Related: "Invasive Diffusion: How one unwilling illustrator found herself turned into an AI model" [0].
It's astonishingly simple to train both LoRAs and Embeddings nowadays -- if you combine it with tech like ControlNet, you can more or less have direct control over the illustration as well (see this example I converted to Hollie Megert's style with barely any effort [1] using the model from [0]) -- but FWIW I was playing with the WhatsApp AI stickers recently and human-made illustrations still have a "je ne sais quoi" that sets them apart, vaguely reminiscent of "realistic" animation vs. "squash and stretch". I would not be worried as an artist*, but I would be worried if I am making assets for corporations in Corporate Memphis [2].
* In fact, I think I'd be excited, it can be a great tool, and people who lack artistic sensibility might find it almost as hard to use as to make art themselves. However, I do think there needs to be some kind of protection from having your 'style' commoditized without consent, using unauthorized images of your own art.
Corporate Memphis seems to have that very "flat almost clip-art" style to it.
Generative art is ideal for this type of stock art use case where 90% is simply good enough for the vast majority of situations.
I feel like the gig economy (things like Fiverr) are unfortunately going to get absolutely crushed by generative art systems since most of their customers aren't that particular and are likely fine with anything that DALL-E / SD can pump out.
> I feel like the gig economy (things like Fiverr)
Predictably this is already happening, it doesn't take much browsing in categories such as tattoo design to spot the hallmarks of AI generated artwork. To add insult to injury they often pretend they aren't using AI, so they can up-charge for more intricate artwork that would take more effort if they weren't really just adding more synonyms for "detailed" to their Midjourney prompt.
Fiverr does have a dedicated category for AI art, but they don't seem to be making any effort whatsoever to keep AI art out of every other category, so it's pointless.
Soon enough generative art will be everywhere and the style uniformized so when some brands will want to differentiate themselves they will hire people again. They could charge more for the service, whoever survives the generative AI squeeze…
Can you tell when art is AI generated? I already can and that is what I am talking about. In time this will become more and more clear to the the market and in cases where it matters they will hire a human.
> it doesn't take much browsing in categories such as tattoo design to spot the hallmarks of AI generated artwork
Which is kind of interesting. My first tattoo has its roots from it. Where I used ChatGPT together with DALL-e to express the concept I wanted on my body, and not a direct translation of it.
Worked out pretty well, got a lot easier to discuss how and what things would work and how to apply them.
Isn’t it their job to put it on you? My understanding is many people bring in designs. That’s the easy part that is so replaceable, whereas needling someone’s skin is not strictly ai replaceable quite yet.
It depends on the person buying the work I guess, but most tattoo artists that I know wouldn't be overly happy with being given an image and asked to just tattoo it.
Has GenAI _really_ been good enough in stock art use cases, or are they given temporal leniency? I feel they're mentioned too often for that to be the case.
This comment thread could be from 3 years ago. What you're predicting I think has for the most part already happened. Everyone I know who used to commission Fiverr stuff has switched to AI.
> However, I do think there needs to be some kind of protection from having your 'style' commoditized without consent, using unauthorized images of your own art.
1) Hire artist to copy the style for 10 images
2) Train model on those
problem solved.
If you choose "using your images specifically" to steal your style as the hill you want to die on, then you're going to find it's a very useless line of defense.
Just because you hired artists to copy someone else's work doesn't mean your model isn't the result of copyright infringement, indeed it makes it somewhat more obvious imo.
Maybe that would be Fair Use, but things like purpose, weak transformation (in the hired artists copies), competition against the original, and business motive would all suggest it relies on non-Fair Use copying.
I don't know why you think something that involves direct copying shouldn't be a 'hill to die on' for a copyright infringement claim?
Not copy the work. Copy the style. Because style is not copyrightable.
What I'm saying is that if you believe that your personal style is your valuable asset and choose to protect it by saying people can't train on your images; you ought to be aware that it's trivially easy to build a model that replicates your style and does not require your images.
Style is largely in the eye of the beholder; SD1.5 was used to make a lot of images that looked like Greg Rutkowski, and people presumably believed they looked like his style, but he wasn't in the training set and it was a coincidence.
Of course this is mostly only possible if you limit yourself to looking at digital art. A JPEG of a gigantic abstract art oil painting is pretty different from the real thing.
That's common enough that there's a name for it [0]. "Stealing the style" is not so much a problem as commoditization, which may find some measure of defense in terms of trademark (not copyright). In this case, for example, the model is named hollieMengert_v1, and we are referring to the style as "Hollie Mengert's art style".
There is absolutely not defense unless "hollie mengert's style" is trademarked; and even then the protection extends only to that name, not the artistic style itself.
Clean room design has nothing to do with this. It is perfectly acceptable to overtly copy a style.
