IMO, the mechanism isn't the important thing, the behaviour is. If you look at the step-by-step, we are also looking for the next word or motor action (and for whoever is about to suggest that we humans plan ahead, Transformer-based LLMs have been shown to also do this); as this is not a useful description of what it means to be a living brain, I'd say it's also not a useful description of what makes everything post-InstructGPT different from what came before.
<noob> Where do birds go when it rains?
<expert> They
I know the answer: because it leads to model collapse. But why is that? Wouldn't a smart model not collapse? It's seeming like they keep getting smarter because we keep pouring more of our own knowledge into them, not because they are actually getting smarter. And yes, sometimes a dumb but persistent bruteforcer can make new discoveries.
and i think this is exactly the crux;
the really big models need really big datasets
and current gen LLMs get a lot of training data beyond "all books + all of the internet"
the objection is then that producing this additional data would already confound it with pre "virtual cutoff date" knowledge (since the training data probably implies mathematical and SWE concepts that were developed post "virtual cutoff date")
But to prevent model collapse you need a way to pump down the entropy. Much like in thermo, it's an expensive and slow process.