Gyokuyoutama

Things have gotten worse.  Outlook now does markdown formatting.  If you don't know what I mean, it's that thing that's becoming more and more common where you can do stuff like **text** for bold, ~~text~~ for a strike through, etc. The problem is that _text_ is one shortcut for italics.  This is a problem because I often have to send e-mails with subscripts and superscripts on variables.  For subscripts I use _, like a_1, a_2, etc. This works well because it's intuitive and everyone is used to writing things that way in LaTeX anyway.  But now if I say "consider the variables a_1, a_2" outlook will change it to "consider the variables a1, a2."  As with all autocorrects, there doesn't seem to be any way to disable this in the current version.
 
Another annoyance is that it always capitalizes after a period.  So if you use phrases like "i.e.", "et al.", "etc.", "ex." it's going to mess up the formatting on the next letter.  This is annoying enough for normal sentences, but can also cause the types of issues with variable names that I discussed in the last post.
 
It's to the point that when I need to write an e-mail on my work computer I just write it in notepad and then copy paste it.

Things have gotten worse.  Outlook now does markdown formatting.  If you don't know what I mean, it's that thing that's becoming more and more common where you can do stuff like **text** for bold, ~~text~~ for a strike through, etc. The problem is that _text_ is one shortcut for italics.  This is a problem because I often have to send e-mails with subscripts and superscripts on variables.  For subscripts I use _, like a_1, a_2, etc. This works well because it's intuitive and everyone is used to writing things that way in LaTeX anyway.  But now if I say "consider the variables a_1, a_2" outlook will change it to "consider the variables a1, a2."  As with all autocorrects, there doesn't seem to be any way to disable this in the current version.
 
Another annoyance is that it always capitalizes after a period.  So if you use phrases like "i.e.", "et al.", "etc.", "ex." it's going to mess up the formatting on the next letter.  This is annoying enough for normal sentences, but can also cause the types of issues with variable names that I discussed in the last post.
 
It's to the point that when I need to write an e-mail on my work computer I just write it in notepad and then copy paste it.

I've been experimenting with having copilot translate images with Japanese text.  In some ways it's impressive in that it can recognize features in images and talk about them and sometimes pick up specific bits of text.  For example, sometimes it picks up no character names which it couldn't have gotten from elsewhere.
 
But of course, this is all buried in the usual strategy of "just make plausible shit up."  For example, I fed it the pick I modified for Wicket's Birthday:
 


 
It made the claim that the Japanese text is ハッピーバースデーお誕生日おめでとうございます!  If you look at the first part you will see that it is wrong.  It's the katakana for "Happy Birthday" phonetically (i.e. happi-ba-sude-) when in reality this is the phonetics for "Wicket."  Now why would the AI have done this?  Probably because it either recognized the cake and/or the English "Happy Birthday" and thus drew from its birthday related JP responses.  Honestly, I'm kind of skeptical about whether it recognized the お誕生日おめでとうございます, as that is a standard Japanese birthday congratulations and so would reasonably appear in a birthday picture with JP text.
 
Now what makes this weird is that if I specifically ask it what text is in the white box it will say "Wicket-san", meaning that it can parse that text, but it lied about it in the initial response anyway.
 
Here's another example.  I gave it this image of Nora Cat in the rain:
 


 
It claims the image has the text 雨だから便利でしょう? With the meaning "it's convenient because it's raining, right?" referring to an umbrella found that Nora apparently plans to use personally.
 
The actual text is 雨が降って来たので、傘を持ってきましたよ which means something like "because it started raining, I came with an umbrella" with the implication that this is being given to the viewer.  So it looks like this time the AI just made up some phrase that a character in the rain with an umbrella might use.  Again, if ask it for what text is specifically in the text balloons it gives me the right characters (though out of order since it's too dumb to read a manga panel right to left) so it has that capability, but it doesn't use it when asked to translate.
 
