Carlos Perez argues the restrictions fail to prevent global misuse.
@BlackHC @sdmat123 Leaving even aside the business interests of orgs like MSFT, NVIDIA, the whole startup ecosystem which all want/need open models to flourish.
Banning open source models is anti-freedom and authoritarian
@BlackHC @sdmat123 Sounds good, what should bad actors do against that? Just… use DCs in another country with less regulation?
Many users criticized bans on open source AI models as unrealistic and harmful because they would let China lead while rendering US labs obsolete.
No Digg Deeper questions have been answered for this story yet.
@BlackHC @sdmat123 Also: Let’s say Mythos 2 is out and restricted in the US. There also is OpenMythos 2 by a Chinese company.
No US provider is allowed to host it. If you don’t have Mythos 2 access, you *have to* use OpenMythos 2 and send your data to China!
@BlackHC @xeophon @sdmat123 It only helps with the US falling behind as the rest of the world will just adopt open-source and render frontier labs obsolete.
Just because it doesn't solve all issues doesn't mean that a partial solution is not a step in the right direction
If all you can do is local regulation that already helps against many bad cases
You can also make access in foreign data centers illegal and/or block their services
Protecting against bad actors elsewhere is not solved by this but it creates an obstacle
This is the same with many other regulated services
@BlackHC @sdmat123 So the threat actor you outline is someone technical enough to fine-tune big models but not technical enough to use a foreign DC?
The bad cases are jailbreak prompts in Claude Code, not people fine-tuning trillion parameter models
@xeophon @BlackHC Benjamin Franklin's famous saying comes to mind.
Relying on global suppression of a mathematical concept is not a great plan.