we'll burn that bridge when we get to it

That doesn't mean literally every application of machine learning or language models goes away forever, because machine learning and generative models existed before Silicon Valley decided they were the infinite money algorithm, and will still be around after they stop thinking that. "Generate ten videos of cat people working in the baby mines for free" or "have the chatbot generate a summary every time I google" are not sticking around.

Roughly ten years from now, we'll have completely different approaches to using machine learning-based software, based on the applications of these algorithms that actually make sense given the resources used. They'll likely look a lot like the applications of machine learning that already existed before OpenAI, because those were motivated by finding places where these models were useful first, then developing the models afterwards.

this does not. they're playing world of tanks which is like the "call of duty" of tank games (casual, players only slightly bad-smelling). it also doesn't have support for tank peripherals. no game does. no trainers do afaik. which means that (assuming this isn't just a video editing) all of that shit they are fucking with translates into mouse/keyboard inputs that the game understands. that's weird/hard and perplexing, uh, and considering that "tank peripherals" aren't a thing that exist i can only guess they built them theirselves