Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Awaclus]

Yeah; there's also been experiments with people grabbing objects you hold somewhere in front of them. No way you could do that with sound.

Well, it makes a sound when you extend your arm. If it's a quiet environment, I would expect blindfolded people with perfect vision and zero ear training to do much better than chance at that task. Something like being able to describe a picture on a computer screen, or being able to choose the brightest colored object out of multiple otherwise identical ones would be a lot more convincing.

[silverspawn]

I would expect blindfolded people with perfect vision and zero ear training to do much better than chance at that task.

That's interesting because I would very much not expect this. I think picking out the brightest object or describing a picture are both too difficult.

But I think recognizing shapes has been done, and probably they've been presented on a screen. I'll keep the "make sure sound can't play a role" criterion in mind while diving more into it

[silverspawn]

(No connection between the first two sentences in my reply, there should have been a paragraph break)

[Awaclus]

I would expect blindfolded people with perfect vision and zero ear training to do much better than chance at that task.

That's interesting because I would very much not expect this. I think picking out the brightest object or describing a picture are both too difficult.

But I think recognizing shapes has been done, and probably they've been presented on a screen. I'll keep the "make sure sound can't play a role" criterion in mind while diving more into it

Colors and brightness are basically the only things that you can only detect with vision. Something like "there's a spotlight here, what direction is it facing?" would probably work well too, because even if the lamp makes a sound, I can't imagine its orientation being distinguishable unless you can see the light.

I don't think even people who were born blind could tell the shape of a relatively small object (like 20-30 cm in diameter) from several meters away, but it would obviously be simple for a machine using echolocation, so I sure wouldn't bet my life savings against there being a person who can do that.

[silverspawn]

Men (2022)

So this is a horror movie where everything is a metaphor about the behavior of men toward women.

And as far as gender critique goes, this is definitely the best movie I've seen. Unlike with Barbie, where I thought all the critique was just so obvious and boring, here the critique it makes is much more narrow, and also feels much more like something that deserves to be said.

I didn't watch this because of YMS; I watched it because I looked on Wikipedia what the director of Annihilation had made since. But YMS did review it; rewatching it now, he says that this movie is much more interpretable than the directors previous movies -- but I disagree. Yes, everything here is metaphorical, but it's really clear what the metaphor is. I don't see anyone convincing me that it means something else. Whereas Annihilation I think is also metaphorical, much more than you'd think, and I didn't get that on my first watch.

Anyway, what I didn't like so much is how gross it was. I like horror; I don't like gross-ness. Although I do admit that it has a thematic purpose; in fact, I can't really name another movie where being gross is more justified by the larger point. So I guess I can't complain too much.

Sort of feels like a 7 right now but I think this is the kind of movie I'll like more with time, so I'll update pre-emptively and give it an 8

[silverspawn]

And I gotta say it's very cool that Garland made this. Lots of respect to him.

[silverspawn]

I would expect blindfolded people with perfect vision and zero ear training to do much better than chance at that task.

That's interesting because I would very much not expect this. I think picking out the brightest object or describing a picture are both too difficult.

But I think recognizing shapes has been done, and probably they've been presented on a screen. I'll keep the "make sure sound can't play a role" criterion in mind while diving more into it

Colors and brightness are basically the only things that you can only detect with vision. Something like "there's a spotlight here, what direction is it facing?" would probably work well too, because even if the lamp makes a sound, I can't imagine its orientation being distinguishable unless you can see the light.

I don't think even people who were born blind could tell the shape of a relatively small object (like 20-30 cm in diameter) from several meters away, but it would obviously be simple for a machine using echolocation, so I sure wouldn't bet my life savings against there being a person who can do that.

I'm not exactly sure I understand your point; surely, if someone can tell apart a square from a triangle that's shown on a screen, that requires visual input. Sound isn't going to hepl there.

[silverspawn]

I just like movies that actually mean something rather than being deliberately vague so you can interpret whatever

[Awaclus]

I'm not exactly sure I understand your point; surely, if someone can tell apart a square from a triangle that's shown on a screen, that requires visual input. Sound isn't going to hepl there.

That's because it has to do with brightness and color (of pixels on a screen).

[silverspawn]

ok fair enough

[silverspawn]

So yeah, those experiments have been done

[Awaclus]

So yeah, those experiments have been done

[silverspawn]

Apropos of nothing, the fact that the creator of bitcoin is still anonymous is kind of incredible. It feels much more like something that would be a plot point in a novel rather than something that would happen in the real world. Especially for something as impactful as the thing that kicked of cryptocurrency.

[silverspawn]

Sometime during the past year, the claim "LLMs will not scale to AGI" seems to have gone from "the uncool position that contradicts what Eliezer and gwern believe" to "the majority position that it's acceptable to voice" even though I haven't seen many people publicly change their mind. (In the rationalist sphere that is; outside idk what people thought but also who cares.) Unfortunately it seems like companies have realized this as well are not, in fact, just working on GPT-5. This sort of puts us back to where we were before before, which is that we don't know how to build AGI. Well, except that we now have LLMs as a tool.

... and also the fact that LLMs can now be incorporated into AGI designs. Which, unfortunately, I think has a chance to work.

[silverspawn]

I still weakly hold that LLMs in and of themselves are probably a net positive. If we could somehow only get current applications without applications for AGI systems, I'd be for it. But they come as a package and (surprise) I don't think it's a good package.

