(Imagine if you will: everyone constantly doing research into new lightbulb technologies — so we’d get the LED a century sooner! — but nobody ever researching the implications of evacuated incandescent bulbs [vacuum tubes! CRTs!], nor into the implications of miniaturizing/printing LEDs, etc.)
It’s kind of like where fusion research and ML research were during their “dark decades” — everybody doing breadth-first searches of the possibilities, nobody digging much deeper into the implications of research that showed promise.
In all three cases, it’s because it’s hard to model those implications without a real experimental prototype; and because, in all three cases, costs to do real experiments (for reactors; for training datasets; etc.) shoot up exponentially as you “go deeper” — as you need to do those experiments in the context of a working example of everything you know so far.
The true issue is one of team size dynamics. Particle physics has figured this out: to continue foundational research, you need a few large teams with huge grants/budgets (to build particle accelerators), rather than many small teams (that can’t individually afford to build particle accelerators.)
But other fields haven’t managed to have the same epiphany, and so in those fields, small teams just continue to explore the parts of knowledge-space they can “reach” with a small team — which isn’t really too productive.
(ML got going again because large companies figured out basic research into ML could be a loss-leader for renting out time on specialized ML-accelerator hardware. Nuclear science never slowed down, because governments were doing the basic research with huge teams+budgets from the start. Etc.)
My favorite was when they showed the image with the working microscope.
Me> They only get significantly ahead of the industry when nobody else is doing research.