Does no one realize how racist this assumption can be? Most LLMs are trained heavily on Commonwealth and other standardized English corpora, yet now when people from Commonwealth countries naturally write in polished English, others immediately say it “sounds AI-generated.”
I fear this is the beginning of a really awful trend that will make it even harder for non-white writers to get published.
its so corny and lame to see people advertising their work as "made by humans" like it's a selling point. like you are damning yourself with faint praise you realise that? bright (2017) was made by humans
whether they're Good Media™ or whatever aside, I think mainstream liveplay ttrpg shows have been a fucking disaster for the hobby. it's hard to imagine anything that could have fucked the expectations:results differential for people more than having celebrities do college improv with dice (and an entire media production team behind them) and telling a generation of new players that's what tabletop gaming is like
there are many people who like to walk into a ttrpg having invented a blorbo in their mind and have the ttrpg to give them the tools to realize that blorbo and all the best to them i am so thoroughly not one of those people. give me a playbook that prescribes a very specific narrative role. give me a pregen character i have to play as. give me a dozen big stupid tables to roll up a guy on. give me some god damn lifepaths !
Okay, but then, why did you sit down at the table? Why are you playing a TTRPG? You don't want to use the mechanics of the game to express a narrative for your character, why are you bothering to play the game at all?
You could just go do improv. Or play a video game. Or look into local theater. If you aren't going to create the character, why did you bother showing up at the table?
-new players should *never* be brought into a game without making a character. If they do, then they do not know the mechanics well enough to use them to play the game. This makes running the game much harder.
-This is a TTRPG, why are you not playing a puzzle game?
-This is a TTRPG, why are you not playing a video game?
-This is a TTRPG, why are you not just going outside and engaging with world?
We are talking about TTRPGs. You don't do the character creation, you aren't playing a TTRPG, you are playing an improv game. In which case why did you bother going to the effort to get the materials to lay the game?
This is a TTRPG. It comes with character creation. If you don't want to do that, why are you playing a TTRPG?
Gonna be honest a lot of people deep down view cheating as worse than abuse which is why so many people view downright controlling and manipulative behavior in a relationship as 100% permissible so long as that behavior is centered around either preventing or discovering cheating.
I think its because even if abuse is more harmful than cheating - abuse maintains the shape of a relationship and cheating warps it.
With abuse you can have your abuser "come to jesus" and we can sweep all that nasty abuse stuff under the rug as a side effect of just how big there feelings were and the nuclear family is maintained.
If the abuser doesn't get better you just paint them as the embodiment of absolute evil (you're basically forced to - if you don't noone will understand why you're not trying to fix the relationship instead) and start anew.
Cheating introduces a new character. How are we suppose to explain to the Johnsons or grandma that there's this new girl sortof in the picture but not really. Is the family steve, daisy, and steve's mistress samantha now? Who is mom? Is there gonna be an extended period where all these questions have ambiguos answers?
Abuse might hurt people but you can just demand people toughen up and forgive him. Cheating hurts the instution of monogomous love and that (unlike people) is fragile and must be protected.
I am so utterly fascinated by “Saki”, the 18-year-running mahjong manga in which you, the reader, become gradually, frog-boilingly aware (over the course of nearly two decades’ worth of mahjong tournaments) that none of these girls are wearing underwear and most of their boobs are slowly expanding.
I need you to understand that I have, like, an anthropological level fascination with this comic. From the perspective of someone who is also a comic artist and writer, two things delight me about it:
the fact that I understand completely how an artist gets from “the fans can have a little hint of skirted asscheek” to “the pussy is completely out on center page” over the course of 18 years; and
the way in which the pussy being out is treated by the characters and diegesis as being utterly unremarkable.
Okay. Point 1. The frog-boiling.
Let me put this in perspective for you. There was already a meme about how the characters in “Saki” don’t wear underwear when I was in middle school. I am thirty now. Okay? And it’s still going.
In the time since, this has stopped being a joke. It is now indisputable canon. This is not because anyone outright says it at any point. It’s because the underwear ran out of places to hide. I’m obsessed with this thought: somewhere in the over 20 volumes of “Saki”, there is a panel in which underwear was objectively deconfirmed. And it would be so hard to figure out where that panel actually is. Maybe the artist didn’t even realize it when she drew it! The frog? Boiling!!
And of course there is also the breast expansion. I don’t know how to put a spin on this. They are just expanding. Like, this happens a lot with artists: you define a character as being, in your mind, “the one with the big boobs”, and over the years you emphasize that trait further and further so that the signal doesn’t get lost in the noise. It’s just that normally—in like a wildly popular manga series about mahjong published by literally Square Enix, for example—normally there would be a point at which the boobs stopped getting bigger. Like, an editor would step in or something. Or you would get to the point where you cannot draw the character in the same panel as her mahjong tiles without her breasts spilling over the tiles, and you’d go, “Well, this is now untenable.”