That last part is the issue. Once the material is released publicly there is zero way to prevent it from being used for training. I like the idea of these things but they're predicated on theft and exploitation.
They don't work, and nothing in the category can ever truly work.
Models and adversarial data are equally powerful - you can find an adversarial example for any current model, but you can also train a model that can handle any existing adversarial data.
Any image is a good example of /something/ - at worst it's only an example of itself and irrelevant for any other class. Don't know if there's a name for this principle but it's kind of like a Church-Turing thesis for data.
Not really, because Nightshade should have made image labeling more difficult, but if you try it, you'll see that it doesn't do anything; multimodal models are too powerful nowadays to be fooled by small adversarial noise generated using CLIP LPIPS (small enough not to be too noticeable to us).
And Glaze does not try to interfere with labeling.
> I do think there needs to be some kind of protection from having your 'style' commoditized without consent, using unauthorized images of your own art
I worry any attempt at that will strangle all creative endevours. Copying others style is not a new thing. It is the stuff culture is made of.
Can we just talk about the fact that the most popular fine-tuned model for SDXL is called PonyDiffusion?
The circumstances of its creation, and the whole entire world of like, booru websites that you have to explain to boomers in suits with VP titles who are trying to invest in this space is just surreal.
It's like a reverse of when they were raising their kids and giving us "the talk" as children. Now we have to give them "the talk" about what the word "doujinshi" means... for science!
I’ve been using generative AI in my art for a good few years now and have a few thoughts on this.
Yes it’s trivial to train a model on a particular style, I’ve done it many times, including on my own art. It can make things that look pretty good!
But it can’t make anything great, not without a lot of luck or creative input from the user.
It can make something that looks similar to a style, but it can’t explain what makes that style good, why it should be chosen over another and the thinking behind it. An artist would have no problem answering that.
As I understand it, generative AI is a sort of average machine, in that if you ask it for a picture of a dog, it’ll give you the most average result for the prompt, model and whatever randomness there is in the system.
Therefore, the lowest common denominators - stock photos, Corporate Memphis, clipart etc are in big trouble. AI images don’t need to be great to displace them, just good enough. Average.
However, for creatives, AI tools are a wonderful opportunity. It’s not everyday a brand-new medium comes along.
When photography arrived, it was possible to make a perfect representation of something at the click of a button. Instead of killing art, it unleashed a wave of creativity. Art no longer needed to be about drawing pictures of things or people and could get more abstract and experimental.
So with AI. It won’t kill creativity, it’s just another creative tool. I’ve had a lot of fun over the last few years exploring what it can do, and it’ll be exciting to see how art changes with its arrival.
As a follow on: One of my friends is a professional illustrator and was concerned about the impact of AI on their work.
As a demonstration I trained some models on their styles. They were worried right up until they saw the output. Yes, it looked superficially like their work, but the slightest glance showed it to be absolute crap. Anyone who would consider using the AI version of their work over commissioning the actual artist would absolutely get what they paid for.
However, they have found the models to be very useful in exploring concepts and compositions, becoming an essential part of their creative process.
It's possible that your model wasn't great, for my experience, in the past I uploaded an AI image in the same style of a certain artist to a website that organizes anime images, someone tagged it with the artist's tag because they thought it was made by him, also the image I uploaded got much more "likes" than anything made by the artist in months, and it's not like I'm spamming images, that was the only upload. (I later removed the tag, of course)
Part of the reason for that is those artists are doing commercial art, which means the "style" is imposed on them by the customer/series character designs/etc, and they're not putting their whole ability into it.
It's like how it was easier to automate office workers because they just followed steps in a process and didn't use creativity.
I mean, the artists you see on Twitter or other platforms usually have a dominant art style, their own, but even without changing the art style, you can still be really creative with the composition, actions, and so on.
Just call them what they are, "booru" websites. The AI community needs to actually understand what these are and why they just happen to give us extremely large, high quality, well taged datasets. Yes, the AI community will have to admit that coomers have been really good for AI development. Yes that will rustle feathers.
A lot of AI researchers are either playing dumb or are actually ignorant of this space. A lot of really talented folks are operating in the shadows and should be at NeurIPS and ACL presenting and bartering for VC funding rather than posting anime waifus on /g/ and staying anonymous on discord...
The model was good, it’s just that the illustrations needed to convey specific concepts and were created for particular purposes, something gen AI is not good at without significant guidance.
Superficial style is easy to replicate, the underlying meaning and thought process is not.
>It won’t kill creativity, it’s just another creative tool.
I've been banging on this drum since AI's inception, and will continue do so.
Even if machines were capable of directly reading our mind and outputting a perfect representation of exactly what we wanted at the mere thought of it, it's still a mere tool bound by the creativity of its artist, even if the artist were the AI itself.
Art is dead. Art is always dead.
Did people complain when digital paintings became a thing?
