Hm a bit, although I don’t think there’s a non-continuous break with it.
pku
Karma: 114
[Story] Human Alignment Isn’t Enough
Interesting. Do you know if it’s possible it looked up a solution someone posted online?
> A sincere effort to figure out the features of X and replicate them without some undesired property Y, seems like a good thing to try
I agree that it’s a good thing to try, but it often (especially in early efforts) end up compromising on some of the things that make X good (which makes sense, if X wasn’t optimizing for avoiding Y it’s probably easier to optimize it for its primary goals).. Sometimes Y is bad and these are worthy compromises, and sometimes after some early mistakes you can make the disadvantages go away, so it’s still often worth putting effort into despite the downsides.
I run into this in coding a lot too—I’ve found that I can consistently get better results from Claude by telling it to rewrite the code more concisely before I go through it (which makes me wonder why Anthropic hasn’t already tried to engineer this into Claude Code via prompt or something).
AI slop is a vegan hamburger
Translating this to the mental script that works for me:
If I picture myself in the role of the astronauts on the Columbia as it was falling apart, or a football team in the last few minutes of a game where they’re twenty points behind, I know the script calls for just keeping up your best effort (as you know it) until after the shuttle explodes or the buzzer sounds. So I can just do that.Why is there an alternative script that calls to go insane? I think because there’s a version that equates that with a heroic effort, that thinks that if I dramatize and just try harder (as shown by visible effort signalling), that equates with making a true desperate effort that might actually work in a way that just calmly doing my best to the end won’t. But since I know that script is wrong, I can just not play it.
(Why does that script exist? I think for signalling reasons—going insane over something is a good way to shallowly signal I think it’s significant. But it’s not a good way to solve the underlying problem when it’s the underlying problem that needs solving, so I just choose not to do it when that’s the case.
A similar example: If I imagine seeing a news article about a child going missing, it’s easy for me to picture myself remarking “oh that’s terrible, I’m crying just imagining the parents”. If I imagine a child of mine or of a close friend going missing, my mental script’s next step is “okay track down where he was, call the police, think of more action steps”. Because there I care more about finding the child than about signalling that I care about finding the child).
After thinking about this a bit more, this says something about Voldemort’s perfect occlumancy in that while he could convincingly become a lot of different characters, there were still parts of his underlying person (CEV?) that genuinely were so unlike other people he couldn’t easily fake not having them.
(Which I guess is semi-confirmed in-story in that it says occlumancy and veritaserum enters the mind through specific surfaces that can be defended, but powerful methods like unbreakable vows/parseltongue/the mirror can’t be blocked with it)
Iirc when they discover filch is a squib Ron explicitly says this is what a squib is (“like muggle born wizards, but in reverse and much rarer”).
Another nitpick, but footnote 14 is wrong: Tonks was a metamorphmagus, not using polyjuice.
Actually now I want to write a spin off short story about this. It always bothered me that hpmor Dumbledore didn’t really get any onscreen moments of awesome, I want to write a story set early in the Voldemort war where, for the first time, a relatively young inexperienced Voldemort has someone pull a successful plot against him by being proactive and he barely escapes (and rewrites some of his list with lessons).
Minor comment but Quirrel doesn’t have a time turner; he just figured out a way to hack the protective shell around Harry’s (which, given it was made to protect against interference by an eleven year old, it’s pretty reasonable he can do).
My open question is: how did Dumbledore know Voldemort was Tom Riddle? When he set himself up as dark Lord he presumably didn’t intend to leave a trace (he didn’t with his David Monroe persona). Did Dumbledore discover it in some plot? Did he conclude it because Voldemort openly use parseltongue somewhere? Was it in the prophecies?
Manhattan – ACX Meetups Everywhere Spring 2025
This assumes speed limits were correctly calibrated at some point. I think the actual cost of road deaths (which are arguably the top single cause of QALY loss even at current historically low rates) is high enough that I suspect it was originally set way too high and is still unreasonably high given the costs.
I think this describes how Eliezer’s grudge against academia has set back AI alignment (even the parts that aren’t related to his organization, since his cultural influence has made this a wider norm).
Partly street design to reduce speeding, partly encouraging other mode shares over private cars (see e.g. here https://www.theguardian.com/world/2020/mar/16/how-helsinki-and-oslo-cut-pedestrian-deaths-to-zero )
(Disclaimer: I’m ideological about disliking cars, which makes me less objective than I’d usually prefer to be on LW)
We haven’t solved traffic or auto accidents
Worth noting that this is a policy failure, not a technological one. Some places have solved this—e.g. Oslo has 0-1 car deaths a year—but American cities are unwilling or unable to make the infrastructure changes it takes. I think this is related to the lack of celebration issues—we celebrate change and progress less, and achieve less progress that would require change, because we value progress less than we used to.
I think I was at about the same place for most of it, but unfortunately I didn’t write that one down and can’t go back and check :/.
There’s also a piece here of expressing “I didn’t go to the right because I was biased in its favor, I has the opposite bias and it took a lot of evidence to shift me over, which is meta-evidence in favor of my position”
(I also don’t generally see “The left went crazy and drove me to the far right”—usually it’s said by someone who went over the line from moderate left to moderate right. People who jump between extremes usually have different arguments)