Will Brown suggested labs also stop promoting recursive self-improvement
Easy solution to slow down recursive AI self improvement:
- The lab with the top-ranked model must agree THEY must not use it for working on frontier AI - But everyone else should have access to it.
By definition, this means the frontier doesn't advance.
Positive users support the proposal restricting top AI labs from using their own models to slow self-improvement because it avoids power imbalances and favors open-sourcing and democratizing AI.
It also has the critical benefit of avoiding a dangerous power imbalance.
Easy solution to slow down recursive AI self improvement:
- The lab with the top-ranked model must agree THEY must not use it for working on frontier AI - But everyone else should have access to it.
By definition, this means the frontier doesn't advance.
@jeremyphoward they probably also shouldn’t release a big blog saying “btw we’re doing recursive self improvement now it’s sick as hell haha”
Easy solution to slow down recursive AI self improvement:
- The lab with the top-ranked model must agree THEY must not use it for working on frontier AI - But everyone else should have access to it.
By definition, this means the frontier doesn't advance.
@finbarrtimbers If they believed that, they'd be doing the opposite of what they chose.
Easy solution to slow down recursive AI self improvement:
- The lab with the top-ranked model must agree THEY must not use it for working on frontier AI - But everyone else should have access to it.
By definition, this means the frontier doesn't advance.
Anthropic has chosen the *opposite* of the safe path: they are allowing themselves, the current top lab, to use their top model for frontier AI research. They've said they'll sabotage others who try.
This means the AI frontier advances, & power imbalance increases.
(To be clear, *I* don't think we should try to slow down recursive AI self improvement - I think we should open it up and democratize it as much as possible. My point is: if *you* claim we should slow down, and you have the best model, you should ensure your org can't use it.)
@jeremyphoward What is the incentive to train the next frontier model then
@jeremyphoward And if they don't, then it's just another case for why:
https://opensourceaimustwin.com/?share=v2
@jtlicardo If your goal is to slow down frontier AI development, then you don't want create that incentive.
@jeremyphoward That’s why AI should be open source, so that there is a check and balance in power / knowledge
@jeremyphoward I'm not sure how this would be enforceable. What would the mechanism be?
@jeremyphoward Why the second point?
Why does that slow down rsi?
Will Brown suggested labs also stop promoting recursive self-improvement
Easy solution to slow down recursive AI self improvement:
- The lab with the top-ranked model must agree THEY must not use it for working on frontier AI - But everyone else should have access to it.
By definition, this means the frontier doesn't advance.