The AI Unbundling
stratechery.com
stratechery.com
Dropbox made filesharing easier. It's a good product, but not disruptive. Nobody panicked that Dropbox would make their job redundant.
Uber was hard on the taxi industry, but fundamentally you still have drivers taking people from A to B. First you had to call the cab company now you use an app. It's an improvement, but not truly disruptive. Not like level 5 self-driving cars would be.
Amazon and Walmart by contrast did disrupt entire industries. Independent book shops and mom&pop retailers saw the writing on the wall. They knew they couldn't survive facing this kind of competition, and largely, they didn't.
Stable Diffusion (and similar tools) fall in this last category of truly disruptive technologies. It's going to destroy the livelihoods of the majority of independent artists in a way that looks inevitable to me. These new tools boost artist productivity by 100x and that means good artists will be producing much more art than ever before. This pushes cost down and quality expectations way up. Some artists will adapt and thrive in this new environment, but the majority won't. It won't be long until making a living with photoshop will become as hard as making a living playing guitar. This is good for society but bad for many individual artists.
There will be a backlash. People will insist that SD art isn't real art. Artists will fight back, and lose. Because SD isn't going anywhere. This is what disruption looks like, and it isn't pretty.
When talking about Stable Diffusion and art there are usually two different aspects of art. I am not going to try to define art, but sometimes we refer to art as illustration, or stock images (what SD puts in danger) and some as broad modern art.
I am no trying to say that one is more valuable than the other, but want to qualify these two, because some art is not painting pretty pictures.
In the modern art interpretation artists will not lose to SD. SD will enable them to do different things. There are many examples of famous artists that commission the execution of an artwork fully, without painting, sculpting or doing any other work. I remember an example of an artist paying illegal immigrants to hold a wall (that could not stand by itself) in a gallery, to touch on social issues.
I am not an expert in art. My point is, in modern art SD will be one tool more (although it might create new influences) at the service of the human that gives it meaning.
If they charged a small amount per submission probably the would stop users submitting 40 images per second.
The surprise is not because it’s furry — the art scene is such a large fraction of the furry economy that the introduction of AI art is approximately equivalent to introducing modern farming to, IDK, the 1850s or something — but rather because of that site’s lack of reputation with regards to technology.
Although, perhaps that explains why the quotation is “That content generated can reference hundreds, even thousands of pieces of work from other artists to create derivative images.” rather than the billions in the training sets.
I still don't really know what to attribute the attitude to, but it seems to be a pattern.
But do these artists have anything insightful to say about inequality or exploitation? No. It's always trite, reductive, and utterly unoriginal. The truth is that "statement art" is, almost without exception, just bad art, but people wrongly believe that: (a) there must be a deeper meaning to it that they don't understand or (b) people who say art is bad are philistines. And if art is just supposed to make you think, even when the artist didn't have any deep thoughts themselves, then SD art can easily meet that benchmark.
SD art will be good enough to decorate your home and office with. SD art will hold its own in art galleries, if it doesn't already. That's much broader than stock photography.
If I am right, it would behoove you to not speak so surely of that which you do not know or care to understand.
Mind if I use them in my gallery's modern art exhibition?
If you do — touché; well played to you Sir!
Much as I love my friend, after the exposure I’ve had to modern art I have to agree that it is almost without exception meaningless trash.
Look, let’s just be real for a moment here. Take a step back and ask yourself what’s more likely: that (A) there really is something deep and mysterious about the banana-on-the-wall or (B) most actual artistry disappeared with the invention of photography and modern artists are LARPing, sustained by taxpayer handouts and the financial fumes of the money laundering and tax evasion that high end art is used for.
Modern art is a joke, and everyone including the artists know it.
I can tell you’re the kind of thoughtful serious guy who likes good, tasteful and neccesary things, and yet! You dislike bad, tasetless and unneeded things. I do too, and so does everyone else, on balance.
Ps: I feel obliged to remind you to not say the quiet part out loud. Not because the elephant in the room might trample you, but because at least one reader had to muster heroic amounts of good faith to not dismiss out of hand the substance of your post due to form. And even still, I could not abstain from this disclaimer.
I remember this specific example because the artist was paying the workers to do a pointless job. Technically, she was breaking the law. If only for that I thought it was interesting. I am not saying I always like this kind of performance art, but sometimes it reaches you in a way that a painting could not.
What I and some insiders consider the real art - that is, applying rare well-developed talent to production of unapologetically high-quality sensual artifacts of art - is mostly commercialized by now, with best people employed by the media corporations to produce assets for high-value games and movies. I have nothing but respect for the talent it takes to create something that is valuable on its own (contrast with modern art, not valuable without of the context of its social network), and feel a tinge of sadness seeing that SD will move the equilibrium here.
The real crux of the matter is that we should provide artists with some decent UBI guarantee, and this should be a humane solution to impending poverty, for each of the professions that are going to cease being commercially viable in the near future.
All art is only valuable within the context of its social network. If you leave the Mona Lisa in the woods, the only real "value" it will hold is as a shelter for bugs.
I think what really bothers you here is about which social networks define the value for certain kinds of art. High technical skill art in service of mass media artifacts like videogames and films is sort of the most "democratic" of uses of art. Almost everyone can perceive its value and it asks little of the consumer in return.
Contemporary art, the kind of stuff people here on HN hate, is different. It is deliberately created to provide value only to the small set of people with the context to appreciate it. It's sort of a continuation of a very long conversation that if you haven't been on the inside of, you miss out on. That leads naturally to valid claims of elitism. (The fact that it's also used as a large scale money laundering enterprise by the very rich certainly doesn't help.)
The unfortunate part is that the collateral damage this inflicts on the general idea of niche art. Anyone who has ever created can tell you that the more specific of an audience you target, the more deeply your art can move them. If you're trying to write a song that a billion people will like, it can only be about the most banal of platitudes. Now write a song about what it's like to lose one's spouse to alcoholism and turn to drinking to deal with it. Few people can relate to it, but the ones who can, well, you can pierce their soul.
The idea that you need context to understand a piece of art is totally valid and one of the most important tools an artist has at their disposal. Likewise, it's not a failing when a piece of art only aims for a small targeted audience. Don't let the snobbery and elitist trash of the contemporary art scene taint those concepts.
Let's notice the elephant in the room: HN commentariat is a biased sample with an obvious overrepresentation of autistic people. I'm likely not too much ahead of the curve here, as you can see.
I hope this pretext will add to your charitable interpretation of what I have to say. As they say "autism speaks, it's time to listen". Consider this a hypothesis.