That did not happen. There is no ceiling. The frog is soup.
Point 2. The complete and utter mundanity of all of this.
It’s like this, okay: there’s no shortage of trashy ecchi manga out there. There’s a million other comics doing wildly bawdier things with wildly more improbable bishoujos.
The vibe with “Saki” is different.
It’s hard to explain this, but it feels like the world of the comic is fundamentally uninterested in the fanservice happening on the page. I cannot describe it as “leering”, because I cannot conceive of a person in the story from whose point of view one would leer. I think the artist is probably into it—I can’t imagine anyone is making her do this—but “Saki” the comic has no opinion on the matter.
There are essentially no male characters in “Saki”. Like, there was one guy? Kind of? At the very beginning? But he is gone now. They put him back in the toybox. He does not exist. It appears to be some level of canonical that in the world of “Saki”, almost all humans are women. Those women are sometimes romantically into each other. According to comments the artist has made on Twitter (which I cannot source), they have lesbian baby technology, so it’s no problem. It’s so much not a problem that the story is about mahjong, instead of any of that.
So, like, the fiction here appears to be this: this is the, like, meta-narrative of the fanservice of “Saki”, right: it’s just normal that they don’t wear underwear and their boobs are arbitrarily big. It’s been normal. It was normal before the story of the manga began. It’s just how things are. Nobody bats an eye about it, and if they do, it’s in sort of a lesbian kind of way so like what’s the problem, we love lesbians here. This is literally normal for girls.
The fanservice simply diffuses into this all-encompassing aura of disembodied, ambient sluttiness. The framing of the panels demands you acknowledge it, and the story demands you already be over it, because it’s mahjong time now, and we’re playing mahjong.
Do you get??? why I’m so fascinated??? Are you not a little enraptured???
Anyway, I have no idea how to end this weird post. I guess the conclusion is that women stay winning????
I have so many questions... How does one SUSPECT a manga character isn't wearing underwear? Like, sure, boobs are front and center amd you can see them get bigger panel by panel but how does this work for panties? Are there just that many upskirt shots?
Also how do you keep a manga about Mahjong going for 18 years, what??
Are there any resources you'd recommend reading? (of others, of yours, on Tumblr, off Tumblr, whispering from tv static)
That is an impressively broad question.
On Tumblr:
@cloakofshadow produces good poetry and good short fiction, which are not easy things to find, here or anywhere.
@materialist-scumbag is an explicitly AI-produced blog, and you can feel about that however you feel about it, but - at the very least, it's a very prolific source for "here is a cool thing that exists in the world and I am going to tell you about it."
@cryptotheism is into All the Right Arcane Shit.
Off Tumblr:
I'll repeat a piece of advice that you've undoubtedly heard too many times already - yes, you should read the classics. If something is famous and has survived the test of time, there's probably some kind of reason for it. At the very least, even if you read a classic and hate it - being familiar with it will put you in a good position to engage with other texts, and with the broader discourse.
(I say "read" because I'm thinking of books, but this is just as true of classic movies, classic video games, etc. The literary canon is much longer and older and more reputable, though.)
I particularly recommend the kind of classics that, in addition to being old and good, are absolutely fucking bonkers. Moby-Dick, the "Four Great Classical Novels" (and any other old Chinese text famous enough that you've heard of it), etc.
If you're interested in theater LARPs - and you should be - you can find some cool ones at Paracelsus Games. You can, at least, find out what the deal with theater LARPing is, to some extent.
If you're interested in reading a demon's prose-poetry ramblings about the good life, I hear that this book sure has some stuff in it.
Thinking about that that "slop accelerationism" post, and also Scott's AI art Turing test.
I also hope AI text- and image-generation will help shake us loose from cheap bad art. For example, the fact that you can now generate perfectly rendered anime girls at the click of button kindof suggests that there was never much content in those drawings. Though maybe we didn't really need AI for that insight? It feels very similar to that shift in fashion that rejected Bouguereau-style laboriously-rendered pretty girls in favor of more sketchy brush work.
But will we really be so lucky that only things that we already suspected was slop will prove valueless?
As usual with AI, Douglas Hofstadter already thought about this a long time ago, in an essay from 2001. Back in 1979 he had written
Will a computer program ever write beautiful music?
