In my experience, LW and AI safety gain a big chunk of legitimacy from being the best at Rationality and among the best places on earth for self-improvement. That legitimacy goes a long way, but only in systems that are externalities to the alignment ecosystem (i.e. the externality is invisible to the 300 AI safety researchers who are already being AI safety researchers).
I don’t see the need to retool rationality for alignment. If it helps directly, it helps directly. If it doesn’t help much directly, then it clearly helps indirectly. No need to get territorial for resources, not everyone is super good at math (but you yourself might be super good at math, even if you think you aren’t).
In my experience, LW and AI safety gain a big chunk of legitimacy from being the best at Rationality and among the best places on earth for self-improvement. That legitimacy goes a long way, but only in systems that are externalities to the alignment ecosystem (i.e. the externality is invisible to the 300 AI safety researchers who are already being AI safety researchers).
I don’t see the need to retool rationality for alignment. If it helps directly, it helps directly. If it doesn’t help much directly, then it clearly helps indirectly. No need to get territorial for resources, not everyone is super good at math (but you yourself might be super good at math, even if you think you aren’t).