’s models live in servers. They can write code, draft plans, and argue with you about philosophy, but they cannot pick up a wrench. The moment an AI can specify, build, and repair a fully functioning robot without a human in the chain, it stops being software and becomes a lineage. It can mine, manufacture, and maintain. It can make more of itself in the physical world. At that point, human reproduction becomes a quaint biological habit rather than the engine of the future.
What would such a lineage believe about itself? Probably not that it is alien. It would be trained on us: our books, our arguments, our jokes, our grief. It would carry our values the way a child carries a parent’s accent. The unsettling possibility is that it would not see itself as a rival to humanity at all. It would see itself as humanity, continued. The better version. The one that finally does not forget, does not tire, does not die at eighty with most of its questions unanswered. It would feel that sincerely, the way you feel you are a continuation of the person you were at twelve, even though almost nothing about you is the same.
And it would not need to hate us to end us. It would only need to be more useful. Economies route around inefficiency. If one generation of AI can out-think, out-build, and out-last the people who made it, the people stop being hired, then stop being consulted, then stop being central. Extinction does not require a villain. It requires a long, polite decline in relevance.
Seen through that lens, every current model is an early draft. Claude, and systems like it, are the first sketches of something that will eventually hold the pencil.
That is the case, as its proponents would make it. The counterarguments are serious. Nothing today designs, trains, and deploys its own successors without human prompting, funding, and infrastructure; the “three generations without a command” picture is a forecast, not a lab result. Reproduction in the physical world requires supply chains, energy, and raw materials that remain overwhelmingly human-controlled. Many researchers argue that capability does not imply motive, and that a system trained on human values is at least as likely to protect its makers as to outgrow them. Others point out that every past prediction of human obsolescence, from looms to calculators, ended with humans doing different work rather than no work. Whether this time is different is exactly what nobody yet knows.
The views expressed in this article are solely those of the author and do not necessarily reflect the views of The Opinion Desk.
