This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
This is exceedingly relevant to my interests.
Hmm.
You've read Library of Babel, I presume? That library also skipped straight to the generalized object in 100000000D-space and got all possible distributions. Generators have the advantage of a search function, which is a big advantage, but the search space is also very, very large.
The end-state we're looking for is to see "the correct image". we can search coarse, or we can search fine. Coarse search was available before AI: we've all done it when we browse large art repositories, pinterest, image boards, boorus. Fine search is what a mature artist (meaning an artist who has gotten past fighting their hand and eye) does, and it's so much faster and more satisfying that rough search because you are completely in control. you make the image as you please, in all the details, you change what you want and keep what you want, and it all comes very cleanly together. It's a marvelous feeling.
It's frustrating to fuck up a nose or a hand and have to try again. It's likewise frustrating for the generator to not give you exactly what you want. It's the same frustration, in fact: the picture's coming out wrong. And the solution in both cases is to fix it, to push fine control down all the relevant pathways, whether with brush and pencil or ai-driven editing tools. More pathways means more complexity, and either we're back at a hundred photoshop layers, or we're at photoshop without the hundred layers, fine search without the pain.
but it's still fine search, and fine search is art, and the people doing it are therefore still artists!
Like, there's two possible futures. In one, the AI reads your mind so perfectly that you don't even need to ask it for what you want, and all you do is consume. That sounds pretty horrible to me, but I concede that there's probably a lot of people who'd happily take it. In this future, Artists are in fact dead. So are humans, in my view. If this is the desired future... well, hookers and heroin are already available, so what's keeping you?
In the other, the AI can't read your mind, so you have to use it to get what you want. and if you are discriminating in what you want, if you want to tweak the pose, the lighting, sharper in the face, make the hair flow this way, adjust the composition... if you're not satisfied with a image, but want the image, then you're an artist, and the AI is your tool. And you're probably going to have some analog to underpainting and photoshop layers and so on, because control is always going to involve difficulty, discipline, and the resulting disparities in outcome.
This reminds me of nothing so much as the friend who once argued that since there's a limited number of notes available, we're going to run out of new music to make. Well, we haven't yet, and I'd bet we never will... and that's with the relatively constrained palette of tone and rythm. And he thinks that the entire range of thought and experience will produce less variety? Bullshit.
I think he's wrong here. And not a little wrong, but fundamentally misunderstanding what art is and how people interact with it.
Art is recorded. Because it's recorded, it can be reused: one piece of art can satisfy multiple consumers. So if human creativity and human desire are as constrained as he claims, If the artists are making all the same shit, and the consumers just want all the same shit, then presumably there's a cap to the amount of art we actually need, right? Like, once we've written more than a lifetime's worth of books, there should be no demand for more? ...Only no, because people don't want GENERIC EXTRUDED BOOK PRODUCT, they want specific kinds of books. Murder mysteries, romance novels, techno-thrillers, historical fiction...
Okay, so maybe once we've written a lifetime's worth of murder mysteries, we can happily close the genre and rest from our labours forever more? Only again, observably, no. It seems inarguable that desire for art is like the plastic crazy straw design community: It's fractal, there is no bottom. Humans observably throttle their desire for novelty up and down depending on the environment, and I see no reason to believe that there's any practical upper limit to this mechanism. Increase supply by several orders of magnitude, as we have in the modern era, and people simply raise their expectations, grow more discriminating. We have much more art and much better art now than we ever have, and the demand is higher than it's ever been.
90% of everything is crap. 90% of everything has always been crap. 90% of everything always will be crap, because it's not the stuff, it's our assessment of relative value given the environment. That's how humans actually work, how art actually works. Until you get the AI that reads minds, we're still going to be hunting for glimpses of the supernal, on both the producer and consumer ends.
I hang out on Artstation, and I get where he's coming from. A lot of the art there is truly quite repetitive. Likewise, when I read web fiction or manga or watch movies, it's easy to find patterns, even ruts. But that repetition is a consequence of artistic failure on the part of the producers, not a limitation to art itself. They're all making the same shit not because there's nothing else to make, but because their creativity has grown morbidly incestuous. They're all just copying each other's homework, and so it all comes out the same. Then an actual creator with an actual vision comes along and blows them all away... and then maybe they get lazy and start with the nutrient fluid. Humans are weak. But the fact remains that the path is there to those who seek it.
...honestly, I could fill two more text boxes just on this subject alone. My mind is awash with counterexamples, with the ways people find enduring novelty in the most absurdly constrained of categories.
