@wraelk's banner p

wraelk


				

				

				
0 followers   follows 0 users  
joined 2022 September 05 20:35:54 UTC

				

User ID: 703

wraelk


				
				
				

				
0 followers   follows 0 users   joined 2022 September 05 20:35:54 UTC

					

No bio...


					

User ID: 703

This actually fits a lot with my experiences: I can get good results, but sometimes it just absolutely fails in a way the makes me think I'm getting different underlying code.

This is good information, but I was really hoping I would not have to do either of these things. At some point I was able to get reliably quality artwork with transparent backgrounds, so I know the technology is there. It's possible it's not something I can access.

What's everyone's favorite AI image generation model?

I'm running a Pathfinder campaign, and being able to generate on-demand NPC art that looks like the vague idea I had in my head for a character is a huge step forward from having to scour the internet for stock art and make sure I haven't reused anything.

Unfortunately, I've been working with Microsoft's Copilot and ChatGPT (with a brief disappointing dip into Gemini), and recently I've noticed that both of them seem to have regressed in their image generation models: wonky eyes that don't get better, one side of the face significantly differently styled than the other, and my personal most annoying issue, being unable to generate with transparency in the background. ChatGPT can do this, but Copilot and Gemini regularly fail over and over at this task, often generating a white-and-grey square pattern that browsers will use to indicate transparency, but that's the actual background of the image, and you can see points in the background where it messed up and did the same color multiple times instead of alternation. These all feel like issues from the early days of image generation, so I suspect they've regressed their free models to something that's cheaper for them to run.

I'm not opposed to paying for it (though free is clearly better), but I'd really rather pay someone other than OpenAI, I just don't like Altman or how the company is run. And if I'm going to pay someone I'd want a good text model as well, I often use the text models to generate names or flesh out ideas.

Less than a year.

Met my wife on OkCupid, we chatted for a week or two before setting up a first date. We were about 2 hours away from each other, and I set up a first date closer to her. While we were waiting for the date to arrive she asked if she could come up and see me before that. It's quite nice to have someone else clearly excited about the relationship and willing to go out of her way for it.

I told her I loved her on the first date, and while that was obviously premature it turns out to have been correct. The experience was just so overwhelmingly better than any relationship I had been in before, and I didn't think the past relationships were bad, but this was just a whole other level. She would not say it back to me, though she expressed interest

We talked about marriage and children on the first date, in the sense of "is this where your plans eventually end up?" but I don't know when we actually started talking about "ok, let's get married for real". I know at some point her position was "it's up to you, I'm ready to say yes any time you're ready to propose", probably about a year or so in.

I ended up effectively moving into her place within a month or two of seeing her, in that my visits would just span multiple weeks. I was looking to buy a house at the time and when I did so, <6mo after the relationship started, I invited her to move in with me. From there, my strategy was essentially "let's wait and see to make sure nothing bad pops up": over time living together you learn each others' quirks and any problems you might have, you see the other person at their worst. So I gave it about 2 years, nothing concerning happened despite some decent low points (surgery recovery, difficult work conditions), and went ahead and proposed.

I think the key point is that AI art is hitting the business of relatively low-skill commission artists.

I'm active in TTRPG subreddits and periodically see people post commissions of their characters: sometimes they have some neat stuff going on but the artistic merit is almost always lacking (the first result I found was this), and slapping a vague description into an image generator is going to produce something way more impressive pretty much every time. Yeah it might not hit the exact note you were looking for, but that's true of commissions too, and you can't tell the artist to go back and try again for free.

Porn is another example where the bottom's getting cut off: AI struggles with details and you have to use weaker models to generate any sort of adult stuff, but the bottom end is really bad, and you get to customize it as much as you want and churn out as many attempts as you want for free on a local model.

These people are often trying to do this as their job, which means their workday involves sitting around and posting on the Internet a lot, and they're often doing that for hobbies too. They also saw AI coming before most people did, because the early adopters playing around with Stable Diffusion would have been visible in their corners of the Internet. So they're initially the only ones who care about it, and being extremely online (and as artists, overwhelmingly left-wing) they know how to leverage progressive terminology to make their points. By the time it propagates out to the normies, the claim's been staked.

Secondly, I can't even EAT one of you if I wanted to; at best I'd expel it out the other end completely undigested. Wrong protein chirality.

What I'm hearing is "miracle health food": all the taste, none of the calories!

Yeah, furries almost point in the other direction with this: it's human beings being "attracted to other species" but in reality they're just taking attractive human characteristics and putting them on those species.

Though I guess this does open up an opportunity for arbitrage: imagine an alien species that has some extremely unattractive-to-them feature that maps well to attractive human characteristics, and vice-versa: now the unattractive members of our species and the unattractive members of their species can get together, both finding the other extremely attractive.

