> seems like it would be much easier to identify the handful of well-compensated but problematic industries
That would only incentivize a brain drain. A good example of this is software engineering in UK, Germany, Canada, South Korea, and Japan, as SWE salaries in those countries are not significantly different compared to other vocations.
Also, I'm not a fan of cryptocurrencies, but they are one of the few industries left in the US that incentivizes NAND-to-Tetris level knowledge, and it did help subsidize the GPU buildout that made foundational models easier to train cost effectively.
You can't "command economy" innovation - it can only be nudged.
The solution I've seen most industrial planners use is provide tax holidays and subsidizes for targeted industries, as this helps reduce the upfront cost of hiring, and does give wiggle room to raise compensation. Linking that with production, timeline locks, or even tariffs tends to help force an ecosystem to develop - which is what Japan used to build their shipbuilding and automotive industries in the post-war era.
> Also, I'm not a fan of cryptocurrencies, but they are one of the few industries left in the US that incentivizes NAND-to-Tetris level knowledge,
And what does it do with that knowledge?
> and it did help subsidize the GPU buildout that made foundational models easier to train cost effectively.
Crypto and AI both use GPUs, but I'm not under the impression that AI people repurposed old crypto mines for anything (e.g. crypto mines used janky racks of consumer GPUs, AI typically uses specialized high-end equipment).
Notice how I talked about HPC, distributed systems, and cybersecurity in my comment?
> I'm not under the impression that AI people repurposed old crypto mines for anything
The crypto boom helped incentivize the scaling out of GPU design and fabrication, just like how video games helped with the first iteration in the 2000s.
There's a reason the concept of "dual use technology" has gained currency
That would only incentivize a brain drain. A good example of this is software engineering in UK, Germany, Canada, South Korea, and Japan, as SWE salaries in those countries are not significantly different compared to other vocations.
Also, I'm not a fan of cryptocurrencies, but they are one of the few industries left in the US that incentivizes NAND-to-Tetris level knowledge, and it did help subsidize the GPU buildout that made foundational models easier to train cost effectively.
You can't "command economy" innovation - it can only be nudged.
The solution I've seen most industrial planners use is provide tax holidays and subsidizes for targeted industries, as this helps reduce the upfront cost of hiring, and does give wiggle room to raise compensation. Linking that with production, timeline locks, or even tariffs tends to help force an ecosystem to develop - which is what Japan used to build their shipbuilding and automotive industries in the post-war era.