From an article about a boutique brand that sells books to rich people:
Assouline has made its name publishing tomes that sell for $1,000 or more.
Oh, so they publish textbooks.
"They represent stealth wealth, intended to tell you what your hosts are about and to provide visual evidence: that the owners are people of wealth, education and taste."
🎶 Please allow me to introduce myself 🎶
now where did I leave my hammer and sickle
"...37%.... That means nearly one in four..."
Eh, no it doesn't, it means nearly two in five. Which is worse.
So you're saying that they've got books worth at least a grand which their owners are literally using to flaunt their wealth?
I'm legally obligated to say stealing is legally and morally wrong buuuuuuuut
@BlueMonday1984 Betcha the authors aren't getting paid industry-normal royalties (10-15% of net receipts) on those Veblen goods …
(A few of my novels have been sold as limited-run signed first editions. Typically for 50%-100% more than the normal hardcover price, so maybe 3-5% as much as this nonsense. Cost of goods for a leatherbound, gilt-trimmed luxury edition is maybe $5-10, plus 10% of the cover price for the author. So someone in the middle is making serious bank.)
Of course the books are just vapid art books or travel books.
just heard a podcast ad for amazon prime saying it causes "involuntary deal squeals" followed by a categorization of different kinds of customer grunts and squeals according to product. not making this up
one of OpenAI’s cofounders wrote some thoroughly unhinged shit about the company’s recent departures
Thank you, guys, for being my team and my co-workers. With each of you, I have collected cool memories — with Barret, when we had a fierce conflict about compute for what later became o1; with Bob, when he reprimanded me for doing a jacuzzi with a coworker; and with Mira, who witnessed my engagement.
I am in awe of the sheer number of GPUs... whose lives ChatGPT has changed.
If it was just this one line, this would be in the top 10 funniest things ever written around genAI. Too bad the rest of the rambling insanity ruins it.
I can't be the only one reading that super passive aggressively right? "Thank you Barret, whom I hated. Bob, for ruining my hot Jacuzzi date. And Mira, for existing."
I am sure this is totally not sketchy in the slightest and the people behind it have no nefarious agenda whatsoever.
Found a good one in the wild
Didn't you know LLM stood for Limited Liability Machine
Not a sneer, but some truly beautiful karma:
Judge to approve auctions liquidating Alex Jones’ Infowars to help pay Sandy Hook families
A lobsters states the following in regard to LLMs being used in medical diagnoses:
If you have very unusual symptoms, for example, there’s a higher chance that the LLM will determine that they are outside of the probability space allowed and replace them with something more common.
Another one opines:
Don’t humans and in particular doctors do precisely that? This may be anecdotal, but I know countless stories of people being misdiagnosed because doctors just assumed the cause to be the most common thing they diagnose. It is not obvious to me that LLMs exhibit this particular misjudgement more than humans. In fact, it is likely that LLMs know rare diseases and symptoms much better than human doctors. LLMs also have way more time to listen and think.
nothing hits worse than an able-bodied techbro imagining what medical care must be like for someone who needs it. here, let me save you from the possibility of misdiagnosis by building and mandating the use of the misdiagnosis machine
Also please fill in the obligatory rant about how LLMs don't actually know any diseases or symptoms. Like, if your training data was collected before 2020 you wouldn't have a single COVID case, but if you started collecting in 2020 you'd have a system that spat out COVID to a disproportionately large fraction of respiratory symptoms (and probably several tummy aches and broken arms too, just for good measure).
Turns out trump really does understand cryptocurrency perfectly. Who’d have thought? Some folk seems surprisingly unhappy about this, though.
