The technical content is pretty much readable to schoolchildren, yet somehow readers are convinced that something magical is going on and they can contribute, not just read and learn.
I don’t think that’s a matter of writing style. It’s a matter of whether the prospective “research area” is simple enough that all of its general prerequisites can be stated in a popular blogpost, and otherwise be assumed to be known to the reader. (For example, many OvercomingBias/LessWrong readers have enough of a background in rational-action theory to know what “precommitment” and “dynamic inconsistency” mean, and these notions are indeed necessary for a proper understanding of EY’s point.) At one point, that was true of the general area of timeless/updateless decision theory. It seems to be less true of the logical induction problem.
I think logical induction could’ve been popularized with just as much effort (that is, a lot). For example, the second problem from the post linked by endoself was discussed by Wei and me in 2012, with >40 comments each. If we’d been better at mass appeal, instead of coasting on the audience attracted by Eliezer, we could’ve had even more engagement. (Note the comment from thescoundrel in the second link, that’s the kind of good idea out of nowhere that mass appeal is all about.)
Does popularization produce the goods? Lots of people have the background and skill to contribute to this problem who aren’t currently in our community and don’t have day jobs.
Choosing the right problem is certainly important, but I don’t think it’s the bottleneck. There’s plenty of low hanging fruit. Knowing how to play your audience seems like more of a bottleneck, and it takes a lot of effort to learn.
I don’t think that’s a matter of writing style. It’s a matter of whether the prospective “research area” is simple enough that all of its general prerequisites can be stated in a popular blogpost, and otherwise be assumed to be known to the reader. (For example, many OvercomingBias/LessWrong readers have enough of a background in rational-action theory to know what “precommitment” and “dynamic inconsistency” mean, and these notions are indeed necessary for a proper understanding of EY’s point.) At one point, that was true of the general area of timeless/updateless decision theory. It seems to be less true of the logical induction problem.
I think logical induction could’ve been popularized with just as much effort (that is, a lot). For example, the second problem from the post linked by endoself was discussed by Wei and me in 2012, with >40 comments each. If we’d been better at mass appeal, instead of coasting on the audience attracted by Eliezer, we could’ve had even more engagement. (Note the comment from thescoundrel in the second link, that’s the kind of good idea out of nowhere that mass appeal is all about.)
Does popularization produce the goods? Lots of people have the background and skill to contribute to this problem who aren’t currently in our community and don’t have day jobs.
Choosing the right problem is certainly important, but I don’t think it’s the bottleneck. There’s plenty of low hanging fruit. Knowing how to play your audience seems like more of a bottleneck, and it takes a lot of effort to learn.