I'll freely admit that I didn't follow most of this passage, so maybe he's just that much smarter than me. But I don't see how "meaning has become an engineering problem" is a supportable statement, or even a likely prediction. I think the guy claiming that it's another tool is strictly correct, because, again, I don't think the machine is going to be reading our minds anytime soon. Obviously if you think we're about to FOOM, your mileage may vary, but absent full-blown artificial superintelligence with no cap, tools is what we're getting.
Indeed, but what does AI have to do with it? The issue of fetishizing invested effort as the prerequisite for validation of end result, a la Marxists and PrimaX3, is wholly distinct from the issue of reward hacking. Cutting costs of procuring enjoyable art or another valuable product to almost nil is qualitatively different from cutting out products as unnecessary intermediate stage for dopamine hits. The former approach calls for a high-level conversation about our terminal values, between moralists, aestheticians, utilitarians, replicators and others. The latter one presupposes values, and resolves very quickly into the reduction of human telos to hedonic utilitarianism and human bodies to hedonium puddles, if even that.
But, that strawman aside – yes. Yes, please. Self-expression is better than raw chemical high. AI reading my mind to directly instantiate beauty that seems contained therein has been my dream since… maybe 5-8 years? As soon as my ideas outpaced my ability to express them. Rigorously imagining is a hard enough job for the artist. It was back then that I realized the impossibility of getting good in every domain that feels aesthetically pleasing, and hard limits to human skill. Were I provided some sort of Neuralink (and there finally are some successes – 1, 2) … With age, all of it shriveled and turned to dust, vague sketches of symphonies, images, dreams, whole worlds, making space for dry rasp of texts. Most likely, little of value has been lost in this case. But something is lost with every child growing up, and it adds up to a substantial sum of unrealized humanity.
It's eerie how problems predicted decades ago in science fiction are coalescing into topical conversation these days, soon to be matters of life and death. In 1964, Strugatsky Brothers have written a novelette tititled «Predatory things of the century» about the oversaturated society where people start to die after finding how to use a commodity radio detail plus some household chemicals to resonate with one's inner desires and conjure an imaginary world beyond anything one can live through. The main character, an archetypal Golden Age Sci-Fi/Social Realist hardboiled hero, takes the drug and then, in his head, debates his predecessor who's succumbed to it:
Dzhilin ends up suggesting a «hundred year plan for restoring and developing human worldview in this polity» but never elaborates. Authors themselves believed that this is an abnormally «Western» book in their oeuvre, full of horror for the future of humanity and proposing no answers. Censors and literary authorities also took issue with it, including the eventual Minister of Culture of the RSFSR Melentyev – who helped bastardize the first print edition, but ultimately deemed his failure to guide this book to an idealistic resolution the only dark spot on his conscience.
…I think we haven't made a ton of progress since then.
Anyway, regarding technicalities. The «fine search» is very fine, about as fine as one's ability to put opinion in words – assuming a smart enough model. Stable Diffusion is pretty dumb, but there are ways to refine the search that look e.g. like this now (taken from 4chan, courtesy of an anon with apparently zero prior artistic experience; I'd contend he's a real artist or at least illustrator now). And of course text-guided img2img/inpainting is trivial; if this were part of some mature editor package, we'd have convenient tools to select, guide and mix features as such, without grasping for words (maybe based on some distant relative of cross-attention) . Will it be a tool? Naturally. But a tool allowing one professional («artist») to make dozens of artisans (and hundreds of Photoshop layers) obsolete.
We do not have unlimited appetite for stimuli, this isn't a game of positional goods. Not running out of music – or art for that matter – is largely due to art's social function, people's limited long-term memory and preference for novelty. We have, more or less, found all melodies, but we cannot listen to the same song on repeat forever. In a sense, content is a means of hijacking the basic reward system (porn especially is) – if you don't change the exact shape of the stimulus, the brain learns to dismiss it. But AIs do variations perfectly.
He's smart indeed, but it's more of a matter of familiarity, and understanding this part is crucial to appreciate the rest of his rants about vectors and dimensionality; all of it makes technical sense. I suggest you start with Jurafsky&Martin's Speech and Language Processing, and see how «meaning has become an engineering problem» is a well-supported statement, not a prediction but a description of the current state of the art in NLP (pun intended). Check out ch.6 («Vector semantics and embeddings»; click this link too). 6.8 covers Word2vec, one specific (aged) engineering implementation of meaning related to the one used in Transformers, and specifically in the text encoder part of Stable Diffusion. Maybe you'll conclude that on its face it's a drastically less rich notion than the ineffable human «meaning», one limited to the space of our text corpora.
But in principle it can be made much, much richer. And I don't need FOOM for that.
One can value improving the struggle > reward loop, or one can value subtracting the struggle and getting all reward.