Maybe we’ll pay off the $35 trillion US debt in Crypto. I’ll write on a little piece of paper ‘$35T crypto we have no debt.’ That’s what I like.
a twitter thread by sv ceo where comment section wants to do some recreational union busting, political assassinations and automating away longshoremen (lmao) over checks notes black friday bringing slightly less profit to mass retailers. to which i say, fuck your black friday then
and he says that it'll affect elections? specifically in "don't do anything visible in interest of unions or trump will win" kinda way? what kinda madhouse is this americans explain https://xcancel.com/typesfast/status/1836498432510562788#m
fuck your black friday then
the institutionalisation of it is/was also just fucking nuts
and then in recent years it's slowly been creeping out into other countries too, with other vendors in other places aping "black friday deals"
I have no mouth and I must scream
Somehow I managed to mention the wordpress lawsuit on last week’s thread instead of this one, so let’s try again.
Matt Mullenweg, the wordpress(.)com guy and current owner of tumblr, tried to shakedown competing blog product WP engine (which builds on the same open source software that his company does) for 8% of their revenue (https://goblin.band/notes/9yjrc2logimd1tr3 h/t to froztbyte who was also on the old thread for some mysterious reason) or he’d say mean things about them at a conference where they were one of the sponsors. And they didn’t pay up, so he compared them to cancer.
And now they’re suing him.
https://notes.ghed.in/posts/2024/matt-mullenweg-wp-engine-debacle/
Mullenweg's the same guy who publicly harassed a random transwoman on Tumblr and had a general meltdown to the point where Tumblr staff had to distance himself from him, so I'm not shocked.
(That its Tumblr is the only thing that shocks me - you'd think he'd have realised its queer-friendly rep was one of the main things going for it)
@BlueMonday1984 @blakestacey I can't believe I'm siding with Michael Bay over James Cameron on a matter of artistic integrity. Make it stop!
I have to go run an errand soon but someone better have posted some commentary about the a16z anime blog post (as seen on the hell site) by the time I get back or I'll be sorely disappointed.
Ok.
The usual lifecycle of an anime fan looks something like this: they are introduced to the format with great IP – the Attack on Titan anime or the One Piece live action show or one of the miHoYo games.
I don't know how things are in Japan, but I'll be damned if I ever meet someone who gateway series into anime was a live action adaptation of One Piece.
AI companions, an evolution of classic visual novels, are the most popular for anime characters and IP.
The most popular what for anime characters and IP?
Anime studios are adopting new AI technologies to create content faster and more cost effectively, but they are also iterating on new core loops with AI-native character interactions.
Some of them probably are. Screw them.
VTubing has transformed the way millions of anime fans interact with their favorite characters in new social and parasocial relationships by allowing any fan to roleplay as the characters themselves.
You can't just casually throw "social and parasocial" in there and then describe a purely parasocial relationship. Apologize to Shannon Strucci.
Also this is like saying television has allowed us to roleplay our favorite Radio announcers. They seem to be under the impression that the vtuber phenomenon is about people digitally cosplaying their favorite anime character together when it's more like an actor putting on a performance as an original character. And for the big ones, a bunch of Japanese style idol industry bullshit layered on top.
While audience inteeaction is usually a part of it, the nature of the medium remains highly asymmetric.
Ready to dive in? Let’s jam.
Keep Cowboy Bebop's name out of your filthy mouth.
Anime entered the mainstream in the 2000s with popular shounen anime like Naruto, One Piece, and now Attack on Titan.
I might be behind the times but even I don't think AoT is new. At least say Jujutsu Kaisen or something.
This affinity has led to one of the most popular use cases of AI recently – AI waifus and husbandos.
May all your subculture in-jokes die a dignified death before a VC firm references them in a blog post.
Waifu / husbando culture derives from visual novels, and AI companions are the logical extension of these animated storybook games.
"Mai waifu" was originally a funny engrish quote from Azumanga Daioh and was used to refer to any favorite character. The non tongue en cheek relationship simulation aspect merged with the meme later on.
Originally, visual novels were serialized books with anime-styled pictures in between.
This doesn't seem to be what the linked Medium article is saying and seems like they're just mixing up light novels and visual novels.
While there are many practical use cases for AI-simulated human interactions – AI as therapist, as teacher, as assistant, etc.
Practical, huh?