But then I feed it this complicated image from Doitsuken, which I thought for sure would stump it due to the highly unusual situation and the amount of text which isn't over a flat image:
 


 
And it pumps out this translation:
 
“私は出来る狐” (Watashi wa dekiru kitsune) - which translates to “I am a capable fox.”
“大丈夫ど私は出来る狐” (Daijoubu do watashi wa dekiru kitsune) - which roughly means “It’s okay, I am a capable fox.”
“運転中に人型解かないでね?” (Unten-chuu ni jingata hodokanai de ne?) - which translates to “Don’t transform into a human shape while driving, okay?”
“君·だんだん戦っぽくなってるけど” (Kimi, dandan ikusa-ppoku natteru kedo) - which means “You’re gradually becoming more like a warrior, though.”
 
That nailed it, there's only three mistakes:
 
1.) As before, it reads in left-right order and not right-left so this is all reversed.
2.) It confuses the kanji 獣 (beast) for 戦 (battle/war).  That is, he's saying that she's shifting back into an animal shape during the driving exam.
3.) It interprets 人型解かない as "don't transform into human shape" when it actually means "don't undo your human form", i.e. don't transform into a fox.
 
Let's try one more, using the intro page that I translated way back before I got too lazy to continue:
 


 
It gives this translation:
 


 
The names and races are accurate, at least.  But "Kitsune Musume", while technically something Kotoha is, appears nowhere in the text.  Shihiru is of course not Kotoha's older brother, being both female and unrelated to her, and no such text appears.  The Personality, likes, dislikes are all made up out of whole cloth.  I tried asking it specifically to translate the text under Kotoha's name, since in other cases asking it to translate specific lines would get something accurate even when it had previously lied.  It tells me  "The text under Kotoha’s name reads “性格：さっぱり” (Seikaku: Sappari), which translates to “Personality: Refreshing/Clear-cut"", which you can verify does not appear in the image.  What it did is take the made up translation and translate it back into Japanese. 
 
I asked it to translate a second time and got this:
 


 
Interestingly the character names are now wrong (though Kotoha's name does appear in Shihiro's description.)  Things aren't perfect; in particular it missed the fact that Kotoha specifically can't hide her ears and tail but otherwise can look human, but far better than the first try.  The problem is definitely not that the AI can't be used to trsnalte this sort of thing but that it won't.
 
I also asked it to translate some chickenscratch handwritten images, but it failed hard on all of those, not that I was expecting it to get anywhere.
 
But all in all that's pretty impressive.  Considering I'm using a free tool, it's certainly clear that current AI can be made to have powerful text recognition, image recognition and text translation capabilities.  However, it's also clear that the current methods for doing it are unreliable, because the AI is apt to make up something at random.  If someone were to optimize each part of the problem, such as text recognition, text translation, image recognition, sentence writing, etc., and then create a general framework to feed information from one part of the model to another, then I think we could get some pretty amazing stuff.  This seems to be more or less what Vedal is trying to do with Neuro-sama and the results so far have been pretty good for a one man team (and let's be honest: Vedal isn't a genius coder.)  But the CS academics are all in on the idea that if we stuff in enough training data then we will have competent "general intelligence" appear out of nowhere, even though every single time these models end up spending more time giving plausible sounding answers than doing what you want.

Was bored. Decided to mess around with AI. Asked to create an ASCII art, but it got it wrong every time.
With each attempt, it got more and more disturbing.
 
 
I believe a curse has been placed upon my crops.

Speaking of DS9 (several weeks ago) it always bugged me how
 
 

Thank you for tying this back to gaming because I'm too much of a dumbass to pay attention to which thread I'm in.
 
But that does mystify me everytime I hear that over a mic.  How can you live that way?  I gotta replace a smoke detector tomorrow due to the dreaded "end of life" chirp, and right now it's in "hush" mode which should theoretically last three days.  But I know from the last time I had to replace one that it never lasts as long as the manual claims, so who knows if it's going to decide to start being noisy again after I've been asleep for two hours?  Just that is enough to wear away at my peace of mind.
 