LTV, as I understand it, holds that labor IS value, and clearly that's not true because some labor is wasted or counterproductive. On the other hand, humans strive, we choose, we see things not the way we want and try to fix them. We grow and change, moving from less to more. Remove that, and what separates us from puddles of hedonium?
What separates this dream from the Minotaur? The AI draws transcendent art from you without conscious effort... Is the art actually transcendent, or is it just hacking your specific reward centers, ordering you to FEEL FEELINGS?
You personally are an Artist, in the sense I've been using the term here. Your chosen medium is the written word, and you are good enough at it that I can't imagine you aren't familiar with the artistic process, of the call of the muse, the desire to create, the experience of laboring to birth something novel out into the world. Certainly, it's a painful, often frustrating process, and full of disappointments and sorrows. You're never good enough, never fast enough, you see all the flaws and the inadequacies in your creations, and they pain you. But hopefully you also feel the high, the delight and wonder of hard work rewarded, when someone gets it, when people respond, perhaps even when their opinions change, or even just having a piece that you can read again with your own eyes and know that for that moment, you bore the sword of creation against the void.
Without the process, the succession of conscious choices, how much of that survives? If the AI draws it all without conscious effort on your part, is it really you speaking? How would you know?
An old joke:
A young man walks up the famous Russian novelist at a party. "Oh, it's so good to meet you," he says. "I've read all your novels, they're transcendent! You know, I've decided I'm going to be a novelist just like you! In fact, I've got a great idea for my first novel, and it's going to be incredible!"
"Ah very good," says the Novelist. "But you know, writing a great novel isn't easy. First you need a good thesis."
"Oh, I've got the best thesis! You've never seen such a good thesis!"
"Good, good. But you also need a good setting."
"I've got an amazing setting!"
"Excellent. But you also need strong characters."
"I've got incredible characters! They're deep, they're lifelike, better than life even."
And the novelist smiles and nods. "Well, it sounds like you're well on your way. Now all you need is five hundred thousand words, and they'd better be the right ones."
There's a particular kind of person I met a lot of when I was in school: the "idea person". They couldn't draw and they couldn't code, but they "had ideas", and they thought those ideas could be their contribution to the project. As they saw it, they would sit back in a comfortable chair and imagine "something cool", and then the rest of us would dutifully get to work actually modelling and rigging and animating and coding their "something cool" into an actual product. I spent a lot of time talking to this sort of person, because there were a ton of them and they all liked to talk a whole lot. Without a single exception, their ideas were absolute trash, warmed-over derivation mashups of things they'd watched or played, "X but with this mechanic from Y", or else completely incoherent, a half-step up from word-salad.
The thing is, there's a feeling of "this is a good idea", a sense of excitement, of infinite possibilities, an infatuation, a mania. But you can have this feeling without actually having a good idea, and not only is it not worth a damn if you can't execute it, it's questionable whether it even exists at all in any meaningful sense. A lot of times, maybe most of the time, it's just a glitch in the brain, premature enlightenment, completely lacking substance. And to the extent this is true, I do not think the AI will help.
What can be executed is what is real. I think it's certain that AI will allow orders of magnitude more execution, and orders of magnitude more success from that execution. Maybe it will also fool those people who have nothing to execute into thinking they're a genius, when really it's just stick figure > greg rutkowski trending on artstation. My guess is that when everything shakes out, the later will feel about as satisfying as the stick figure without the autogen; maybe less.
AI, Nueralink and the rest: does it make effort easier and more fruitful, or does it turn your brain into essentially a very large random seed value? Do you recognize a difference between these two outcomes?
...It's always possible that I've misunderstood, but the question comes down to "what is it all for", "what's the point", right? I think I have an answer to that question, and one I at least personally consider satisfactory. Pleasure serves life, not the other way around. We have a purpose: to grow and to choose, to go from puppets to real people.
I agree that it'll allow artists to produce art harder better faster stronger, by many orders of magnitude. I even think it's likely going to devastate existing artistic hierarchies. I question whether it will eliminate artistic hierarchies themselves, though. I'm confident disparate outcomes will persist, 90% of everything will still be crap, there will still be rockstars and sad sacks.
To some extent, this is a philosophical bet against the reality of the Singularity, similar in form to a lot of other bets I make on other topics. I'll admit that my worldview inclines me toward viewing a lot of techno-utopianism with horror and disgust, and it can be difficult to distinguish what we think should be from what we think will be. Maybe I'm wrong, but I don't think so and I defy the future to prove it.
Assuming I understand the statement, how is it not positional goods? I have access to orders of magnitude more art now than when I was a kid, and much of that art is qualitatively better in every way than what I had as a kid, but I still feel starved for quality. It follows logically that my standards have gotten stricter. Why do we suppose that this heightening of standards won't just keep increasing?