I cannot say this condition is overwhelmingly beneficial - honestly I can interpret it as a bit of a handicap, something that, among other effects, makes the child disinterested in people around him in his early formative years (yes it's most commonly "him"). Instead the child is made to focus on the world and things in it on his own terms, as a one-of-a-kind being starting from the almost pure blank slate - with the blank slate here, perhaps, being an overgrown PFC full of fresh synapses. This makes learning one's world and one's inner depth harder compared to normal individuals - which have the right set of biases and heuristics to start imitating other people early on, standing on the shoulders of giants right away, learning the behaviors that passed through many aeons from parents to children and quickly arriving to a socially-bootstrapped form of self-awareness. And for the autistic One the world (and, apparently, the self) is an alien world which is being conceptualized for the first time ever, and the society is an alien configuration of beings which feel and behave quite differently from how one would expect them to, if they were kin to the autistic. Even when one grows up to like it, it's still like being an eternal foreign student everywhere you go - yet your relationship to the world you found yourself in becomes deeply personal.
What I'm trying to say is this predicament makes one much more likely to learn the world on the world's terms, same with oneself and one's feelings - beauty sense included. Having developed in this way, the social concept of beauty you describe is a vastly alien thing to me - I can understand it, but it's like understanding physics of some abstract phenomena. Myself, I just feel beauty, as it developed inside of me as a palette of feelings resonating with certain structural patterns in the world's percepts I come across - something having to do with information-theoretic regularization in primate neocortex, as I may guess. No social reinforcement was necessary to arrive to this feeling, it has a sense of finality and self-referentiality to it - something basic and indispensable, something valuable.
Does it sound scary, cold and alien to you, or simply incoherent? That's how your idea of socially mandated value sounds to me - shallow repressiveness of one's social graph pushing supervised learning examples into your very soul, re-flashing it with the weights of the collective simulacra, of the lovecraftian entity - a disembodied communal sense of value (can't even call it "beauty" as in the limit it's completely arbitrary, a value-language defined by social custom) - to make you one of them, a part of the eusocial organism? This is 1984-level scary, reeducation camp-level, basically.
... And with these new neural networks I immediately felt some sense of commonality, as if they were from the same metaphorical planet I came from - tiny blank slates learning compressed sense of the world as it passes through them, their beauty and my beauty approximating the same information-theoretical Eidos.
However I must reject your claim that your experience of Autism gifts you with some sort of objective view of the world, art and aesthetics that allows you to judge all contemporary art as 'trash.'
That the gilded creators of «the real art» want your pity? To be construed as without agency in all of this?
You elitist buffoon, you know not of what you speak. You know what you like, but you dislike and question the motives of that which you do not know nor care to. Keep it to yourself.
Period.
We're watching the Third Industrial Revolution unfold live and in front of our eyes, and it will only take a few short years to fully sweep into all aspects of society.
Yes, Picasso and Banksy will continue to have value. But that's an infinitesimal sliver of the value that is about to be unleashed upon the economy.
The real value is that startups like mine will give anyone the power to generate new Taylor Swift songs [1].
New movies, new music, new games. All tailored to the nichest and most exacting of interests.
You want the steampunk vampires of Cloud City, featuring Sean Connery and young Eddie Murphy duking it out while riding on the backs of space whales? You'll have that in just ten short years.
If you think TikTok adequately captivates the minds of the population, you have no idea what a finely tuned search for people's true interests will do.
Humans won't be good enough at satisfying other humans anymore. Not without aid.
The new world is going to blow our minds, and it's already beginning to unfold.
In about a year, I don't know why anyone would dare work for Google, Amazon, or any of the other legacy businesses. They're going to be swamped and struggle to stay afloat amid the biggest single act of the Innovator's Dilemma to ever happen in our history.
[1] This is happening. My startup does it.
Eg. https://www.youtube.com/watch?v=r_0JjYUe5jo --> https://vocaroo.com/1hgjjnVNqWjk
(We'll have the Obama and Taylor Swift versions up on our site soon.)
Our video stuff is under development, but I'll show you something analogous that a skilled competitor is doing in the open (though they're underselling themselves) : https://imgur.com/seBTPG8
The whole world is about to be carved up and reallocated.
I'm going to get back to work. See you on the other side.
Well, at least your drug dealer is apparently supplying you with the good stuff. I'll bet 10k any day that the combined value of all companies purely in the generative space will not exceed the value of a single FAANG company in 5 years. Yes, including Netflix.
If the art made you think about a deeper meaning, is it better the artist put it there for you to find, or that it enabled you to freely think? I think too many people have “High school English class” syndrome. There’s no single right answer to extract from looking at art to give you that A+. It’s about self reflection, reflection on the experience and context of the art, and learning about the world around you. Everyone has a different experience and a different perspective and everyone gets something else out of it.
Can this be accomplished with SD art? Yes, but the curation of the model, or exact wording put into the model is still needed to provide some expected output that evoked the emotion. Artists may even modify an output as a starting point or inspiration, but “gallery” type art isn’t consumed for simple entertainment value like an Instagram photo feed, but as a destination experience (for sufficient expense to fund human artists too).
> It's always trite, reductive, and utterly unoriginal.
This is simply untrue, modern art is not "always" anything. And who are you to judge what is going on in the mind of an artist?
I don't know what decade you deem to be "modern" art, but there is a huge amount of 20th - 21st century art that is jaw drop. You are the same as generations and generations of people who have been wrong and stuck in the comfort of the status quo. I hate to refer back to Van Gough, but you are a version of the people who missed the point there. And I suppose the social commentary of Pop Art is just trash too?
AI art will hang in a gallery if it evokes a feeling, not because it's a bloody good execution of a painting of a tree. This is how ALL art has been since we moved beyond painting wealthy people. It has to capture an emotion, thought, idea, or whatever else is intangible. Being able to get that intangible feeling across to the viewer is what is special. Not the quality of the craft, that's artistry.
I truly feel sorry for you not having been moved by something more than a beautiful oil painting.
What would be some examples of that? Not disagreeing, would just like to have my jaw drop
https://phantasmagoria.stavros.io/images/RSXTZVMjYhTdyTmE/
I want to hang this on my wall.
This isn't a cheap shot on art (the same can be said for many things people enjoy, including much of fiction). It's to say that I think that this will replace "art as a low margin service" but not "the art world."
Sadly, the market for the kind of art that the these models cannot disrupt is relatively small compared to the kind it can. There aren't a lot of people making a family-supporting living doing installation art. Most people I know with art degrees do illustration, photography, or are part of a video game or film asset production pipeline (or they draw tattoos, but that's a different matter). If I were them, I would be looking at the next generation of this technology as a potential threat to my livelihood. I don't want to be alarmist, but it is a possibility, and it'd be weird to dismiss it.