But living with a loud beep every 30-300 seconds (depending on the manufacturer)?  For days on end?  Forget how people have that happen and don't try to fix it, how do you live like that and not go insane?
 
EDIT: I mean, even if you were in full on "fuck it, I don't care if my house burns down" mode, wouldn't you eventually take the detector off the ceiling and throw it out?

Or you have a friend that refuses to acknowledge it yet keeps an open mic, gaslighting you into thinking it's yours until you take off your headsets and check them.
 
Luckily I can bring them up since they don't post here anymore.

Everyone needs an automobile, even those with many tails. This was the largest auto I could afford. Should I therefore be made the target of fun?

decided to replay the episodes on a whim. without a guide for get some grub I just tried to be thorough. I was off by fucking ONE AAAAAAAAAAAAAAAAA

Sparkling Generation Valkyrie Girl Yuuki was a webcomic about an otaku guy named Yuuki who bought a magical girls anime which was actually a ploy by the Norse god Hermod to recruit a new valkyrie, thus making Yuuki into a valkyrie themed magical girl.  Things are made more complicated by the fact even when he undoes the magical girl transformation, he remains a girl.

A couple years later there was a tokusatsu show in Japan called Holy Girl Warrior St. Valkyrie ( 聖少女戦士 St.ヴァルキリー) about a guy named Yuuki who inherits a magical item from his mother which turns him into a valkyrie themed magical girl.  Things are made more complicated by the fact even when he undoes the magical girl transformation, he remains a girl.
 
No, the webcomic author wasn't involved in the tokusatsu show.  I wouldn't be surprised if most of the staff of the show wasn't even aware the webcomic existed, and this was all done by one writer who snatched the idea without telling anyone.
 
From what I can tell the tokusatsu show seems to have taken major divergences from the webcomic overall.  For example, I can't see any Norse gods showing up despite Hermod and Thor being mainstays in the webcomic.
 
Both of these things are pretty damned obscure these days, so if anyone plans to make a "top 10 things YOU didn't know about Japanese shows" videos this would be a topic to hit.

Was reading the newspaper (), when my eyes stumbled on this particular picture.
 

 
I never watched the videos, but I gotta say I feel really bad for Gooseworx.

Didn't see Hellsinger in the thread. The soundtrack is really good.
 
 
 
 

Updated Red's design.
 
 

Same TBH. It's been bothering me for a while lol.
 
Especially because the show itself is basically forgotten even when way more overrated shows from the same era (*cough*Babylon 5*cough*) still get constantly praised and talked about.

I haven't really gotten into most of it yet, but I thought I'd tell you about another Solitaire game because it seems pretty obscure.  It's A Solitaire Mystery and it's by the Baba is You guy.  As far as I'm aware it's only available on this itch.io page:
 
https://hempuli.itch.io/a-solitaire-mystery
 
Not on Steam or Gog or anything.  The only reason that I know about it is that Jelly from Phase Connect did a stream where she played through the Zachtronics solitaire collection, and then she immediately played through this one too.
 
There's some pretty obvious inspirations from the Zachtronics games.  In particular Tap Solitaire is very, very close to Cluj solitaire: sort cards into stacks but you can cheat by putting a card where it doesn't belong.  In this one suits matter and they are given Baba is You flair, but it's pretty similar.
 
However, there a staggering 23 solitaire games in this collection and many of them are unlike anything that Zachtronics has put out.  There is a bona fide Baba is You solitaire and it's everything you would hope.  The suits are Baba, Keke, etc. There are also four noun cards, four "IS" cards and four attribute cards.  So you can arrange BABA IS +5 which causes the value of all cards in the Baba suit to increase by 5 (looping around on the top.)  Or KEKE IS ANY SUIT which allows you to use the cards in the Keke suit as if they were any suit (except for putting them up in the goal stacks.)  It's easy to overlook, but you can also do things like FOFO IS BABA to make all Fofo cards count as Baba cards.  Captures the spirit of the game nicely while also working as a solitaire game.
 