[silverspawn]

What I really want to know is, did the bitcoin author name themselves Satoshi Nakamoto because they're Japanese, or because they think Japanese culture is cool?

[silverspawn]

Also apropos of nothing, I think I've never mentioned Shinzen Young, so I should probably do that. I think of him as basically the most enlightened person I know of. If I had to point to someone to advertise what mindfulness can do, he'd be it. Everything I've seen and heard from him seems consistent with him being basically not ever suffering and also being completely ethical. Also everything he says about meditation seems like he's speaking with incredibly expertise. And he's also quite smart, but if anything causality goes the other way there.

E.g., https://www.youtube.com/watch?v=pmR6SepZlwY

Unfortunately he's also very old now, but what can you do.

[silverspawn]

And it goes... well actually it really doesn't go without saying now that I think about it, but either way, he's also super humble

[silverspawn]

Wtf why does the bitcoin forum look exactly like the dominion forum this is weird af

[silverspawn]

aw

[silverspawn]

Death Note (2007)

I remembered this as one of the best written shows out there. Rewatching it, it's both better and worse than I thought. (Not gonna use spoilers because cmon.)

At this core, I think the construction is even more brilliant than I'd remembered. Light and L are both great characters, but I didn't quite appreciate how the story is constructed such that they confront each other so directly. This is of course what you want to do to make it maximally cool, and the story really pulls it off. The way it happens feels completely believable.

I also didn't appreciate how much of an idiot Light is. You know what I'd do to confront the police if I were in his shoes? Nothing. N-o-t-h-i-n-g. Because while I might be far less brilliant and have nowhere near the same work ethic, I understand the basic fact that finding me requires about 30 bits of information that are impossible for investigators to acquire unless I do something to help them. It literally doesn't matter how smart the investigators are; their intelligence only matters as a function of how many of the bits I reveal they actually receive, but if I don't reveal bits at all, they can't do anything. I wouldn't provoke the police, I wouldn't kill any of them; I would just figure out a way to randomize victims and follow it methodically. And then the police would make no progress and we'd have no story. The story only works because of Light's character.

Like man, killing the FBI agent who was investigating you is just so unfathomably stupid. Seriously, the degree to which this is stupid is hard to put into words.

Eliezer has talked about this thing he calls security mindset, which at its core means always acting such that the story of why your thing works is as simple as possible. This basically sums up everything that Light doesn't understand.

But it's interesting because L really doesn't have the same weaknesses. Despite the show portraying them as equals, I think he's just straight up smarter. He's got comparable analytical skills without the weaknesses.

Anyway, the first 12 or so episode are really amazing. I think I've praised L before but it's really hard to exaggerate how good of a character he is. The big thing is of course that all of his deductions actually make sense. But then there's all the details like his posture or the whole thing with constantly eating sweets. And I love that he gets depressed and demotivated during the middle arc, it makes so much sense. And I like how genuine he is; there's no hiding behind irony.

The thing where he shows Light three pieces of a message that criminals have left before dying, asks Light to interpret how they fit together, but then claims that Light failed because there's actually a fourth (in which fact L made up) -- that bit is absolutely brilliant. That feels like something out of Methods of Rationality.

He shouldn't be amazing at Tennis though, that's stupid. And neither should Light.

The part after L dies is worse than I remembered. I still have the opinion that I expected to have, which is that people exaggerate how bad it is, when in fact it's still better than most other shows. So yeah, that's still true. But the writing is noticeably worse even if we don't care about the characters. There are also a few plot details that I didn't understand, which could be my fault, but I understood everything up to that point, so if they're not plot holes, then the communication wasn't as good. Also a few instances of weirdly bad music in the second half.

I also noticed some ... well maybe not exactly plot holes but definitely questionable elements in the first half. For example, the fact that everyone immediately carves up the phenomenon as "a person with these well-defined abilities" even though the abilities are supernatural. This is definitely not what would happen. Or the part where L puts all of the cameras into Light's room and then just decides that Kira wouldn't reveal himself even if he's watched. Like okay, he technically didn't, but man you're making it so much easier for him. the cameras were one hell of a restriction, and if you'd tried a bit harder, you could have probably gotten a lot more out of them. And Ryuk just naturally go along with a lot of what Light wants in the middle part of the show, which is a bit questionable since he was supposed to be neutral.

Anyway, on the whole I'd say 9/10. Great premise, amazing characters, great writing throughout, and utterly brilliant plot construction in the first 12-ish episodes.

[silverspawn]

The overarching irony of the entire show is that Light may have plausibly done good overall. It's not obvious since there are many higher order consequences of killing criminals in this way, but certainly not using the Death Note at all would be a catastrophic moral error. At the same time, Light is an obvious psychopath,  so it's not hard to root for his opponents.

[silverspawn]

Also I mean someone like L doesn't exist in the real world, but I don't count that as a plot hole since it's required to make the story work

[silverspawn]

Trump doesn't have 51% chance of winning the election if he's the nominee; that's ridiculous. Betting Markets still have this strange right wing bias.

[silverspawn]

Listening to a podcast with a legal expert on the Trump cases. This bit was pretty funny:

Interviewer: people say that others who behave like Trump wouldn't get away with it.
Expert: well actually we don't know that because all other criminally charged people are smart enough not to antagonize the judge or prosecutor so there's no precedent