Divergence uses a nabla not a delta

To clarify: "coherence" here means that your credences obey the probability axioms?

I wrote this tl;dr for a friend, and thought it worth sharing. I'm not sure it's accurate. I've only read the "Recap"

Here is how I understand it.

Suppose that, depending on the temperature, your mirror might be foggy and you might have goose pimples. As in, the temperature helps you predict those variables. But once you know the temperature, there's (approximately) nothing you learn about the state of your mirror from your skin, and vice versa. And! Once you know whether your mirror is foggy, there's basically nothing left to learn about the temperature by observing your skin (and vice versa).

But you still don't know the temperature once you observe those things.

This is a stochastic (approximate) natural latent. The stochasticity is that you don't know the temperature once you know the mirror and skin states.

Their theorem, iiuc, says that there does exist a variable where you (approximately) know its exact state after you've observed either the mirror or your skin.

(I don't currently understand exactly what coarse-graining process they're using to construct the exact natural latent).

(Remember that, IIRC, we still have the misfeature that you can't strong upvote your own comments. Perhaps you mention this, I haven't read much of your comment or these threads)

This strikes me mostly as an argument for cheaper housing!

Curated. I thought this was a pretty interesting result. I'm not sure if I should have been surprised by it, but I was. They also point to a decent amount of interesting follow-up work, though I expect that to generalise less well than "existence proof" papers like this one.

I often enjoy this group's work finding interesting "info leaks" in LLM behaviour, like their previous lie detector work.

If by "very widespread" you mean like ~10% of votes, I disagree. Do you mean that?

If only a handful of people did as you propose to do—then much of the usefulness would be lost, though not most.

If by "much of the usefulness would be lost" you mean something like "people would see comments that they liked <90% as much" or "people would get less than 90% of the information about what some kind of weighted LessWrong-consensus thought", I disagree. Do you mean that?

The obvious consequence of such a norm is comments having to say things like “don’t upvote this comment too much, because otherwise you will be robbing me of replies which I would otherwise get” or “please downvote this comment, so that it gets pushed down, so that I can get people replying to me, instead of staying silent because of a weird and incidental fact of comment section sorting”. This would be very bad, obviously.

I agree it's obvious that it at least pushes some in this direction. I think some versions of this could be very bad, though mostly it would be not that bad.

The things which you are trying to do with the karma system would destroy its usefulness.

By "destroy its usefulness", how much less useful do you mean to say it would become?

Comments that hijack an unrelated thread can be downvoted (and thus hidden), or their authors censured for abusing the commenting system. This is a non-problem on Less Wrong.

I didn't mean, in that comment, to imply that Ben was hijacking. I was just trying to provide at least one example of a pathological interaction with threading and karma.

Having refreshed myself on Ben's comment and its parent, I now think he was doing something continuous with hijacking.

I think you're asking if the whole mod team agrees with "the LW karma system is NOT robust, well-implemented and generally used very properly by users".

I think in general the LW team thinks that the karma system is generally used properly by users (not sure about "very properly", for example I think we're probably not skilled enough, as a userbase, at using it. Habryka might even disagree with "properly", because he so strongly wants more. downvotes).

I don't know what opinions people have about the implementation. I think, for example, most people on the team think that agreement voting is quite good, that having weak/strong votes is good, and that our vote scaling is good.

For "robust", I think most people think it fails sometimes on the actual website, and not just in possible corner cases.

I think your comment was a little bit "cheating" against LW's systems, and thus deserving of a little downvote. I don't know if a norm exists against this kind of cheating, but I think it should.

IIRC, I kinda perceived that you were trying to pushback against a general vibe spread throughout the comment section. Your comment is basically not engaging with cata's comment at all. You reference the video, which cata doesn't, and you reference "believing everyone is doing the best they can", which is not something cata says. You were pushing against the general zeitgeist, and you did it in a way that uses a quirk of the commenting system to give it prominence.

I think you should have written a top-level comment pushing back against the other comments, perhaps linking to them. And then the karma system could have buoyed it to the top, or not.