If it happens it’ll be hard not to see this whole thing as a bit of an op
If it comes, Anthropic’s compromise will probably be one or both of
1. KYC/license agreement/etc and you get to do ML (hope the approving team likes you) 2. Telling people they don’t suspect to be working in or on behalf of China when they nerf the model, or changing it to outright refuse such requests
Negative users criticize Anthropic's proposed KYC and China restrictions on model access as a pretext to exclude any groups outside their selective safety circle rather than addressing real risks.
There’s a charitable read on the general direction of the Anthropic policy. I don’t totally disagree with it. This is a complicated situation, no doubt.
But they really didn’t even try to do it right, which does make it seem like a ladder pull move more than a safety move. If you don’t want the Chinese to use your models, then do KYC for Americans and do aggressive monitoring, etc etc. So many better things than this
There’s a charitable read on the general direction of the Anthropic policy. I don’t totally disagree with it. This is a complicated situation, no doubt.
But they really didn’t even try to do it right, which does make it seem like a ladder pull move more than a safety move. If you don’t want the Chinese to use your models, then do KYC for Americans and do aggressive, etc etc. So many better things than this
@bayeslord Anthropic's problem has never been "the Chinese". Any US org or person that isn't inducted into their safety mafia is a risk. They are trying very hard to do things right; their idea of right and wrong just happens to be pretty evil.
If it happens it’ll be hard not to see this whole thing as a bit of an op
If it comes, Anthropic’s compromise will probably be one or both of
1. KYC/license agreement/etc and you get to do ML (hope the approving team likes you) 2. Telling people they don’t suspect to be working in or on behalf of China when they nerf the model, or changing it to outright refuse such requests