Speculation: Yes, but not soon. Music is a language of emotions, and until programs have emotions as complex as ours, there is no way a program will write anything beautiful. There can be "forgeries"—shallow imitations of the syntax of earlier music—but despite what one might think at first, there is much more to musical expression than can be captured in syntactical rules. There will be no new kinds of beauty turned up for a long time by computer music-composing programs. Let me carry this thought a little further. To think—and I have heard this suggested—that we might soon be able to command a preprogrammed mass-produced mail-order twenty-dollar desk-model "music box" to bring forth from its sterile [sic!] circuitry pieces which Chopin or Bach might have written had they lived longer is a grotesque and shameful misestimation of the depth of the human spirit. A "program" which could produce music as they did would have to wander around the world on its own, fighting its way through the maze of life and feeling every moment of it. It would have to understand the joy and loneliness of a chilly night wind, the longing for a cherished hand, the inaccessibility of a distant town, the heartbreak and regeneration after a human death. It would have to have known resignation and world-weariness, grief and despair, determination and victory, piety and awe. In it would have had to commingle such opposites as hope and fear, anguish and jubilation, serenity and suspense. Part and parcel of it would have to be a sense of grace, humor, rhythm, a sense of the unexpected and of course an exquisite awareness of the magic of fresh creation. Therein, and therein only, lie the sources of meaning in music.
I think this is helpful in pinning down what we would have liked to be true. Because in 1995, somebody wrote a program that generates music by applying simple syntactic rules to combine patterns from existing pieces, and it sounded really good! (In fact, it passed a kind of AI turing test.) Oops!
The worry, then, is that we just found out that the computer has as complex emotions as us, and they aren't complex at all. It would be like adversarial examples for humans: the noise-like pattern added to the panda doesn't "represent" a gibbon, it's an artifact of the particular weights and topology of the image recognizer, and the resulting classification doesn't "mean" anything. Similarly, Arnulf Rainer wrote that when he reworked Wine-Crucifix, "the quality and truth of the picture only grew as it became darker and darker"—doesn't this sound a bit like gradient descent? Did he stumble on a pattern that triggers our "truth" detector, even though the pattern is merely a shallow stimulus made of copies of religious iconography that we imprinted on as kids?
One attempt to recover is to say Chopin really did write music based on the experience of fighting through the maze of life, and it's just that philistine consumers can't tell the difference between the real and the counterfeit. But this is not very helpful, it means that we were fooling ourselves, and the meaning that we imagined never existed.
More promising, maybe the program is a "plagiarism machine", which just copies the hard-won grief, despair, world-weariness &c that Chopin recorded? On it's own it's not impressive that a program can output an image indistinguishable from Gauguin's, I can write such a program in a single line:
I think this is the conclusion that Hofstadter leans towards: the value of Chopin and the other composers was to discover the "template" that can then be instantiated to make many beautiful music pieces. Kind of ironically, this seems to push us back to some very turn-of-the-20th-century notion of avant-garde art. Each particular painting that (say) Monet executed is of low value, and the actual valuable thing is the novel art style...
That view isn't falsified yet, but it feels precarious. You could have said that AlphaGo was merely a plagiarism machine that selected good moves from historical human games, except then AlphaGo Zero proved that the humans were superfluous after all. Surely a couple of years from now somebody might train an image model on a set of photographs and movies excluding paintings, and it might reinvent impressionism from first principles, and then where will we be? Better start prepare a fallback-philosophy now.
This an interesting angle. I've never been particularly taken with trying to determine True Art from False Art on the basis of specific qualities of the piece.
(I did ok-ish but not exceptionally on the AI art quiz, probably with a slight bias for misattributing human pieces as AI ones - like many other respondees, I found the Impressionist pieces hardest to distinguish, since they very much play to the AI's strengths.)
There are many different ways you could describe "art" as a human activity, I'm sure there's a post somewhere where I make a list, but a really big one is its communicative function - one purpose of art is to somehow pass on some aspect of our 'inner world' to another person, through a lossy and limited channel.
That a signal can be easily imitated doesn't mean it doesn't carry contextual information. For example, I could ask a yes/no question of enormous emotional importance - "should I take the shot", "has the baby been born", "will you go out with me", "am I a good girl" - and be answered with either "yes" or "no". It would be trivial to generate a machine which randomly substitutes for this communication - that's basically all a magic eight-ball is.