For instance, character.ai’s top characters are all from Genshin Impact; Raiden, Yae Miko, and Hu Tao take some of the top spots at 390M, 202M, and 113M messages respectively as of the time of this blog, compared to Elon Musk at a mere 40M messages.
To be fair I'd rather take almost anyone, gacha game character or not, other than Elon Musk as my conversation partner, whether simulated or real.
The majority of top anime games and visual novels are role playing games that feature a romance mechanic, and so it’s natural for fans to want to deepen their connection to their favorite IP and characters through active interactions.
Factually dubious claim aside, how hard is it to write "series" or at least "anime" like a real human being with feelings instead of "IP".
I've watched some anime series and felt things about them. I've never given a shit about an anime IP. Why would I, never owned one.
UGC Democratizes Creation for Anime Fans Anime is the new playground for content creation. Fans often engage with anime IP by creating their own versions of art, novels, and games, and innovation is happening across the stack.
Pixiv has existed for ages. Even before that was doujinshi, and people have made art, original and derivative, since before the beginning of civilization. Your idea of modding custom animu avatars for shovelware Love Plus sequels is not new.
There are a few notable reasons for the popularity of these games. The first is that there’s clear player demand against a shortage of high quality anime IP games; one example is Palworld’s recent success as the “Pokemon with guns” game, selling over 25M copies in a month across Steam and Xbox Game Pass.
Palworld is evidence of a lack of high quality anime games much like all nonblack nonravens are evidence of a lack of nonblack ravens.
The second reason is that the anime IP licensing landscape is notoriously difficult to navigate for developers, creating a potential undersupply of games.
It's actually incredibly easy to create and publish media based on anime and get away with it. You just can't do it too professionally. If you love democratizing art so much, go to Comiket.
Also there are tons of licensed games based on anime what the hell are you talking about?
Some startups like Kasagi Labo, Layer, and Story Protocol are tackling this issue to make IP more democratized and easier to access.
Misspelled "plutocratized" there. Also had a double take checking out the third one: "Story is the World’s IP Blockchain, onramping Programmable IP to power the next generation of AI, DeFi, and consumer applications."
Beyond UGC platforms, AI models and tools are enabling first-time creators to make compelling anime content that previously would only have been possible with a team of professionals.
I'm sure I will continue to be as thrilled as I have been up to now to see more art made by people who can't make art and filling the gap with statistical average of all art ever.
On the other side of the spectrum, professional game studios are leading the charge for high production-value consumer experiences that build on or create new IP. Anime games are some of the highest grossing in the games industry, accounting for 20% of spend on the mobile app store despite only having usage penetration of <3%.
Sounds great (not), but I heard someone say there was a lack of high quality anime IP games. Surely you can't both be right?
There are two ways that anime game studios broaden the horizon for players. First, they usually create the highest quality games of the most popular IPs like Dragon Ball, Pokemon, or Dragon Quest.
Consistency, what's that? Maybe invest in a bigger context window so you can remember what you generated a few paragraphs ago.
For now, we’ve been covering mostly free-to-play (F2P) mobile games. However, there are several successful PC/console anime games as well: Doki Doki Literature Club, the Persona series, the Final Fantasy series, the Fire Emblem series, and Phoenix Wright, just to name a few.
Doki Doki Literature Club is a fully original freeware pay-what-you-want indie game that became a viral sleeper hit. You're comparing it to Final fucking Fantasy? From a business perspective? Hell, despite the art style it's not even Japanese! The only connecting thread between these games is that they have vaguely anime style art in them.
Anime is also leading the way for digital play, turning previously passive consumption of linear media into a new dynamic form of entertainment.
It's really not.
Some notes:
-
Who told Mark Andreesen about the overlap between possible AI ~~suckers~~ customers and weebs? Are we going to get a16z's next hot take - "Furries are eating the world?"
