I think you’re missing the point of the Many Worlds posts in the Sequences. I’ll link to my response here.
Regarding HPMoR, Eliezer would agree that Harry’s success rate is absurdly unrealistic (even for a story about witchcraft and wizardry). He wrote about this point in the essay “Level 2 Intelligent Characters”:
In the real world, everything is harder than it is for characters in stories; clever insights are less likely to be true and clever strategems are overwhelmingly less likely to work. In real life, I have to try literally ten good ideas before one of them works at all, often putting in years of effort before giving up or succeeding. Yes, I’ve been known to pull off implausible tricks like “Write a Harry Potter fanfiction good enough to recruit International Mathematical Olympiad gold medalists” but that’s not the only implausible-sounding thing I’ve ever tried to do. You just don’t hear as much about the clever ideas that didn’t work, over the many years I’ve been trying weird and nonweird ways to get my task done.
In fiction you as the author can decide that the bright ideas do work, being careful to accompany this by an appropriate amount of sweat and pain and unintended consequence so that the reader feels the character has earned it. You cannot evade the curse of building your story out of clever ideas that would be far less likely to work in real life, not just because you have no way to test the ideas to find the ones that actually work, but because in real life we’re talking about a 10:1 ratio of failures to successes. We get to see Harry fail once in Ch. 22, because I felt like I had to make the point about clever ideas not always working. A more realistic story with eight more failed ideas passing before Harry’s first original discovery in Ch. 28 would not have been fun to read, or write.
I would agree with you, however, that HPMoR lets Harry intuit the right answer on the first guess too much. I would much prefer that the book prioritize pedagogy over literary directness, and in any case I have a taste for stories that meander and hit a lot of dead ends. (Though I’ll grant that this is an idiosyncratic taste on my part.)
As a last resort, I think HPMoR could just have told us, in narration, about a bunch of times Harry failed, before describing in more detail the time he succeeded. A few sentences like this scattered throughout the story could at least reduce the message to system 2 that rationalist plans should consistently succeed, even if the different amounts of vividness mean that it still won’t get through to system 1. But this is a band-aid; the deeper solution is just to find lots of interesting lessons and new developments you can tell about while Harry fails in various ways, so you aren’t just reciting a litany of undifferentiated failures.
There’s a difference between succeeding too often and succeeding despite not testing his ideas. The problem isn’t having too many failed ideas, the problem is that testing is how one rules out a failed idea, so he seems unreasonably lucky in the sense that his refusal to test has unreasonably few consequences.
I think you’re missing the point of the Many Worlds posts in the Sequences. I’ll link to my response here.
Regarding HPMoR, Eliezer would agree that Harry’s success rate is absurdly unrealistic (even for a story about witchcraft and wizardry). He wrote about this point in the essay “Level 2 Intelligent Characters”:
I would agree with you, however, that HPMoR lets Harry intuit the right answer on the first guess too much. I would much prefer that the book prioritize pedagogy over literary directness, and in any case I have a taste for stories that meander and hit a lot of dead ends. (Though I’ll grant that this is an idiosyncratic taste on my part.)
As a last resort, I think HPMoR could just have told us, in narration, about a bunch of times Harry failed, before describing in more detail the time he succeeded. A few sentences like this scattered throughout the story could at least reduce the message to system 2 that rationalist plans should consistently succeed, even if the different amounts of vividness mean that it still won’t get through to system 1. But this is a band-aid; the deeper solution is just to find lots of interesting lessons and new developments you can tell about while Harry fails in various ways, so you aren’t just reciting a litany of undifferentiated failures.
There’s a difference between succeeding too often and succeeding despite not testing his ideas. The problem isn’t having too many failed ideas, the problem is that testing is how one rules out a failed idea, so he seems unreasonably lucky in the sense that his refusal to test has unreasonably few consequences.