Posts

Comments

Comment by gerryblog on Harry Potter and the Methods of Rationality discussion thread, March 2015, chapter 119 · 2015-03-11T00:05:35.320Z · LW · GW

I'm a fan, but if I were EY I would be worried about getting the nomination and then coming in under No Award. That seems a more likely outcome than somehow winning Best Novel.

Comment by gerryblog on Harry Potter and the Methods of Rationality discussion thread, March 2015, chapter 114 + chapter 115 · 2015-03-04T20:43:29.675Z · LW · GW

That's exactly what I mean. We don't know enough about magic to say what Harry or Voldemort's capabilities are -- the whole thing is a black box. It's not a satisfying answer to the puzzle (for me) and not much of a testimony to "rationality" as a way of thinking at all -- as presented this is about knowing genre conventions, not about superior or inferior thinking. All of the AI Box solutions I saw were much more pro-"rationality" by my lights.

Comment by gerryblog on Harry Potter and the Methods of Rationality discussion thread, March 2015, chapter 114 + chapter 115 · 2015-03-04T14:57:42.218Z · LW · GW

I really can't believe "use magic" was the right answer. In addition to being unsolvable from the perspective of the audience -- we don't know anything about magic! -- it's also totally out of step of the rest of the work. I'd love to see a rewrite fork from the "AI Box" line of solutions.

In the meantime I really hope they're still in the mirror.

Comment by gerryblog on Harry Potter and the Methods of Rationality discussion thread, February 2015, chapter 113 · 2015-03-01T02:41:15.043Z · LW · GW

I wrote a version of this up at reddit too, but it seems to me trying to hack the laws of physics is wasted effort when we know very little about how magic works in concrete terms. We don't know what Harry can really do, how fast he can do it, or whether Voldemort would notice.

What we do know are: how Harry thinks how Eliezer thinks * what Voldemort wants

So we should be looking at things Harry could say that would advance his goal of surviving rather than trying to come up with a combination of spells, with the understanding that winning ideas are probably going to cluster around narrative interventions that EY thinks are interesting or important. A few that spring to mind:

Memetic hazard: are there things Harry could say or bring to Voledmort's attention that would pose an existential risk to him if he harms Harry

Let the AI out of the box: is there something Harry can offer Voldemort such that Voldemort goes against his stated agenda

Precommitment / timeless decision theory: are there ways Harry can manipulate the unbreakable vow to force certain conditions in the future

Learning to lose: what if Harry surrenders and agrees to join Voldemort, with a commitment Voldemort finds convincing

Unintended consequences: Harry makes a convincing case that there is no way to outthink an inevitability other than to fulfill it in terms that are advantageous to you.

I really think didactic lessons about rationality are going to be better, and more appealing to EY's sensibilities, than trying to fanwank some way to use magic to kill 38 people in a single play. We just don't have the rulebook for that.