4 ms·

China AI and Semiconductors Rise: US Sanctions Have Failed

9 points by themgt 3y ago

alephnerd 3y ago
The US Sanctions weren't meant to stop sub-14nm processes despite what Redditors said. The main point was to keep a 3-5 year lead on Process R&D. That said, I've found the policymaking around the whole Semicondutor space very infuriating. There's a lot of blowhard bullshit going on, and in an actual shooting war we'd be seeing larger sized chipsets used (eg. 14nm, 28nm, etc). You don't need computational performance for a precision missle - it's anyhow going to crash itself. And any attempts as slowing Chinese simulation/Exascale compute would have anyhow failed as 1. The sanctions only target military, not educational and private sector players despite China's robust MIC 2. Parallelism can make up for any raw compute losses. I have some additional thoughts about this having worked with some of the people writing these policies but ehn
- throwaway4good 3y ago
  It is a massive breakthrough of the US sanctions in particular those directed at Huawei.
  - alephnerd 3y ago
    It was always going to happen. It's like how the Soviet Union got nuclear weapons - they were able to develop an indigenous program because there was already a decent talent base and a national priority. No one argued that these policies would ever stop China, but they would at least slow down any military technology R&D. Also, we shouldn't be surprised when South Korea, Japan, Germany, Malaysia+Singapore, Israel, and India end up building similar capacity within a decade. The same thing happened with 4th Gen Fighter Jets and is happening with 5th and 6th Gen ones too. It's all engineering, and talent+knowledge is very diffused now in 2023. It's just a capital allocation and execution problem at this point for most large countries. This is the policy paper that inspired the "Chip Wars" talk btw - https://cset.georgetown.edu/publication/silicon-twist/ https://cset.georgetown.edu/publication/silicon-twist/

9 points by themgt 3y ago

The US Sanctions weren't meant to stop sub-14nm processes despite what Redditors said. The main point was to keep a 3-5 year lead on Process R&D. That said, I've found the policymaking around the whole Semicondutor space very infuriating. There's a lot of blowhard bullshit going on, and in an actual shooting war we'd be seeing larger sized chipsets used (eg. 14nm, 28nm, etc). You don't need computational performance for a precision missle - it's anyhow going to crash itself. And any attempts as slowing Chinese simulation/Exascale compute would have anyhow failed as 1. The sanctions only target military, not educational and private sector players despite China's robust MIC 2. Parallelism can make up for any raw compute losses. I have some additional thoughts about this having worked with some of the people writing these policies but ehn

throwaway4good 3y ago
It is a massive breakthrough of the US sanctions in particular those directed at Huawei.
- alephnerd 3y ago
  It was always going to happen. It's like how the Soviet Union got nuclear weapons - they were able to develop an indigenous program because there was already a decent talent base and a national priority. No one argued that these policies would ever stop China, but they would at least slow down any military technology R&D. Also, we shouldn't be surprised when South Korea, Japan, Germany, Malaysia+Singapore, Israel, and India end up building similar capacity within a decade. The same thing happened with 4th Gen Fighter Jets and is happening with 5th and 6th Gen ones too. It's all engineering, and talent+knowledge is very diffused now in 2023. It's just a capital allocation and execution problem at this point for most large countries. This is the policy paper that inspired the "Chip Wars" talk btw - https://cset.georgetown.edu/publication/silicon-twist/ https://cset.georgetown.edu/publication/silicon-twist/