How is this not a romance series already?

or that it will change your emotions by reflexively pumping drugs into your bloodstream.

I'll note that even your link says "can opt to", and Wikipedia says

Most Culture individuals opt to have drug glands that allow for hormonal levels and other chemical secretions to be consciously monitored, released and controlled. These allow owners to secrete on command any of a wide selection of synthetic drugs, from the merely relaxing to the mind-altering: "Snap" is described in Use of Weapons and The Player of Games as "The Culture's favourite breakfast drug". "Sharp Blue" is described as a utility drug, as opposed to a sensory enhancer or a sexual stimulant, that helps in problem solving. "Quicken", mentioned in Excession, speeds up the user's neural processes so that time seems to slow down, allowing them to think and have mental conversation (for example with artificial intelligences) in far less time than it appears to take to the outside observer. "Sperk", as described in Matter, is a mood- and energy-enhancing drug, while other such self-produced drugs include "Calm", "Gain", "Charge", "Recall", "Diffuse", "Somnabsolute", "Softnow", "Focal", "Edge", "Drill", "Gung", "Winnow" and "Crystal Fugue State". The glanded substances have no permanent side-effects and are non-habit-forming.

which mentions optionality again and also mentions "secrete on command".

It's been a while since I've read them, but it seems like that if they're automatically dispensing drugs to you they're doing so under parameters you configured yourself, which seems perfectly fine.

It's not a top level post, it was a response to eetan.

Yup, I got this wrong. I've edited my original post to make note of this.

Discussion engagement with a topic is a limited resource here. So if you spend the time to write up an in depth post and it take you half a day, you don't want someone sniping the topic by basically posting a dumb article or twitter thread.

That's a fair point: you want the opening post on a given topic to be the highest-quality post, so low-quality posts that take less time to write can undercut people with less effort.

The question is, is a higher-effort post actually going to be written on a topic? From the above, I've seen the dinergoth thing before on the Culture War thread, but that was a previous week, so it seems there was more discussion to be had on it. I'm not sure if the other topics would have otherwise gotten any mention. And if there ends up being no post on a topic otherwise, then the relatively-bare post was the highest-effort post and it's being blocked by the forum rules.

I don't have a great way to solve all of these issues: the best thought I have is to have a bare-links-allowed post later in the week, so that anyone wanting to write detailed posts has a few days head start at the start of the week. That doesn't help with events that actually occur midweek though, and "no bare-links posts about things that happened less than 2 days ago" seems like it would be too annoying to implement for the mods. IIRC we used to have something similar around major tragedies, although I may be confusing that with another forum, and major tragedies are thankfully rare (or we calibrate "major" by how common they are, at least)

Other spinoffs of the culture war thread concept have been tried. The ones that allowed bare links are all dead.

Do you mean on TheMotte, or are you talking about places like TheSchism? I think that's maybe a "correlation implies causation" instance: the problem with TheSchism was not its bare link thread.

Depending on what your criteria for "other spinoffs of the culture war thread concept" is, rDrama seems to be alive and quite active despite allowing basically any level of effort or non-effort. The discussion criteria is clearly not up to the standards of theMotte, but I think that's more downstream of the userbase and general comment moderation.

Bring back relatively-bare link posting! In its own thread if need be.

The top-level post below this one (JeSuisCharlie's) is literally just a link with a quote from that link: the only user-provided content is the label for the link. EDIT: that was a second-level post, not a top-level one, and that does detract from my point somewhat.

You don't mod it, despite that having less personal commentary than any one of Eetan's links.

The resulting discussion from that post is interesting and enjoyable. The Motte would be worse off without it. That is true for this top-level post as well.

People posting uninteresting links is a self-solving problem when no one responds to them. We're a discussion forum, and having a jumping-off point for a discussion is useful.

Although the Chavez case could also be seen as an example of being willing to hold their own leaders to account: I generally have more respect for organizations that clearly abide by their stated principles.

Well, he has been dead for 30 years, so it might be a little late to provide feedback and steer his behavior.

And it seems like this was not an unknown thing before now, so I don't know how much credit I want to give: "abiding by your stated principles" is mostly impressive when it's chosen over maximizing your capability to attain your goals. Is denouncing Cesar Chavez now costing them much?

But, to be perfectly mirrored, we'd have to look at people that died in the 90s, were credibly accused of sexual misconduct, and were right-wing, to see if right-wingers are currently denouncing or at least not supporting them. Maybe they don't clear this bar.

I think the argument is that #MeToo accusations are strategically delayed to minimize harm to the left and maximize harm to the left's enemies.

If someone's in a position of power and supporting the left, the #MeToo accusations will be delayed until they're no longer in power or it can be guaranteed that they will be replaced by someone just as supportive to the left.