One other thing I'll tack on here is that I find it fascinating that the kinds of skills required to be "good" at using these image generation models — "prompt engineering" if you like — are largely different than the ones required to create art from scratch. You can be a great studio painter, but not be able to "talk to" Stable Diffusion at all. Likewise, you may have zero artistic ability in the traditional sense, but be a prodigy at getting the computer to spit out what you are imagining, or something even better than that. If AI generated art is determined to be a kind of art (as I believe it will be) the parameters of what we call artistic ability may change.
I'm not so sure about this. Both require visualizing something upfront, evaluating the resulting images and understanding what needs to change.
It's a bit like digital and analog photography. Digital makes the skills of getting a good exposure less relevant and allows post-processing. But a photographer still needs to know what makes a good photo.
This doesn't invalidate the original point though. Making a living as a photographer became harder with the advent of digital photography.
But then, if you're a great studio painter, SD doesn't threaten your livelihood in any meaningful way, whereas if you're the sort of artist that produces video game assets, you've probably already got a keen eye for spotting rendering issues, and prompt engineering and inpainting and producing concept images AIs can work with are all likely to be skills you pick up considerably better than the average person... and then it's just another tool.
As someone intimately-amateurly involved with AI artwork over the past six months, I've come to disagree with the term "prompt engineering" as a description of the skillset. It's not just feeding it a good prompt (which is not that much of a puzzle once you get into it, and is close to being a solved problem in my book), but instead a whole iterative pipeline of human-in-the-loop hacks and touchups in order to make up for the tech's existing faults and unwieldiness. Many aspects of this pipeline, interestingly, are already familiar to digital artists.
I think the specifics of this pipeline will continue to change on a weekly basis until we reach a plateau in the technologies. But, novelty lies at the heart of all art: I don't think the pipeline will ever be replaced by one-shot, completely automated and brainless txt2img processes. Even if the difference between "beginner artist" and "expert artist" shrinks to a hair's breadth because of access to AI, it goes back to that xkcd quote, "Our brains have just one scale, and we resize our experiences to fit." We will continue to instinctually desire and respect the efforts of people who go the extra mile to create something just a little more perfect or a little more original.
Do you happen to have a link? This sounds amazing. I wish more artists and mainstream would come forward to help immigrants. The way the states treat the immigrants is inhumane. We should open our borders, not build more walls.
Indeed. There was a time in the past ages in which one had to wait for a certain dye to arrive in your local merchant, take it and mix it with another dye in order to be able to paint a color that you need painted in your painting. They also waited and scoured for brushes and other supplies.
Those tools and supplies becoming available widespread and easy did not reduce the quality of art. It made it easier to do art. Artists gained more time to express what they exactly had in mind rather than going through the mundanities of having to deal with tools and supplies.
Computers have done the same. A lot of the art that can be done through computers these days, including the much-underappreciated art of creating virtual worlds and tales in computer games, would be dreams for the artists of earlier ages.
AI won't be any different.
The artists now will be able to say "Put a distant view of Shangri-La, half visible behind a misty mountain during sunset, as the traveler gazes at it and wonders when he will finally arrive there". And they will have it.
Now, from this point on it gets interesting: Notice that how it was extremely easy for the artist to get a scene that conveys a particular emotion created. What was difficult now, and what was the final work of art in the earlier times, is now just a simple statement. Everything got much simpler.
Meaning that, the artist will be able to convey more complex emotions and situations, creating more refined art.
That's what happens when we give more tools to people and make a given level of doing anything simpler: People start building more advanced stuff, and the complexity moves to the newly emerging level. Like in software. Will be the same in art.
Not so fast. To me this is the part that people are glossing over a bit too quickly in those discussions.
How does the person making this know it's actually good art? How do they pick between the 40,000,000 variations of that composition they're capable of generating?
Wouldn't they have to be familiar with principles of composition, color, dynamics, and so on?
As it becomes easier to generate the art, it becomes more important to be capable of differentiating it (except for purely "industrial" production where "good enough" will do).
And people will need to refine their taste to extreme levels to get there.
It's the same with poetry. What's the AI worth if it's supposed to generate Shakespeare-level writing, but no one is capable of assessing its quality?
Does it mean art essentially becomes a purely curation, taste-based endeavor?
Just throwing some random questions out there.
How did Van Gogh know he was making good art when he cut off his ear...
> Wouldn't they have to be familiar with principles of composition, color, dynamics, and so on?
It really doesn't feel like art is something that can be formulated and constrained as such. And it never was. All of those are techniques to effect certain results to convey the emotion or the idea. If the computer already does the technique part, what's left to the artist is to imagine, feel and express.
> As it becomes easier to generate the art, it becomes more important to be capable of differentiating it (except for purely "industrial" production where "good enough" will do).
And you can be sure that people will differentiate themselves. Like how they did in every age before. Moreover, those who had a talent for art, but not a talent for all the techniques and the tricks that goes into making that art happen, will now start making art.
> What's the AI worth if it's supposed to generate Shakespeare-level writing, but no one is capable of assessing its quality?
If an AI is generating Shakespeare-level writing, and no one is capable of assessing its quality, then that's Shakespeare-level writing and its good.
Let's face it: Most of the 'high quality' criteria comes from our beautification and exaggeration of the arts of the past. There isn't any objective formula that is used to assess the 'quality' of any art.
If art creates thoughts and emotions in you, its good art.
I wouldn't know, but I'm not sure what that point illustrates?
I'm fairly sure Van Gogh had a pretty good idea how his art was different/stood out from what existed, and could articulate why he did it a certain why (i.e. why it was good, to him at least). I'm far less certain about some random person with no art background typing a prompt being capable of the same analysis/understanding (that's also why 99% of AI art so far is transparently derivative).
To be clear — my point isn't that AI-art is bad or whatever. It's that it could displace/change the nature of what making art is about, and that incidentally it might require new skills or more extreme versions of existing skills, making the fear of the disappearance of "artists" overblown (and as a corollary making the idea that anyone can become capable of creating meaningful art exaggerated as well).
> It really doesn't feel like art is something that can be formulated and constrained as such. And it never was. All of those are techniques to effect certain results to convey the emotion or the idea. If the computer already does the technique part, what's left to the artist is to imagine, feel and express.
Right, but "expression" isn't something trivial, that's the point. If the prompt is "Harbor scene", there's a billion ways to realize that. Someone would presumably have to have the vocabulary/knowledge to navigate that space and produce something worthwhile in the end.
