CFAR’s focus on AI research (as opposed to raising the rationality water line in general) leads me to two questions:
Given the long feedback loops involved with AI alignment: Wouldn’t it be better to concentrate at first on other target populations (not AI research) to validate rationality techniques (even if, in the end, the aim is to deploy those techniques/attitudes in the context of alignement)?
Even if it were possible to increase the rationality in parts of the AI research area, in democracies wouldn’t it still be very important to target other parts of society as well with interventions to increase collective rationality? Because, I think, in the end there has to be some sort of regulation even after AI alignment were theoretically solved to solve it in practice.
Based on that: Shouldn’t it be an important goal to test and popularize rationality techniques outside of subcultures in AI research if one wants to solve the alignment problem in practice? (Whether that is a job for CFAR or someone else is a different question, of course).
CFAR’s focus on AI research (as opposed to raising the rationality water line in general) leads me to two questions:
Given the long feedback loops involved with AI alignment: Wouldn’t it be better to concentrate at first on other target populations (not AI research) to validate rationality techniques (even if, in the end, the aim is to deploy those techniques/attitudes in the context of alignement)?
Even if it were possible to increase the rationality in parts of the AI research area, in democracies wouldn’t it still be very important to target other parts of society as well with interventions to increase collective rationality? Because, I think, in the end there has to be some sort of regulation even after AI alignment were theoretically solved to solve it in practice.
Based on that: Shouldn’t it be an important goal to test and popularize rationality techniques outside of subcultures in AI research if one wants to solve the alignment problem in practice? (Whether that is a job for CFAR or someone else is a different question, of course).