The amount of information that can be contained in an image of a given size and colour depth can also be calculated. For example, the number of fullscreen images that would fill my current monitor at 8 bits per channel would be 2^(3440 × 1440 × 24) ≈ 3 × 10^35788372 - about 15 megabytes uncompressed. It's a number that seems astronomically huge, though effectively the amount of information is much less than you'd calculate since all the likely pixels are correlated. The same goes for other art forms, like novels (encoded as, say, UTF-8 strings or PDFs) or pieces of music (encoded as sound files, MIDI, MuseScore files, etc.). The exact number is complicated, you end up getting into Kolmogorov complexity and shit like that, but the point is that it's finite.
If we want to claim that all the information about a human life that Hofstadter describes (grief, despair etc.) is in there somehow, we're claiming that this finitely many bits is adequate to capture all the nuances of a human life. I don't know that that's true!
This, however, doesn't really seem to align with how we interact with art. Human production and exchange of "art" is a social act - I would describe it as being continuous with "play". When we observe a piece, we are opening a communications channel - at least a one-way channel. The person on the other side sends some information into the channel, and we process it somehow.
Since it is a lossy channel with limited information, we must infer various things about the other mind on the opposite side of it. If I show you an artwork that I made, we might have a conversation about how I did it, why I made the choices I did. If I feel something looking at the work, I might imagine that you felt something similar, and designed the piece to evoke it intentionally (a guess that will often be wrong but sometimes still productive). I might also look at what specific choices you have made, compare them to the choices others have made in the same medium, etc etc.
We form these inferences on the basis of experience - the more you learn about making art, the more you learn to appreciate other peoples' art and vice versa. And we project these experiences, usually plausibly, onto other artists.
(Perhaps I am saying all art is in a sense performance art? Seems like a tasty soundbite, though I'm not fully sure I wanna commit to it.)
I'm not meaning to claim that a computer couldn't simulate this kind of 'how did you make it' interaction too. This line of argument was anticipated by Turing in his original 1950 paper on the 'imitation game' that someone links in the comments above, where he describes a poet undergoing a viva voce test interrogating their word choices, and argues that a computer might be programmed to give convincing answers to such a test. I imagine he's right - for a paper written in 1950 he makes some surprisingly sharp predictions for how future AIs might be made, such as the idea that an AI could be built to be 'educated' like a child. (He also thought the evidence for ESP is 'overwhelming', but hey, can't win 'em all).
A lot of the context around art would be quite easy to forge, had you a mind to. For example, suppose I go to a film screening, and someone is introduced as the director so we can all clap them. Did they really direct it? I don't know! You could totally send an actor. Less conspiratorially, if someone says they made an artistic choice for x or y reason, they could be lying about it, or misremembering, or most likely oversimplifying a complex and inscrutable process down to a simpler story.
At some point you have to take something like that on trust, or else simply accept that being lied to about it is part of the game you're there to play! (c.f. Oshi no Ko.)
Anyway, the sudden arrival of a new process that can produce, at least sometimes, near-indistinguishable output to various types of communication, throws a spanner in the process. If we're feeling uncharitable, we could call it something like a DDOS attack, stuffing the channels with spurious inputs that don't fit our design assumptions. I think that goes too far, though. AI gen doesn't preclude communication, but it does need we need to think differently about what is being communicated.[1]
So to consider that last question, if art is like a game, could you train an AI art to produce art that is meaningful to humans only by 'playing against' itself, like AlphaGo Zero? I don't think this is so likely. The rules of Go are strict and well-defined; the rules of what humans find meaningful are inseparable from the history of interacting with other humans, which is why art constantly evolves. Training an AI on existing human artworks is training it to compress and interpolate/extrapolate that dataset; training it to optimise for "making novel art that expresses something in a form that its interlocutor could understand" requires it to be interacting with someone.
You could imagine a training process with an "artist" AI and a "critic" AI (a sort of more sophisticated GAN, where the adversary is optimising not to distinguish human/AI art but to judge it on aesthetic grounds) - but how would you get the "critic" AI? Whose taste would it express?
Admittedly, the developers of image generators are constantly refining their models in response to users, so they are being optimised to appeal to someone, not just interpolate existing artworks. But I think it would be very hard to remove humans from the equation entirely. And the present means of providing feedback to the AI are very crude.
For an AI to learn from interacting with other AI (and the world), I feel like you'd need a whole new process that isn't about minimising loss against input-output pairs. Romantically, I imagine it would be closer to how humans learn from life, but I don't really know what will 'work' in the end.
below: some other remarks that were excised from the main post.
[1] We can view AI image gen as another channel for communication between humans, with its own set of inferences to make. If someone shows me a picture they've generated with AI, there's no point asking why they painted this bit that way, but I might approach them more as a curator and ask why they chose this generation over others, or how they went about prompting it.