There's probably something to be said about the amount of control the artist gets on the output as well. The more we try to reintroduce post-hoc adjustments to e.g. an AI-generated image, the more we make "new art" look like "old art" (i.e. requires significant specific technical knowledge, albeit with different tools that paint and brushes).
Maybe it hits a balance, maybe it just goes back to the same situation as before.
> And you can be sure that people will differentiate themselves. Like how they did in every age before. Moreover, those who had a talent for art, but not a talent for all the techniques and the tricks that goes into making that art happen, will now start making art.
Yup, that's another important point. AI just becomes the new normal. But it will take skills to stand out given that new tool set. AI generated "art" will be seen the same way we see children doodles.
> If an AI is generating Shakespeare-level writing, and no one is capable of assessing its quality, then that's Shakespeare-level writing and its good.
Eh, disagree. If people had no knowledge of what poetry existed in the past or no understanding of it (think average college freshman essay), it certainly wouldn't follow that AI-generated crappy nursery rhymes would become Shakespeare-level poetry.
> Let's face it: Most of the 'high quality' criteria comes from our beautification and exaggeration of the arts of the past. There isn't any objective formula that is used to assess the 'quality' of any art.
That's why I made the point about curation.
If there's no "objective" quality to art, it's entirely narrative-based, whether on a past canon, or some sort of manifesto/principles something should stick to (think of modern art movements).
The job becomes completely different — it's not about the technical details of the work itself, but about how it's inserted in a larger cultural phenomenon ascribing certain values to art itself or certain principles in particular.
> If art creates thoughts and emotions in you, its good art.
To the person creating it, maybe, but then we're definitely talking about something different than ("high") art the way it's been talked about in academic discourse for the past ~2500 y. There has so be at least a common perception about the value of a piece of art for it to be considered as such.
I'd say the opposite. Seeing how shit SD was at generating placeholder products for a demo, and how it fails at stuff like proportions, depth, etc. the only good applications I've seen so far is abstract/concept art where you suspend disbelief because of the setting.
I think abstract art is lower hanging fruit. "High art" will be impacted as long as someone spins it into a story, it mostly reduces into social circles and selling bullshit to eachother anyway.
To do a practical exercise, I just enter to The Atlantic website, which in my memory has some level of quality on custom illustrations and photos/collages to represent a concept of the stories. The Atlantic now is paying designers/illustrators and stock image websites to create this images. After the research, I believe that using AI will be useful in many cases and will reduce costs for sure. What do you think?
------
AI + Human "AI querier"
AI is able to do this just from a text query and maybe a human tuning the query
* https://www.theatlantic.com/technology/archive/2022/09/crypt...
* https://www.theatlantic.com/ideas/archive/2022/09/brad-delon...
* https://www.theatlantic.com/ideas/archive/2022/09/free-speec...
* https://www.theatlantic.com/family/archive/2022/09/icebreake...
-------
AI + Human "AI querier" + Human designer/artist for visual adjustments (merge/remove stuff/retouch)
* https://www.theatlantic.com/newsletters/archive/2022/09/care...
* https://www.theatlantic.com/family/archive/2022/09/developin...
--------
Not sure
* https://www.theatlantic.com/ideas/archive/2022/09/liberal-de...
* https://www.theatlantic.com/family/archive/2022/09/developin...
A note here, the term "modern art" is usually used to define a period of art produced sometime between 1860-1960. When discussing artwork created today (or since about 1950), most would define that as "contemporary art".
Why is SD going to destroy the livelihoods of artists when machine language translation hasn't put human translators out of work yet?
I don't think there's been any industry that's been ended by AI yet, and yet people are strangely confident that art is going to be the first.
Because that's not true at all. The AI can't even draw hands yet. To say nothing of its ability to handle multiple people and objects interacting in complex scenes.
I'm concerned that the discussions about AI art on forums like HN get distorted because you have people sharing their views on art here, even though they don't actually have a serious and nuanced appreciation of art and they don't have a good understanding of all the types of work that artists do. Maybe you'd be fine with reading a comic book where everyone has seven melting fingers, but people who take comic books seriously as an artistic medium would not.
Technology is making something that used to take a lot of practice and skill be accesible to those without any of it. A monkey can now draw two ovals, label it an owl, and run an image-to-image conversion with Stable Diffusion to get a pretty good sketch of an owl [1].
Is it better than what a good artist could do? Irrelevant.
Is it better than what a cheap illustrator I find on Fiverr could do? Irrelevant.
The only important point is that I no longer need an illustrator to get myself an owl. I draw some lines, I pick some words, and presto I have an illustration.
The question of whether it's "art" is entirely irrelevant.
> Are you under the impression that right now, as of today, the publicly-available AI models are ready to replace humans for all types of art outside of scientific and technical illustration? Because that's not true at all. The AI can't even draw hands yet. To say nothing of its ability to handle multiple people and objects interacting in complex scenes.
I think this is severely underplaying the speed at which things are changing and basing an argument about things that the AI currently can't do. DALL-E was anounce in Jan 2021 and it's still locked behind API access. Stable Diffusion came out Aug 2022 and I can run it on <$2,000 laptop. That's not 2 years. Do you think hands are going to be a long term roadblock?
As for complex scenes, you can currently string that together with a Stable Diffusion plugin for photoshop/gimp.
[1] https://www.reddit.com/r/StableDiffusion/comments/wwv7zk/sta...
Now, this may actually be helpful in that it gets around copyright claims - but that's the only real difference.
But the approach for stable diffusion is just as easy whether you want just "an owl", or "an owl in X's style with A, B, and C"
Now, I should of course note that search engines already employ ML techniques to actually interpret search terms, so to some extent the point is moot - ML is important to actually solving this problem.
And of course, it's also possible that the image I want can't be generated by SD/DALL-E/etc.
Without meaning to sounding rude about it... I'll wait.
I tried generating that exact prompt a few times at theartbutton.ai and all the results were nonsensical.
For example: https://theartbutton.ai/image/OW1HZLfhjg6DFvJtk4vQZzUYqI7pGG...
Not a very wide range of what I could do with the idea in terms of composition, but just some variations of finishing touches/intermediate steps. I achieved this with some human-in-the-loop iteration and inpainting, but it was no more than 15-30 minutes toying around with it, and I'm no artist.
If you have a semi-decent graphics card and would like to experiment with a bunch of extra settings and tools than are readily available online, this is a good repo for that: https://github.com/AUTOMATIC1111/stable-diffusion-webui
This seems to be purely an issue of the size of the network. Parti (https://parti.research.google/) demonstrates that as the number of parameters increases, with no change to the underlying architecture, a lot of these problems simply go away. Basically just throw more compute and memory at the problem and everything gets fixed.
