>>60820204 (OP)
It's a bubble and could pop, but it's insane to say it's not useful technology. Obvious comparison but this could be like the Dotcom boom, people sense that it's useful but it's going to take time to get integrated into business processes. It's not disappearing even if the bubble pops.
It's definitely overhyped though, they want 100's of billions to build "god in a data centre" but its unclear (and seems unlikely to me) that more scale is going to fix the fundamental problems of LLMs. They hallucinate, make shit up to please the user and lack common sense. They'll answer something a PhD student wouldn't understand, and fuck up something a 6 year old could answer with ease. Will another 10x larger training run fix that? I doubt it but what the hell do I know, seems like a few more fundamental research breakthroughs are needed.
For the sake of the economy I hope the next Gemini model shows genuine improvement, or else GPT-5 might mark the point where everything collapsed. Reminder, the US would already have been in a recession if it was for Q1/Q2 datacentre capex..
