Project Idea: Challenge Groups for Alignment Researchers
post by Adam Zerner (adamzerner) · 2023-05-27T20:10:12.001Z · LW · GW · 0 commentsContents
No comments
@Henrik Karlsson [LW · GW] writes a lot about what makes people successful. One recurring theme I've found involves the people around them. Successful people almost always have a group of friends who they're constantly chatting with, bouncing ideas around with, learning from, pushing each other, supporting each other, etc. A group who is similarly talented and driven.
Consider the example of MrBeast:
MrBeast: I somehow found these other four lunatics. Three of us were college dropouts. One was a high school dropout. And one, he just like quit his job. We were all super small YouTubers.
And we basically talked every day for a thousand days in a row and did nothing but hyper-study what makes a good video. [. . .] Some days I'd get on Skype at 7 am and I’d be in the call until 10 pm and then I’d go to bed. I wake up and I do it again.
He had summoned a scene.
MrBeast: We all had like 10-20,000 subscribers when we met. And by the time we stopped talking, we all had millions of subscribers. We all hit a million subscribers within a month of each other.
There was, in other words, little luck involved. They collectively extracted the know-how needed to consistently do what they wanted to do, which was to go viral.
Or consider musicians in Sweeden. Or intellectuals who grew up in exceptional milieus [LW · GW].
I like to think of this as a "challenge group". It seems like a pretty huge force multiplier.
I'm not sure about the extent to which alignment researchers currently lack challenge groups, but to the extent they do, helping form them seems like a pretty impactful project. It also strikes me as something that is, and is likely to remain, moderately neglected.
Similarly, you could also help form challenge groups in other areas: aspiring alignment researchers, cryonicists, EAs more generally, rationality dojos. But to increase the likelihood that new members are matched with existing members for whom they're compatible, it's probably a good idea to start off with a somewhat narrow group.
I have more thoughts on this that I'm struggling to formulate, so I'll leave it here for now. In general I think @Henrik Karlsson [LW · GW] has a lot of great things to say on this topic.
0 comments
Comments sorted by top scores.