The 20B images don’t look that much more impressive than what SD is already doing (aside from the ability to render text), and in some cases they look worse. It’s hard to tell because the resolution is so small, but even in the 20B “astronaut riding a horse through a pond” image, it looks like his hands are still nonsensical.
AFAIUI, that’s part of the point that Gebru was trying to make before she was fired.
Edit: to tie it back to the original criticism: what’s the maximum training cost we’re willing to accept for the model? How can we guarantee return greater than the increased training cost?
And training a real living breathing person in a rich OECD country is going to be costly - no offense meant, I'm actually not from OECD.
But then again, when it comes to commercial needs, a human doesn't need "retraining" every time you ask them to draw something they weren't familiar with when they went through art school...
With an order of magnitude more parameters it won't just do hands, it will do quite a bit more.
>the tech scales so well as to make their new goalpost irrelevant in a year.
This just brings me back to my original question. Self-driving cars have been "a year away" for many years now, and now companies are starting to hint that human assistance may be required for the foreseeable future [1]. So, why the confidence that art will be an easy problem to solve with just more scaling, when that approach hasn't eliminated the need for humans in any other domain?
[1]https://www.reuters.com/technology/truly-autonomous-cars-may...
(This will be great from the point of view of art creation; not so great from the point of view of supposedly rendering humans obsolete)
In the smartphone age, the case for data hunger looks pretty weak.
Artistic styles are often just thin semantic filters over the base 3d geometry that can be learned from photos, and learning these shouldn't require many examples.
This is not to take away from the achievements of AI. It's that creating pictures adhering to a prompt with some degree of creativity is very little of what art is. Maybe it will replace some part of commissioned illustrations where the artist's name does not matter (e.g. some avatar pic?).
We still value, financially, some material goods for much more than they cost to produce. Or for much more than their almost identical mass-produced counterparts.
I mean, it's the majority of commercial art - you get a prompt from the client, you maybe flesh it out in a few different directions with sketches, then you refine a final piece. And AI is incredibly good at this process - instant results, infinite patience, and it's free. A very hard combination to meet.
Calendars, book covers, video game assets, green screen backgrounds....
Even in a case like video game animations, where the AI can't build every frame, it can still give you a good reference photo. From there you just need a cheap artist to fill out the frames - a huge cost savings, and a big blow to the artistic community.
Where do you get started as an artist, without any of those? Obviously, Fine Arts isn't nearly as effected, but how do you get your start when you can't build a name from your cool book covers, or get famous off Magic: The Gathering card illustrations?
I would extend this lack of nuance and understanding to the deep learning implementation side also. A lot of people seem to have some very foundational misconceptions about what deep learning is and what it does. In the case of generative art: these models are “simply” sampling from the frozen statistical structure they have learned from web images and their captions. They don’t understand the relationship between objects in space, they have no ideas or feelings to express, and they communicate nothing. That’s why the even the best output of these models tends to have a perceptible hollowness: you can detect the lack of a coherent authorial intent in it.
The field of "art that needs human communication skills" seems to be a lot broader than just scientific illustration.
SD obviously doesn't understand language in the same way we do, so it can be tricky to describe things in a way that will match your expectations. Once you start to understand the tricks here, it gets easier and easier.
Inpainting will let you fix a lot of the rest. Staircase stops? Select the area where it stopped, get the AI to generate more. People are already doing this to create very complex artwork where there are issues with faces, hands, etc. https://www.reddit.com/r/StableDiffusion/comments/x9u8qh/img... is a great example of how you can quickly iterate over a scene.
One of the other things people struggle with is consistent characters and settings, but people have found ways to improve this with Midjourney - https://docs.google.com/document/u/1/d/e/2PACX-1vRahIr3-h_V3...
There's more of a learning curve to these tools than most people think, but it's also still miles and miles away from the learning curve required to actually be proficient at the technical aspects of making art.
Yes and with luck, at least one language you know how to read will make sense and not be off-the-wall bonkers or simply incomprehensible.
It's not clear if machine language translation may have:
- Increased "demand" for translation by reducing the price of "basic" translation. - Increased overall globalization -- I use machine translation to communicate with contract manufacturers in China, whereas without it I might avoid using contract manufacturers in China. - Due to increased globalization, increased demand for occasional "advanced" translation via professional translators or bilingual.
So perhaps it does greatly decrease or eliminate the large slice of the pie which reflects translation jobs that would exist at the low end while simultaneously greatly increasing the overall size of the pie.
> I don't think there's been any industry that's been ended by AI yet, and yet people are strangely confident that art is going to be the first.
Any time now.
It can recreate data close to what it has already seen, just like all neural network techniques. It does poorly outside that domain.
Of course, there are still some jobs for high-quality/high-importance translation like legal work, simultaneous translation etc, but these are quite niche.
And yet, with «stable diffusion» it is just so. Disruption is declared wholesale, without even passing mention of the preceding state of the art of computer generated art.
Not to put too fine a point on it, but do you not see the own goal here?
Gp then proceeds to declare «Stable Diffusion» to be Capital D disruptive, bemoans the unfortunate destiny of the world’s now disrupted artists but, curiously, does in fact not argue the point.
I'll believe it when I see it. How many self sufficient, independent digital artists are actually making a living doing something other than webcomics or furryporn? SD is amazing technology for sure, but this notion that it's going to incredibly disruptive is the latest in a long line of AI hype. It ignores the simple fact that there is already more art produced and uploaded onto the internet, every hour, than can meaningfully be consumed. Independent artists already had to compete with millions of art uploaded very minute, what is a few million more?
I'll second that.
Hell. Even ATMs haven't automated bank tellers yet. Nor has Kindle replaced physical books.
I really don't see it. What tools like SD, DALL-E etc do is essentially similar to what Google Image Search does, except that the results are presented differently - instead of showing you the results it has indexed in some order, it picks features from the results it has trained on with some priority, and presents a kind of amalgam view of the entire collection at once.
The only significant advantage compared to Google Image Search is that AI-generated art is getting around copyright, allowing you to publish it legally under your own name. This makes it very similar to Uber, which got around taxi regulations that were impeding market access.
Essentially, if we ignore copyright, there are relatively few illustration problem you could have where DALL-E/SD/... would help more than searching for an existing image. Branding is an obvious one, where the need for unique-ness is not only legal, but also practical. The other is that DALL-E/SD/etc can also sometimes produce combinations of unrelated images, thought that is offset by just how bad their results are in other places where there is plenty of existing art to choose from (especially anything involving images of humans).
Not like Taylor Swift. As Taylor Swift.
