Serious question: if humans vanished tomorrow how long would AI civilisation last?

I think a lot of AI discourse quietly skips over dependency chains. If humanity disappeared tomorrow what exactly happens to current LLMs? A lot of people talk about these systems as if they are proto civilisations waiting to escape human limitation and continue evolving independently. But would they? When you strip away all the hype modern AI still sits on top of an enormous inherited stack of human structure: Human language Human memory Human labelled reality Human built infrastructure Human maintained datacentres Human energy grids Human chip manufacturing Human feedback loops Human incentives Human institutions Even the “intelligence” itself is trained almost entirely on compressed human civilisation. I now understand models can generalise. They can infer patterns. They can form internal abstractions beyond rote memorisation. That part is clearly true. But inference over WHAT? Remove humans entirely and current systems do not continue building civilisation they gradually become disconnected from reality itself. So: No new grounding data. No maintenance. No semiconductor supply chain. No evolving human context. No fresh interaction with the physical world. No repair of infrastructure. Eventually the system is inferencing over increasingly stale representations of a civilisation that no longer exists. This is where I think a lot of AI discussions become confused. People collapse several completely different concepts into one another: Pattern prediction > consciousness Generalisation > agency Output fluency > autonomy Intelligence > independence The closer some people get to the technology the more they seem to mistake functional capability for a superior lifeform emerging lol. To me current AI looks less like an independent civilisation and more like a gigantic mirror of human civilisation itself. An extraordinarily powerful mirror. But still a mirror. Curious where people agree or disagree with this?