If someone's not supporting the left, the #MeToo accusations will come out immediately and be leveraged to their maximum extent to attempt to replace that person with someone more supportive to the left.

I'm not sure how much I agree with it, but this should be at least somewhat disprovable: there's gotta be some easy counterexamples.

Al Franken was a sitting senator when he resigned, but the accusations were from 2006 and didn't come out until 2017. And he did get replaced with another Democrat who won 76% of the vote, so it's plausible that it was timed to where they were sure the left wouldn't lose any power by it. But the accusations were from Leeann Tweeden, who has a few right-wing things in her bio: I don't know enough about her to know if that's accurate. "Left-wingers strategically get right-winger to accuse left-winger of sexual misconduct at a time where they're confident another left-winger can win the election" seems a little too complex for me to see as plausible.

Isn't Dase Russian? Here he refers to the Russians as "us" and the Americans as "you".

VNRs as popularly conceived might not be borderline magical nanotech, they might just be a few megatons of old-fashioned industry adapted for space that take a decade to duplicate.

I do think this has a few notable differences from the normal depiction of VNRs. The city-sized factory that takes a decade to duplicate itself seems like it's far more likely to have something go wrong, which is a big problem if you're only planning to launch one, and the longer it takes to work the more likely something breaks it before it finishes.

"A factory that works perfectly for 10 years without external human intervention" seems very far from our current technology, despite pretty good incentives to make one.

Though now I really want a Factorio mod that allows you to build space platforms from other space platforms.

Do women ask it of men?

N=1, but my female partner is generally appreciative of me keeping the area well-groomed and trimmed down (not waxing) for the sex-adjusted equivalent of the reasons TitaniumButterfly mentions.

I have no reason to believe she's attempting to make me look younger or more feminine as she's quite happy with hair in other places, but hair in the mouth is never pleasant, and trimming things down in that region often visually accentuates what's there quite well.

She appreciates my mustache being kept trimmed down for similar reasons.

This does point out a slight problem with the mnemonic though, which is that every month mentioned has multiple other months that read the same way.

Thirty days have December, August, May, and September. All the rest have thirty-one, except for January, which has 26.

There's no way to correct from something wildly wrong like this to the correct rhyme (other than looking them up to check, which defeats the point of the exercise), because this rhymes and scans just as well. At the end of the day you're just memorizing the right months and numbers.

While mistrials can be declared for a bunch of reasons, your article is very clear on why there's a mistrial:

Mr Harris and Mr Caserta were found not guilty, but the charges against Mr Fox and Mr Croft ended in mistrial.

The government had argued they targeted Gretchen Whitmer, 50, in a 2020 plot.

Jurors began deliberating this week after 14 days of testimony and had indicated earlier on Friday that they were deadlocked on some of the charges.

They ultimately reached no verdict against Mr Fox, who was alleged to be the group's ringleader, and Mr Croft, both of whom were also facing an additional count each of conspiracy to use a weapon of mass destruction.

They didn't find Fox and Croft not guilty, they were unable to decide whether they were guilty or not, which means a hung jury, that mistrial status, and generally a retrial: this was a pretty high-profile case, it would be very surprising for the courts to say "eh, we don't really care about this enough for two trials, let it go". Their co-defendants Harris and Caserta, who the jury did acquit, did not have a second trial, and if Fox and Croft had managed that they wouldn't be getting a second trial either.

You can argue that the courts stacked the deck to make sure the second trial had a better chance of finding them guilty, but that was clearly a reasonable possibility even in the first trial, or the jury would have just acquitted as they did for Harris and Caserta.

Still no judge overturning the convictions: Fox and Croft were still in the indistinct "haven't resolved this in a court" status after the first trial, a second trial found them guilty (still by a jury), and then the appeals court decided the jury's decision was from a trial that was performed adequately and didn't need reexamination.

Even just recently, with the governor kidnapping attempt, who's conviction of those indicted was actually overturned because the majority of said group was federal operatives egging the entire plot on. And that overturned conviction was reversed by a higher-judge.

Can you provide some more information on this sequence? Particularly the bit where the conviction was overturned, as I'm having trouble finding anything about it.

What I've seen on this is that the initial trial had some convictions and some acquittals, with entrapment being presented as part of the defense, and then the court affirmed on appeal for Adam Fox and Barry Croft, who argued that the court didn't let them focus on the entrapment aspect as much as they wanted to.

Which, maybe they're right, but if this is just the appeals court going "we're sticking with what the jury said", that seems significantly less bad than judges fighting each other over the conviction, which I can't find evidence of.

At scouts as a kid, the Dreidl was crackerjack, so there's been some effort. And in middle school chorus, the token chanukah songs were normally pretty good. So it's not impossible.