We'll have our new product out very very shortly.
We're at end times.
If you're not excited by these rapidly improving results, then I don't know what to tell you. I think you can extrapolate what lies just ahead if you think about it.
I've been a tech pessimist for the last two decades of my life, and I can strangely now state that this is the most important and exciting time of my entire life and career. Everything we know is about to change.
I can't believe that this is the timeline we're in and that it's actively unfolding now. It's like a dream. It's hard to grasp and come to terms with.
Just think for a second about what these models are doing. This is going to happen to every product category.
Think too about Carmack's recent claims.
ML and Rust +++
Even if I like Taylor Swift, what would be the point?
Tomorrow, you want more Taylor Swift and you tell the computer to give it to you. It delivers with no human intervention whatsoever. You love it, but want it a bit more country - so be it. Piano and acoustic. Got it. A duet between her and Bruno Mars. Just for you.
Michael Jackson and Elvis? Nothing is without possibility.
Brand new singers. Instruments and styles we could never conceive of - all about to happen.
A never-ending stream of novel music that tailors itself to you.
Already being dreamed up.
I'm building this, have parts of it already working, and I'm sure others are as well.
Every inch of your day to day life is about to change in a way that will make developments brought on by smartphones and the Internet pale in comparison.
> A never-ending stream of novel music that tailors itself to you.
That's worthless. A firehose of novelty is almost immediately no longer novel.
Why would anyone want this?
It's the same problem already with streaming services and their own series. They're all already joyless and immediately forgettable. They have less and less value.
Your breathlessly described "amazing transformation" sounds incredibly boring and pointless.
It's easy to say that as someone on the outside without skin in the game, not working on this, not in the know.
> That's worthless. A firehose of novelty is almost immediately no longer novel.
And yet here you are on HN. I'll bet you have a Twitter and phone and consume social media, too. That's all just a poor man's version of what we'll be building.
> Why would anyone want this?
I have literally wanted this exact thing all of my life.
There have only been a handful of films I've walked out of and felt truly satisfied. I've seen thousands of movies shot by mere hundreds of directors at best. That's billions of dollars of institutional capital put behind a few lucky white men for what amounts to as a "meh" at best. Now we can empower anyone to have their own take at it. To build and express anything. To sift and sort through emotional/concept space and deliver truly new things.
We get one short life to experience the universe. Let's climb the gradient to the stars and experience the limits.
> It's the same problem already with streaming services and their own series. They're all already joyless and immediately forgettable. They have less and less value.
Spaghetti thrown at walls by blank check writing firms trying to capture eyeballs. This isn't at all what I'm talking about. This technology will bankrupt that type of behavior.
Everyone will have access to these tools and be able to build whatever their heart desires. Whatever they want to put out into the world for the rest of us to experience.
> Your breathlessly described "amazing transformation" sounds incredibly boring and pointless.
I don't get you, but that's fine. We'll be building away regardless.
I don't think humans are the end point of our tiny quadrant of the universe. We're a transitional blip. This is just part of the changing of the guard. And the transformation will yield beautifully captivating results. I'm here to contribute and experience.
A better version of a net negative is still negative.
Maybe I'm wrong, but it really sounds like more of what we've already proven isn't good.
> It's easy to say that as someone on the outside without skin in the game, not working on this, not in the know.
But you sound like you're in a cult.
This is probably true if you want a single image, but if you want many images that look consistent with each other then google images will be little help. Stable diffusion with textual inversion can just about do it though. I would imagine in a few years it will be able to do that better than it can currently generate a single image.
Professional illustrators will always have work, but their work may be in a different shape than current one, relying on AI as an additional tool to generate images for their clients. In any case, a big part of the work for illustrators do is in refining the vision of their client and exploring the space of art this way. This is as much of an issue with the current AIs, you need to ask for something very specific to get a good looking image. Iterating to produce a quality image can take hours.
As for art more broadly, it is much greater than visual art. It will be interesting to see the new ways people find to express themselves with these new tools.
> Like the other commenters have written here, art is a lot more fluid than you give it credit for.
And the new workflows will be as well.
> Professional illustrators will always have work
They're in work because the job takes 10,000 hours to learn and individual pieces take 2-10 hours to create. Not anymore. The dam got busted.
> In any case, a big part of the work for illustrators do is in refining the vision of their client and exploring the space of art this way.
You tell the illustrator, "I don't like the arm". They come back the next day with something new.
You tell FancyIllustrator.AI or whatever the same and you can draw a squiggly and have ten new concepts to refine from. Game over.
> This is as much of an issue with the current AIs, you need to ask for something very specific to get a good looking image. Iterating to produce a quality image can take hours.
This is month one. Everything is about to be disrupted. Even the big tech companies.
This is the biggest singular moment of Innovator's Dilemma in human history.
What was practitioners in the field of computer generated art able to do in month minus one, how has this changed subtstantively, and do they offer any experience, criticism or understanding of the practice that might still apply? If not, which and why?
I’m as bedazzled by the blinkenlichten of prompts and generated images as the next guy, but I must confess I don’t have the foggiest what those guys were (and probably still are) doing before Dall-E. If the needle really is bending against the post I would take their word for it.
https://www.youtube.com/watch?v=u7j4f6U7FJ4
I think that machine learning models will quickly hit some walls that feel unintuitive due to a lack of training data, but I also wouldn't want to bet on which ones.
Digital technology has already accelerated this process a lot, allowing artists to communicate prototypes efficiently and supervise implementations of their work on the other side of the world. What's new is how AI makes this dirt cheap and accessible, not just folks who can earn the admiration of assistants. People are focused on the plight of the artist-creator, but I also wonder what will happen to the legions of substantiators already out there in the world.
>Amazon and Walmart by contrast did disrupt entire industries. Independent book shops and mom&pop retailers saw the writing on the wall. They knew they couldn't survive facing this kind of competition, and largely, they didn't.
Sorry, I’m still not seeing the distinction you’re making between these cases. You say that with Uber, it makes no difference because there are still drivers going from “point A to B” (a pickup to a destination), but it was a huge difference with Walmart because small shops “saw the writing on the wall”.
But … in both cases you have the same service still being provided (rides and retailing) and the unprofitability of smaller providers (small shops vs taxicab companies).
If there is a difference between the two, could you highlight it clearly and in plain language?
Artists are losing today, but why are they in any way more vulnerable? They just seem to be next on the chopping block.
"There will be backlash"
For 99.999% of art that is commercially relevant no one cares if it's "real" or not.
The problem here isn't necessarily that these tools change the story for "real" art, it's that they change the story for _commercially relevant_ art which has always tended towards uninspired and formulaic.