-
I'm sure most of the audience here can fill in their own 700+ word rant about the breadth of anime as a visual style, so I'll leave that as an exercise to the reader. However, unlike the older trends of assuming that whichever shonen is currently most popular (the kids still like at least one Dragonball, right?) is representative or dismissing anything with the relevant aesthetic as "some weeb shit I won't like", here the writers manage a much more impressive feat. They acknowledge the breadth of what anime contains, but completely fail to ask the basic question: "why do people like this?" Similar to the original prompts for this kind of rant, they're assuming the art style and Japanese cultural background are the primary reasons why anyone connects with anything anime, and then expand from that premise. I'm pretty sure this is a root cause of why the whole article feels like it was written by goddamn martians.
-
Are vTubers playing existing characters a thing? What little I've seen isn't linked to existing stories (that's what humans call "IP") but rather focus on original characters who have their own shit going on. Even ignoring the attempt to shove genAI into everything (as though everyone is going to want to make their own vTuber avatar and stream it someday?) this seems like assuming that the people going to watch the finals of the local Battle of the Bands are going in the hope of getting an autograph from Kurt fucking Cobain.
-
There has been some criticism of gacha games as being monstrously exploitative and basically gambling targeted at kids and/or teens, but consider just how much money it makes. These people are ghouls.
-
Going back to the genAI we set aside two bullet points up, I do think anime has a unique property there. It simultaneously has a much stronger visual identity than many other aesthetics, including photorealism, but also has a massive number of scrapable examples to train off of. The more consistent style makes it easier to replicate statistically and what visual abberation you still get is less likely to fall deep into the uncanny valley. The outputs I've seen from even older anime genAI were better than their contemporaries, but still pretty easy to pick out. Something about shading or gradient or something, probably because since anime is drawn rather than captured like a photo there's no detail that's fully incidental. GenAI, of course, has no actual purpose and so all details in every output are incidental. That gives the output a weird unfocused quality I think?
In conclusion, I'm starting to suspect that VCs don't have souls and/or don't interact with any human being outside of potential partners-in-somehow-not-crime or potential victims.
Anime – what started as a niche genre of manga and animation
Off to a strong start I see
iterating on new core loops
We're sorry, but the brainrot is too far advanced. Amputation is your only hope.
Do you think when the Trumps get paperclipped it will look something like this?
Inventor sez "I locked myself in my apartment for 4 years to build this humanoid". Surprisingly, not a sexbot!
Saw an unexpected Animatrix reference on Twitter today - and from an unrepentant promptfondler, no less:
This ended up starting a lengthy argument with an "AI researcher" (read: promptfondler with delusions of intelligence), which you can read if you wanna torture yourself.
Not a sneer, but ~~Kendrick~~ Ed Zitron just dropped.
Its damn good as usual, with Zitron taking aim at the current state of SaaS and tying it into his previous sneers on AI.
Was salivating all weekend waiting for this to drop, from Subbarao Kambhampati's group:
Ladies and gentlemen, we have achieved block stacking abilities. It is a straight shot from here to cold fusion! ... unfortunately, there is a minor caveat:
Looks like performance drops like a rock as number of steps required increases...
correct me if I’m reading this wrong — the results are that LLMs are much, much worse than classical AI at planning block placement for SHRDLU? that seems pretty damning
Yes, the classical algo achieves perfect accuracy and is way faster. There is also a table that shows the cost of running o1 is enormous. Like comically bad. Boil a small ocean bad. We'll just 10x the size and it will achieve 15 steps inshallah.
Imo, this is like the same behavior we see on math problems. More steps it takes, the higher the chance it just decoheres completely. I can't see any reason why this type of thing would just "click" for the models if they are also unable to do multiplication.
I mean this just reeks of pure hopium from OAI and co that things will magykly work out. (But the newer model is clearly better^{tm}! I still don't see any indication that one day that chart is just going to be 100s across the board.)
ah yes, $42, definitely a "the same amount of compute is used" figure
these results are remarkably damning. I knew things were bad, but god damn this is impressively shit
Large Reasoning Models
May the coiners of this jargon step on Lego until the end of days
TechTakes
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community