Roman Ross

This is a timed post. Every 5 minutes while writing this post, I had to stop to do 15 push-ups, and when I could no longer complete my required number, I had to upload it. My friends thought this would be a fun challenge, but it means that it will likely be less polished than some of my other output. Summary: There has been a lot of talk in the EA/AI safety community recently about what it means to be a “generalist,” but many disagree on what this means. In this post, I give my take on the role: a generalist is best understood as someone who does whatever needs to be done to solve a set of problems, rather than as a specific career role. What is a generalist? There’s been a lot of talk lately about the need for “generalists” in AI safety. However, people across the space seem to have different understandings of what it means to be a generalist. Some seem to believe that “generalist” is just a weird title to describe a lump of roles that don’t fit anywhere else. Others seem to think it’s a synonym for “fieldbuilder” or “impact multiplier.” While both of these definitions gesture at what a generalist is, I think it’s wrong to conceive of a generalist as a specific set of tasks. Instead, I think that a generalist is someone who looks for the most important problems to solve and shapes themself across a broad range of skills to address those problems. I don’t think of this as a definitional shift for what it means to be a generalist. Instead, I view “being a generalist” as a disposition to seeking out the most impactful thing across a broad range of skills, similar to how “being an optimist” is a disposition to thinking good things will happen in the future. This might mean that you can’t easily picture a pig wearing clothes in a children’s book being a generalist, but I think that’s okay. Success as a generalist requires agency. Ben Kuhn has my favorite explanation of what this means[1]: agency is working backwards from your end goal and doing everything