Some games are pretty close to normal solitaire versions.  For example, Hanoi Solitaire basically has fewer slots but more generous stacking rules that make the game function like the Tower of Hanoi puzzle.  There is a variant of the common Poker Solitaire game that people play in real life (basically you need to move cards to make poker hands to get a certain number of points.)  Others are more fanciful, like Eldritch invasion where all face cards are Lovecraftian monsters that need to be defeated by totaling the correct sum, while being careful not to reveal too many monsters to get killed (yes, there is a "you died" screen.)  Time Travel solitaire lets you take cards from the future, but you need to then return them back when a timer runs out lest you cause a temporal pardox.  52 solitaire basically throws all the cards in a random fashion in a box, and you are only allowed to pick up cards that aren't covered by other cards (i.e. like you start with the aftermath of 52 pickup.)
 
Many fit into the "fucking evil" category.  For example, in chaotic solitaire your card will get swapped after you move it. Lock solitaire prevents you from moving a card for two moves after it is placed, making it very easy to walk into a corner. Binary solitaire just gives you 1's and 0's, but these need to be combined into the numbers 0-7 (i.e. 0, 1, 01, 10, 001, etc.)  Wrapping your brain around how to combine cards and then swap them is a brain buster.  Single card solitaire is even worse: you just have a bunch of blank cards which must be combined into groups to sort properly.  I am sure there will be card games that you absolutely hate in this collection; but with 23 you'll probably find at least one that you like.
 
One thing I like is that each game has some comments on how easy it is to play with a real deck of cards.  This is something that I did with the games in the Zachtronics collection (they are all playable with a deck of playing cards or a deck of Tarot cards except for the alcehmy game, though you have to be creative when playing Shenzan or Kabufuda.)  I particularly enjoy playing Sawayama and Cribbage solitaire with a real deck of card (Cribbage solitaire is actually good practice for playing Cribbage for real.)  So I appreciate having some advice on these new games.  Eldritch invasion seems like it'd be the most fun, especially with a deck of cards where the face cards look appropriately menacing.
 
EDIT: Bonus Jelly meme:
 
 

Lisa reference pose
 
 

I think the trouble is that there are a lot of blonde characters, often of lesser importance so you forget who they are.  Once you see Marisa's iconic side braid it's clear, but it's easy to miss on first watch.
 
But this made me realize that I haven't played the classic early Windows 2hus in half a decade.  Started up EOSD, couldn't beat first stage on normal.
 
Though that probably had to do something with the fact that it was running at 1600FPS.

All you need is one person to stumble on it and post a link on reddit and then it's lazy youtube content city.
 
Said person could even be someone who knows about SPUF but wants to see what redditors would do with it.
 
He could even be you.  He could even be me!  Well, no not really because I'm not going to make a reddit account for any reason.
 
Anyway, while I'm not about to conduct the experiment in reality, I do wonder what such an investigation would result in.  If they had anyone with the minimal amount of competence, they would realize that you click on profile names.  Then you can look at public posts, which includes everything posted on public forums before they were locked in 2019 and every single status update even since then.  This is just the recent stuff at first, but you can get the full backlog with "see their activity."  Thus if any investigators were minimally competent they'd be able to quickly sus out that this was a bunch of morons posting about anime and vidya and ponies and whatnot.
 
But there is always the possibility that someone creates an off the wall conspiracy theory that reddit hiveminds onto.  In that case it doesn't matter how much actual good data people get afterwards, because the original theories are all anyone would read.  In that case likely someone would eventually realize that the posts end in 2019 and suspect that something sinister is going on behind the login wall.  Probably someone would claim that the status updates are being generated by LLM AI or something.
 
EDIT: Actually I can contribute to the mysteriousness.

Give it a year or two and reddit or nexpo or someone is going to make a video on "5 more unexplained rabbit holes on the internet" and we'll be #4.
 

I wonder how many people have stumbled across this website over the years and wondered what the hell it was, considering the front page is basically blank when you don't have an account.

I watched too many of those HOMM memes and now I get slav tube recommendations