> Isn't it more likely that countries without native frontier models will use open-weights models that will let them be 70% as productive of counties that do?

Possibly, but sometimes being marginally better lets you win all the games. It can be difficult to guess what matters and how.

Also, may depend where the compute is physically located, which may depend on the cost of energy.

▲

Schiendelman 2 hours ago | parent | next [-]

That's exactly why TSMC keeps beating the pants off Intel. Marginally better looks miles ahead in computing a lot of the time.

▲

eru 2 hours ago | parent | prev [-]

Luckily, competing Chinese companies are selling cheaper and cheaper solar panels to anyone who's willing to take them.

▲

ben_w an hour ago | parent [-]

For now.

Globalisation's great right up until your manufacturing base decides they don't like you, or you them.

Reshoring is expensive for all the reasons offshoring was ever tempting in the first place, so we'll have to wait and see how this plays out.

▲

eru an hour ago | parent [-]

If China stops selling you solar panels tomorrow, you still have all the solar panels they sold you until now.

And why do you see reshoring as the only alternative? Seems like a false dilemma. There's more than two places on the globe.

	▲	ben_w 42 minutes ago \| parent [-]
		> If China stops selling you solar panels tomorrow, you still have all the solar panels they sold you until now. Which is great, unless demand for electricity goes up loads. > And why do you see reshoring as the only alternative? Seems like a false dilemma. There's more than two places on the globe. "Here" and "not here" is a complete set. I could also phrase it as "autarky or trade".