I think you're right, but history has plenty of technological dead-ends, particularly in the AI/ML world
Future prediction is always dodgy.
I hope you don't plan for best case scenarios when doing your database backups, much less the future of society on earth.
Change will come eventually and it will be painful. But I don’t want my kids to deal with it - I’d rather take it head on and deal with the consequences then praying that we push it to the next generation.
If not when. At a certain point OpenAI and others could need more and more training data for less quality increase.
VR and AR were also a hype. The headsets just need to become as small as normal glasses, could be impossible without sacrificing crucial features like color space or resolution. Meta invested billions but their best headset is still bulky with smartphone quality graphics.
To use your comparison as an example, mainframes took 40ish years to be replaced by personal computing and distributed systems. Whereas smartphones reached modern parity in 8-10 years since the original iPhone and have made more or less incremental changes since.
Granted those are hardware changes and this is software but I believe the point still stands.
Nobody can predict the future. But I think we have enough experience with AI trends now to see where this is going.
Current LLMs are like the mainframes of the 1960s.
When the technology gets speeded up, streamlined, and wrist-watched we're going to be in completely unknown territory - culturally, politically, economically, and socially. Never mind technically.