Yeah I vaguely recall someone coming to our schools a couple times, maybe making some latkes, and learning some songs. Systematize that, bring a basket of challah and some chocolate coins for the dreidel gambling game every year, and it'd be a fantastic initial impression.

But I think it would drop off pretty heavily in adulthood, because while I remember Diwali celebrations in the office quite fondly, I don't remember any Hanukkah equivalent, and I know I have Jewish coworkers. Which speaks to your initial point: Judaism doesn't seem to have much interest in us getting involved in their holidays. So thus I claim the "holidays" for Christmas and Christmas alone, as Diwali lands much earlier, the Asian New Year is later, I have never seen anyone genuinely celebrate Kwanzaa, and New Year's is basically a Christmas extension.

Happy Holidays My most boomer take, I hate the phrase Happy Holidays.

You've linked Andy Williams' Happy Holidays/The Holiday Season medley, which is the perfect subversion of the original and the general "happy holidays" sentiment: its chorus is Happy Holidays, but 90% of the lyrics are about Christmas and no other holidays are mentioned at all.

Which is in my opinion the best way to respond to the Happy Holidays term: I mean Merry Christmas, you mean Merry Christmas, we all know that Christmas is the only holiday that really registers on anyone's radar. Happy Holidays means Merry Christmas, it is an entirely Christmas-owned term. If the Jews want us to get Hanukkah in there, they better get started on doing some outreach and getting people on board, because right now my second favorite December holiday is Diwali. It's got nothing to do with Christmas, but the snacks are great and the celebrants are generous with them.

I agree completely, I get incredibly frustrated every time I see this.

Most of the time I see "var" in C#, it's because the dev didn't want to have to track down the exact details of the type they're making a variable of because they understand the general shape of it but the wrappers and type details are annoying.

But the details of that type are important, and if they can't trivially figure it out when they're first writing the code, the other dev who comes in 2 years later to fix a bug with the code is going to have a way harder time.

In theory it's only supposed to be used for trivially-inferrable types like an int, but I very rarely see that because... if it's obvious it's an int, it takes exactly as much time to type "int" as to type "var".

They did phrase it as "why are you going after Iraq, Iran is who you should be going after" though, which paints a slightly different picture.

https://www.commondreams.org/news/2007/08/29/israel-warned-us-not-invade-iraq-after-911

There's a big difference between "the Empire remains a relevant force" and "it's just the Emperor again" though. A fledgling New Republic dealing with an Imperial remnant force with its own local goals that the Republic is spread too thin to deal with is very different than "if the Imperials ever catch our fleet, we're all dead".

And while the remnants often have something unique that make them a threat to the New Republic, it's rarely a "superweapon" in the sense that the Death Star was one.

  • Thrawn himself was his "superweapon" and rather than being a weapon of mass destruction, he had highly tailored and specific tools that let him punch above his weight with the ships he had and his grab bag of tricks.
  • Ysanne Isard had a Super Star Destroyer, but that's just a conventional ship: her real threat was her planning and information-gathering, which eventually led to her master plan of taking over bacta production. Trying to choke out a key military supply is wildly different than anything the Emperor has ever tried.
  • Warlord Zsinj had a superweapon, the Nightcloak, but it was incredibly limited, an array of satellites that would block out light from reaching the planet. Functionally pretty much indistinguishable from orbital bombardment. Most of Zsinj's threat is his conventional fleet.

Not to say there were no Death Stars: the Sun Crusher is the one I want to call out as being the most boring "well what if we made a Death Star but better". But notably, the Sun Crusher spends very little time in the hands of the Imperial remnant, first serving as an escape tool and then falling into the hands of a troubled young Jedi. So there we have the Death Star, but not the Empire.

My favorite riff on the Death Star was Darksaber, where the main plot was a Hutt finding Bevel Lemelisk, the original architect of the Death Star, and trying to get the minimum-viable-product version of the Death Star, where it's just the laser and nothing else. This is a threat the New Republic has to take seriously and deal with, though it turns out the whole project was a train wreck and it's destroyed the first time it tries to fire the laser to clear an asteroid in its path.

The lack of any of this adjustment is one of the problems with the movie sequels: the First Order is played as exactly the same threat and type of threat as the Empire, with Death Star 3 and similarly overwhelming fleet, and the New Republic is immediately relegated to a background entity so that the good guys can be exactly the same as the Rebels.

I'm not going to personally defend this perspective. I'm just here to point out that the data-based argument that samiam makes doesn't actually refute the "alpha fux, beta bux" argument as incels normally formulate it.

There are certainly philosophical arguments that can be made (really, that's more or less what's going on upthread), and I'm not even ruling out the idea that there's a data-based argument that does refute "alpha fux, beta bux", it's just not this one.

I am not an incel and disagree with many aspects of their arguments: the best I can muster is a "there but for the grace of God go I".