[sigil]

Ok but isn’t this the opposite of OpenAI’s “nukes are safer when multiple actors have them” strategy wrt AI?

I’m also confused by the threat models earnestly put forth in your blog post. Are we really concerned about deep faking someone’s writing? The plain word already demands attribution by default: we look for an avatar, a handle, a domain name to prove the person actually said this.

[albntomat0]

> Ok but isn’t this the opposite of OpenAI’s “nukes are safer when multiple actors have them” strategy wrt AI?

It seems more like the "nukes are safer when multiple rational state level actors have them", rather than anyone able to pull a git repo.

[sigil]

Yep. Maybe I misunderstood the subtler points of OpenAI’s “democratize AI” strategy, and this has been the plan all along. But AFAIK they haven’t put an “among a few rational state actors” asterisk on anything up until now.

Regardless, I agree with TFA that this is a silly and arbitrary time to yell “fire.” It’s PR.

[albntomat0]

> But AFAIK they haven’t put an “among a few rational state actors” asterisk on anything up until now.

True. On the PR side though, it'd be incredibly hard to say "we want to make replication moderately difficult, but not too difficult." Everyone would end up arguing exactly how much should be released, how it would prevent X,Y,Z folks from contributing to AI, etc.

> Regardless, I agree with TFA that this is a silly and arbitrary time to yell “fire.” It’s PR.

Alternatively, it does provide good insight into the reactions in the community as a whole, and continues the conversation on exactly how much should be released. Maybe I'm not far enough into the ML community, but the decision not to put the "keys to the kingdom" on github for every script kiddie to weaponize seems reasonable to me, especially as a precedent.

[Cacti]

More like “nukes are safer when we control them and the rest of you cite them”