#theor_note #theor_aisafety

"The closed source frontier is ~6 months ahead of the best open weights model;" "Even afterward, the offense-defense balance will lean in our favor." (https://www.astralcodexten.com/p/open-questions-on-open-weights)

This seems unjustified, and I'm concerned if broad AI community is persuaded by this claim. AI is more akin to missiles or misinformation than trench warfare or vaccination (offense / defense -dominated respectively).

I'm concerned AI safety orgs don't near-unanimously take a stance that open source models need to be jointly regulated by China and US. I understand the personal freedom concerns, but saving lives should be a priority.