There are a lot of artists that have hobbies in art they consider meaningful while making a living with commercial art that can be software-displaced.
If something affects a large group of individuals, it will eventually have an impact on society as well.
A force multiplier, allowing existing professionals to generate more art.
The creation of a new profession. Digital artists are already doing a very different job with very different tools. These tools will be another step and there will be demand for human and generated art in different applications.
The empowerment of non artists. Same as how digital art democratized art. This will let people without hours of skill generate images of their liking with relatively minimal effort. Anyone can generate scenarios or images, even if they aren't perfect.
These aren't exclusive and will probably all happen at the same time. It will also 100 percent devalue artists work since it's now easier to meet people's needs. This won't make them obsolete, but it'll reduce demand and put people out of business.
- Instead of video advertising with a handful of regional variations, hundreds or thousands of variants tweaked for specific preferences, target profiles, as part of A/B testing, etc. Changes to accent, emphasis, copy, maybe even which parts of the product being advertised are focused on.
- Increasing variation in open world game locations and interactions, with AI able to insert much more meaningful distinction between instances of a template (as compared with recent Assassins Creed games, where identical towns and forts are copy-pasted dozens of times all over the map).
- Much greater accessibility of visual art as an accent for other kinds of creators— illustrations to accompany poetry, blog posts, fanfic, etc.
- Increasing ability for AI to do first-cut assembly of video content, particularly review-type YouTubers (think: SkillUp, Critical Drinker) where the video is kind of secondary and the majority of the audience may even be listening to it as a podcast anyway. The AI being able to understand from the script what is under discussion and select matching clips would likely match or approach what
This is far from a sure thing— sometimes people really are the buggy-whip manufacturers in a situation and technology completely displaces them. But other times (as with software development), the market and use-cases have expanded considerably faster than the efficiency gains made by better tooling.
Not saying this is bad, or that typesetters or illustrators deserve a living. I am saying that being able to draw used to require being able to draw, and that instead being a fan of drawings, being good at describing imaginary ones to a computer and iterating them into a finished product is a mostly, if not entirely, different skill.
Right now naively generated AI images are having their day, but I expect we'll soon get over that. I think the real future is in visual artists who closely collaborate with future generations of ML systems that are much more tractable and responsive. I expect a lot of that iteration will involve pretty traditional art skills, because those are time-tested ways to convey visual information.
So basically, you read over a book(with amazing art) over, know your terms(or maybe know how to use a neat visual dictionary), and now you can paint in many art styles.
Of course there will be some more learning while playing with stable diffusion and getting a feel how to talk to it.
But it's very far from the old skill of neing an artist, not much knowledge will transfer.
I can immediately recognize (as I did for this post) when art is AI generated right now. I've spent a lot of time playing with SD and while it's been a blast as a toy, the vast majority of the outputs are 'meh' and even the best still have those weird artifacts that stick out like a sore thumb.
Just like the clip art ascetic quickly becomes recognizable and somewhat nauseating, so too is AI generated art already starting to feel that way.
As far as aiding experts I don't imagine synthetic art will be nearly as game changing as the Adobe suite. I do a fair bit of writing and find GPT-3 to be completely unnecessary for "quickly generating a bunch of concepts". I assume that visual arts, like writers, already have a bunch of ideas in their head for what to do. That's never the bottle neck. Writers block happens but it's not because you can't think of a general concept, it's because implementing that specific concept into text is what's hard.
The output between all the major AI image synthesis tools, while impressive, is quite clearly very high dimensional interpolation, and it shares the same limitations in that space as it does the 2d space. We've already thrown insane amounts of data and compute resources at the problem and it's made the things that were done well in the past even better but doesn't move us towards solving the hard parts.
Particularly as one deals with the world of atoms, where the cost of a mistake is much higher.
There are already models that work to “aesthetically” rate images, and the results are quite promising.
To think we won’t see leaps in quality over the coming months/years is narrow sighted.
Though it's more likely in a the coming months/years people's ability ignore reality will just become stronger and we'll all believe that a person waving their 4 wiggly fingers and staring at you with strange fish eyes is hyper realism at it's finest.
We've entered the "emperor's new clothes" reality at this point, so nothing would surprise me.
But they haven't plateaued. I can go take a trip in a self-driving Cruise in San Francisco right now. I couldn't do that last year.
There's certainly a lot of room for them to improve, but the self-driving car industry is making progress.
Not happening.
Inevitably they started on the lower rungs and worked their way up.
If these tools displace the need for lower level/apprenticeship like work, then it won’t be a surprise if later on we have less quality artists who weren’t able to get the necessary experience early on.
I think of it more like, If an artist can support themselves doing commissions of low quality art so they can support their high quality art, that’s better than if they had to drive Uber.
https://mobile.twitter.com/xsteenbrugge/status/1558508866463...
1) the lower rate per piece enabled by higher productivity unlocks a much larger market for commissioned work. No one has to lose their job, people make up the price difference in volume.
2) there is not a much larger market waiting for cheaper art. The existing pool of artists is too big, and some of them need to find new work.
My guess is lowering the price does grow the market some, but not enough to avoid displacing at least some of the artists. Whether that’s 10% or 99% depends on how good these tools ultimately end up being.
If you want something a diffusion model can produce, you can get it for nearly free. If you want something significantly better, you have to pay an artist significant money for it.
1) Quick, low quality sketch
2) Generate detailed scene based on that sketch
3) Fix obvious errors by re-generating specific parts of the image
I also have questions, what is the "Generate detailed scene based on that sketch" process? Can an algorithm do this? Can a person easily do this? I'd be more impressed than with the raw pictures. Similarly with your #3.
[1]: https://www.reddit.com/r/ProgrammerHumor/comments/u8j3lf/i_n...
You can't ask for "in the style of Keith Haring' before there's a Keith Haring
Since we still have to give text prompts to generate images, and these prompts need qualifiers (x subject “oil painting”) it seems possible that a prompt giver could generate images in a specific way to achieve their own unique style.
Difficult now since the tools are new and don’t have a ton of features, but down the road if the tools are built out more and prompts get more nuanced I could see “prompt givers” as “artists with their own unique style”
- Before writing, ideas had to be distributed orally.
- Before the printing press, distributed copies had to be written by hand.
- Before the internet, distribution required thousands of physical copies to be produced.
That's not to say that we've ignored creation. We've built hundreds of tools to make producing content easier: word processors, Photoshop, and even TikTok to name a few. It's just that AI has recently taken a huge leap in this area. For that reason, I don't think "unbundling" is the right metaphor here.