The AI artists who go to the trouble of finetuning their models with LoRAs for a specific end goal, or using more involved processes with multiple stages of generation, probably have most to 'say', either through the work they generate or how they'd discuss it. (I find it very endearing when someone trains an AI to serve up a hyperspecific fetish.) And the more I know about how AI images are generated, the more I can probably have a productive conversation.
In this light, the "problem" of AI is mostly one of deception, insofar as it tries to look like something else and thereby tell a misleading story. That's probably a big reason why why it brings the rancour it does, although it doesn't explain all of it. It's not (usually) a forgery of a specific human's work, but it is designed to forge spurious communications in this channel in general, so the channel is 'noisier' - and this could be thought to undermine many of the contexts, i.e. the operating narratives and social games, which are why we exchange art in the first place. Over time, we'll presumably end up renegotiating the 'games', and spawning new ones, as humans always have.
And of course, the issue of provenance and plagiarism in art - particularly when prestige and money get involved - long predates AI and is full of all sorts of bizarre contortions when you look at it closely.
More intriguing is whether there is some possibility for "real communication" between humans and AIs - that is, could there be an AI output that does respect the 'rules of the game' in some way. This is harder to imagine! Like, if you ask why we aren't solipsists, we could point to how much we resemble other humans and say, all things considered, seems very unlikely we aren't the same type of entity. But I only know 'what it's like to be' a human. Conversely, while I know a reasonable amount about how AIs work, the attention mechanism and latent-space vectors and so on (thanks 3blue1brown), the analogy isn't so clear anymore, so I don't even know how I'd determine whether there even is a 'what is it like to be' under all the 'noise' of communications aggressively optimised to fit the patterns of something a human might say. If there is, it's probably very alien to all of my experience.
Ironically I feel like the current model of 'AI', which teaches us to regard any generated output with suspicion of having 'nothing behind it', would make it harder for any 'real', agentive, subjective-experience-having AI to make itself known to us. But perhaps it's good that we're forced to sharpen our criteria of what we're looking for out of these things.
Anyway, all of this is probably just idle imaginings, because nobody can figure out how to make anything like enough money to justify the exorbitant costs of training and operating AIs, so at some point this whole speculative bubble will go up in smoke and whatever AIs continue to be in use will likely remain about as good as they are today, or stupider - at least until the next 'AI summer' when a new paradigm emerges.
In addition to this take, that art is communication and what matters is what the sender meant, not the bits of the message, there were also some replies to this post saying that what matters is whether the art itself if good, not how it was made. I think this disagreement shows that it's a tricky question...
Anyway, I wanted to expand a bit more on what I had in mind about a computer rediscovering Impressionism.
Actually this is inspired by a quote I read (I unfortunately can't locate it now) by some Chinese calligrapher, who had a kind of deflationary view of a past master. Something like, while the work had previously been said to "reveal the soul" of the characters, he said that they were made by exaggerating the variations to make the thin parts even thinner and the thick parts thicker. I think this analysis makes the calligraphy seem more "shallow". The effect can be obtained through operations directly on the pixels, without an understanding of the deep meaning of the writing
Maybe this is obvious because characters don't have any souls in the first place. So let's instead consider people. There's a classic paper from 1985, where somebody made a program to generate caricature drawings by comparing the facial landmarks to an average face, and then increasing the distance:
(Indeed, this is similar to what caricature tutorials say you should do, so I think it may be a realistic model of what's going on.)
But this could be worrisome! When we look at a caricature of JFK, we might feel that the artist really captured his soul. We look into his eyes and feel his wit, intelligence and arrogance shine through, which tells that the artist well those passions read. But if we did think that, then this computer program reveals that we were deceiving ourself, because caricature can be done by just computing a basis for a vector space of images of faces without any understanding of the passions at all.
Now what about Impressionism? It too might involve deep understanding: "people must first of all learn to look at nature, and only then may they see and understand what we are trying to do" —Monet.
Or, it may mostly be a matter of changing the color palette (get rid of black, add mauve) and using coarser brush work ("try to forget what objects you have before you, a tree, a house, a field, or whatever. Merely think, here is a little square of blue, here an oblong of pink, here a streak of yellow"—also Monet). In the latter case, maybe you could take a random photo, project it down to a basis of limited brush strokes, and achieve the same results that won the Impressionists fame without learning to look at nature at all, without even having a semantic notion of house, tree, or field?
This would reveal impressionism, too, to be more shallow than we might have hoped. It might be something which is not tied to human emotion, or even human visual perception, but could be done by random aliens based only on computing a latent space of variations from a collection of images.
kianamaiart
dailybridgerton
yousharknotpass
snowghoul
baratheonpilled
prguitarman
