Like, why would you think the bubble popping would lead to open-source, lightweight, ethical replacements? It hasn't happened with, say, cryptocurrency: it's just the same stuff, but less of it. And, like, why would you assume OpenAI, Anthropic, etc., just disappear instead of pivoting to something else slightly different, like non-generative machine learning? Why would governments stop finding surveillance tools necessary?

Also, the negative effects, like the resource usage, the massive exploitative industry of people labelling and filtering data, cannot be handwaved away because they are a big reason why the tech works like it does. They are intrinsically connected. It's like asking "you don't want to own a horse? what if there was a magical pony that never pooped and could fly, what then?" It's not a serious question.

All that aside, I personally have no use for the things LLMs are "good" at. I genuinely want to know how to program and learn why it works, and that is more important to me than getting a working program that I don't understand. I am perfectly capable of writing grammatically correct and well-formatted text on my own. And I don't need a chatbot companion.

But in that hypothetical world, I guess I'd see if there's a use case. LLMs are inherently prone to generating errors, and no amount of development will get rid of that because generating stuff without understanding fact value is what they *do*, which limits their usefulness no matter what.