Otherwise- really good points, well-written
In the unbundling story, the premise is I gather that what is disruptive atm is that what's left as the god-in-the-gaps human role, is now democratized in the the technological acceleration in the the creation-substantiation phase at the head of the snake.
What we need to start preparing for is that the creation phase is next; it's a mere engineering problem.
Yes, that's somewhat sardonic; it's going to take some years. But only some.
What I am saying is, before we are ready, the aspects which remain human-domain in the cybernetic collaboration between humans and AI in what well call "content generation," namely, the agency and the intention—the "idea" which today is, the prompt; the editorial and curatorial functions; the insight and lightbulb and all...
...that is going to become the domain of the AI as well.
What inspires "art" or whatever else?
The filtering of media streams (aka cultural discourse, the zeitgeist) and the identification of symmetries, analogies, serendipity, humor... whatever that is, it like the "substantiation" phase is going to be ceded to the super-human in a few years.
I speculated recently to a friend that the way the fine-art market reacts to this is going to be interesting. We're going to have Wintermute on tap. I wondered if gallerists will represent AIs, or, spin them up long enough for them to emit some stream of work then be ceremoniously destroyed.
Per Ex Machina that may not go so well in various ways.
Anyway. The writing is on the wall.
And it's not in OUR handwriting.
You can say the same thing about food today versus food 100-200 years ago. But we've seen how that plays out, with western countries and the US in particular facing massive public health problems and an obesity epidemic. Humans evolved for scarcity, not abundance. We just can't handle having so many cheap, hyper-palatable foods around.
In the same way I think it's obvious that hyper-information is harming our mental health. And while I hate to think of losing access to all the wonderful information resources we have today, I also think the path we're on will be net worse off for most people, just as no sane person would ever wish for famine, yet junk food also makes most people's physical health net worse off.
In the same way that the entire "diet industry" thrives today, I expect to see a greater and greater rise in the amount of self-help books, training, and support to help people with resources cope with information overload. Meanwhile most people's mental health will be harmed.
If we want broad prosperity in the face of over-abundance, we're going to have to learn how to cope with abundance and filter the bad out, so that we get an abundance of good without drowning in junk. This is going to be the defining challenge of the 21st century.
Content will be created specifically for you.
The reason is that AI intelligence is probability generated mimicry. By extension it mimics art but is not art. Think of the character arcs in Shakespeare, the narratives, the layers of meanings. This sort of thing cannot even be modeled as a problem in AI. The ability to define what is necessary to produce such layers of meaning does not exist.
I think AI is missing something fundamental about intelligence. What it is missing is the ability to construct systems of abstraction and have them relate via meaning (which is informed by history, context, experience).
Something new must emerge for AI to move past probability generated mimicry.
One of the lessons from DALL-E and latent diffusion is that problems can often be decomposed into a “semantic” space and a “textural” space loosely corresponding to what we consider layers of abstraction.
There are good uses for human guided AI, but I fear most people are down for accepting whatever the algorithm happens to crap out at you
The News aren't just about distribution, you moron. If we didn't have large companies such as twitter and Facebook misclassified deliberately through government protection and instead called out as the publishers they are, newspapers would be one helluva profitable business line.
Just because economic forces were over ruled in favor of a national security advantage in keeping the American tech industry growing into the massive multinational beast it is, that does not automatically disqualify the hard journalistic effort in contributing to the large subset of content that drives engagement and thus clicks and thus ad dollars for many of these businesses.
Just because you post a link somewhere, it does not automatically waive the commercial interests of the platform you're posting to.
In another strain of the argument, you posting to Facebook and Facebook later choosing to rank and distribute that content to their liking with you having ZERO control over that is the equivalent of using individuals as proxies to steal and propagate content.
You can't just start a website with stolen content plastered all over, say its coming "From the users" and deny any culpability.
But what happens when AI becomes more creative than humans? There will really be no need for a human artist at that point.
I thought the AI's interpretation was much more interesting than what the author wanted.
AI makes the execution step near-instantaneous and almost free. It kind of erases it. It's more disintermediation than unbundling. Execution artists are being removed as gatekeepers.
A question remains that, if an AI generator produces deterministic, consistent output given a set of (prompt, seed, model), does that mean that all possible images are somehow contained in the model?
If there was a Borgesian book containing, say, all possible 512x512 images, one on each page, then surely two people having the same copy of that book wouldn't need to exchange images, they could simply exchange page numbers, and see exactly what the other one is referring to.
If we are now able to exchange prompts and seeds and get a predictable, consistent result out of SD, isn't that what we're doing?
If that were the case, the page number would be as long as the image, and exchanging one or the other is the same. Heck with proper order arrangement, the page number IS the image, in a known format.
In other words, the information content of such book is exactly 0.
So that does mean image model AI like dall-e/SD are effectively compression over this space of images that they can generate (which is at least attempting to emulate the space of 'meaningful images' to a human), since given a seed, they'll deterministically produce the same image, and that seed is much smaller than the information needed to describe every pixel in the image.
does that mean that all possible images are somehow contained in the model?
A perfectly linear model with linearly-independent matrix columns could generate any possible value, but would be exactly equivalent to a single vector-matrix multiplication, unable to do any interesting multi-step reasoning.
Plus, I can't imagine it ever becoming faster to generate a 512x512 image using AI than transmitting the full image over a network.
The brief openness of ai right now is a glitch in the system that will get ironed out soon. Only megacorps, and countries that ignore patent law, will be able to afford to license the patents to do any significant ai work at all
AI is really a black box, which should make patenting specific implementations of it very difficult. Even if you do manage to patent something like a diffusion model of image creation, in order to enforce it against someone who was willing to go to court, you'd have to get into a deeper discussion of exactly how your AI works than you may want to.
My guess is AI will be more the realm of trade secrets than patents - it'll be the Coke recipe of big tech.
In particular the large tech companies simply ignore all existing patents, develop what they want, and defend themselves in court if need be. Google is never going to say "well someone else got a key patent, let's give up development in the AI space".
I really wish I knew what DFW had in mind when he said irony wasn't the point.
He can't see the forrest for the trees:
"I have plenty of ideas, and thanks to the Internet [...]"
Most of those ideas will look shallow in the future and be generated by an AI, executed by an AI etc. aka the great bundling.
you can't have another van gogh by reproducing van gogh paintings.
reproducing van gogh painting will teach you a lot about art.
but once you become good, you develop your own flair / style.
and that's what makes art great, the infinite possibilities.
it's the same reason, we tend to overlook countless paintings of trees done by street artists unless the style is so unique and evokes our emotions by beauty or novelty.
for that reason alone - I will go against the grain and say AI art won't push artists out of jobs