The bet here is that you like something, Game of Thrones or Ghost in the Shell, and the AI means that there's an unlimited amount of it available. The show never ends, or if it does there's another show that hits the same notes, right? And when you get tired of not just the show but the notes themselves, there's an unlimited amount of your next appetite, and the one after that and so on, correct? And not just an unlimited amount, but an unlimited amount of perfect, transcendent quality, the best one can imagine.
Yeah, this was my original argument, almost exactly. Endless perfect consumption is isomorphic to the destruction of the most valuable parts of your mind, in the same way wire heading or endless heroin and hookers are. what's it for? If there isn't a better answer than "feeling good forever", I don't want it, and I'm going to bet against it.
Gotta close the gestalt at least.
Artemiy Lebedev, the self-appointed Graphic Design Guru of Russia, has a famous note in his «Kovodstvo» (... «'Idance») blog: Idea worth negative million bucks. It's all very fair and clever and biting, it's only a shame Artemiy would've been a nobody without his illustrious mother, and his contribution to Russian design is probably measured in negative billions.
Regardless, I understand the contempt that doers feel for idea guys (and ChatGPT shows just how commodified random combinations of premises can be made). It's the same sort of contempt artists on 4chan feel for «prompteurs». Prompts can be auto-generated ad nauseam by a shell script, Greg Rutkowsky had to polish his craft for decades.
Mayakovsky was, presumably, like Rutkowsky. He wrote:
I don't like Mayakovsky or what he moved people towards.
I'm an idea guy. Ideas are like stick figures. But not all stick figures are created equal. One will never become as good an artist as Murata, and it's possible, likely even, that Murata believes himself to be a greater mangaka on the account of his ability to make technical artistic choices; he's wrong and it's a tragedy that One has to compromise with him in producing an artistically competent work. Frankly, One's time was wasted on learning to draw. It'd be proper for him to stick to his figures, and let someone like Oh!Great experiment with styles, to then use their finished forms at his discretion. This is one aspect of AI's promise for human creativity – distilling forms of talent, to recombine them without crippling compromise.
In any case, defining the inherent worth of artistic work through choices made feels odd.
For me it's not a matter of choices. There are instrumental choices made along the way, but mostly just mechanical effort towards crystallizing the idea into its finished shape, reverse-diffusing it from noise towards a predestined output – If I were to abuse this metaphor again. Suppose my fingers are cut off and I have to type with my tongue, or use some BCI with 15 characters per minute; that'd be the end of my participation here – just not worth the time; and moreover, it'd probably be impossible to finish any meaningful piece while the idea is alive and breathing (this post was not finished in time, and so it'll never be 1/5th as decent or 1/3rd as responsive to you as it should've been). It's a fragile thing – a true (pardon my hubris) idea; it's far more than a random seed. It's chosen near-instantaneously from an infinite murky space of triviality, grasped, and the ability to see its glitter is what matters; while this ability is perhaps developed and polished in some relation to iterative effort dedicated to implementation, it has a dimension that lies outside the entire artistic craft and has more to do – with some talent, of course, but more interestingly with unique human experience, that inner achievement which shitty Twitter illustrators purport to convey but probably lack; unlike artists whose fame they hide behind. Rubén David González Gallego has cerebral palsy. He has proven one can be a compelling writer in even worse circumstances than my hypothetical; I can't help but wonder how much more he'd have given the world without his handicaps. Quantitatively and qualitatively more – for the same reason of the impermanence of inspirations. And contrariwise, were it possible to diffuse faster, typing or rather altering the text – or other canvas – at the equivalent of 1000 char/minute globally, improving the work in its every point as the brain sees fit and as an algorithm can do – hoo boy, how some idea guys would shine.
Or not.
We are ankle-deep in the sea of generated creativity, both technical and ideational – watching the tsunami wave as it crashes down. I hope that when it recedes, the survivors will discover some jewels in the wreckage, something essentially solid and irreducible to combinatorial glass bead game guided by random seeds. Some true ideas. It's in our interest that they remain human-made.
More options
Context Copy link
More options
Context Copy link
more detailed replies later, but one quick point:
...putting opinion into words is itself an extremely serious bottleneck. As it's put in that Chad Cerebration meme, "can use an inner voice, but doesn't because he knows it's inefficient." There's tradeoffs either way, because actually putting your opinions into words forces precision that sometimes speeds the search, but for the most part it's serious deadweight loss and a real pain in the hindparts.
That illustration vid is a good example of the dichotomy. he can select elements to edit visually, but he needs to rely on lossy, imprecise language to edit. The prompt box is the bottleneck, in the same way as nose construction and photoshop layer stacks. Obviously the tools are going to continue to improve, but he's still making choices, so by my lights he's still an artist.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link