> Did people complain when digital paintings became a thing?
Yes. A lot. All the same complaints. "It's low-effort cheating. The machine does all the work. It's soulless. Art requires a physical process. It looks like crap. It will put 'real' artists out of business."
As far as I can see, the whole history of art is a long argument about whether something can be called art or not. It's always art
As a recent example, Tron was disqualified for a visual effects Oscar because it was all done on computer and therefore seen as cheating, instead of being recognised aa both an incredible achievement and a precursor for the next 40 years of filmmaking. [1]
Imagine the howls when the first film with significant amounts of AI effects is released.
> As I understand it, generative AI is a sort of average machine, in that if you ask it for a picture of a dog, it’ll give you the most average result for the prompt, model and whatever randomness there is in the system.
"Generative AI" isn't a single technology. Text and image generation don't actually work the same way (have the same model architecture), which frankly makes it absurd we invented them both at the same time.
But for text at least this isn't true; LLMs have fractal complexity, which is why they work. ("Average" text would just be the letter 'e'.) It's harder to say how diffusion models even work.
It makes them easier to handle if you train them to act like this though, but try playing with a "base model" and you'll see all kinds of inhuman results.
AIUI diffusion models of images are trained to make a noisy image into an image without noise (reverse diffusion). Then when an image is required, the system starts with white noise, and maybe a prompt, and then iteratively performs reverse diffusion - guided by the prompt - to produce a final image.
You can enhance an image by [forward] diffusing it, then feeding it to the NN created to perform the reverse diffusion.
As an artist, if I’m not constantly pushing the boundaries of how I’m expressing myself via media then I’m just telling the audience the same thing over and over. Im not actually changing as an artist, I’m simply producing the thing regularly that is expected of me.
This is 99% of professional art you find your style and then you do that over and over for money so that you can survive in a way that you like.
If you did this with a Thomas Kincaid, it would be trivially easy because Thomas Kincaid has built his entire portfolio around a very specific type of art that does a very, very specific type of thing.
Similar for Kahinde Wiley
On the flipside you’re not going to be able to reproduce something from Duchamp or Picasso - their styles changed because they as individuals in the world changed, and they had to change their approach in demonstrating how they interact with the world.
I totally can understand the perspective from professional artists who have developed a style and capability to produce something that they feel comfortable with and live on that
we could sit here and debate about the aim of Art or otherwise, but the reality is, if the core of what you do is reproducible because you have focused on commoditizing your medium or expression type (such that it is so recognizable immediately by any human - eg Kincaid) Then you should not be surprised if your limited vocabulary is fairly easy to replace
The art masters cannot be replaced, you cannot predict what they’re going to do next because they change the rules entirely. This is how we know a great artist from a mediocre artist.
You are presenting opinion as objective fact for a subjective medium, which is surprising to hear from an artist.
This argument, and this article, are not about the "great masters" but about amateur and professional artists who find their efforts easily reproduced and outproduced by AI. These people are not protected by "This is a REAL Picasso" and some of them could lose their financial security because of it.
The article indicates that style can be copied in as few as 20 to 30 images. Could you imagine being an artist who is attempting to gain notoriety, publishing low quality pngs of 20 of their works for their Etsy store, and then someone trains a LoRa on that and profits off of your images? Hundreds of hours of time in an attempt to start your art career, to have it duplicated in seconds by computer. You have no time to grow as an artist, because your required contribution to the data set has already been made.
My takeaway is that as with many of the artists, they are reacting to the philosophical no mans land that cheap computer reproduction is putting them into...From TFA:
>As I’m contemplating this Cronenberg-like transformation of the image, I can’t help to be struck by the triviality of my own work. There’s something confronting in facing a computational doppelgänger, something akin to the uncanny valley. I’m surprised at how much this affects me, even though my whole schtick is to be reflexive and critical about style, what surprises me the most is that even though the output if “objectively” a failure1, I see myself in it. But maybe what I see in the generation, what I find actually disturbing, is the part of my work that has already been objectified and commodified, the parts of my style I spent years making digestible for clients, consistent for social media, and reproducible for easy production.
This feeling is not present in artists who are not commercial artists or otherwise produce VOLUMES of similar work - because it is functionally impossible
Why?
Again, as with picasso et al... they are performing the work of an artist. Namely it is a lifetime oeuvre - not a single period - that defines an artist.
This is my point, and you can disagree, most commercial artists are not producing art, they are designing propaganda for corporations. It takes artistic skill, but it's not what I would consider relevant for the author
The author sits at the intersection of capitalism and art - that you can produce designs for money repeatedly and so predictably as to be identically imitated is proof enough for me
Nobody can reproduce or copy my art because it's PHYSICAL and hanging on my (and many others' including our own jacquesm) wall. They can try, but forgeries are very hard to do.
> This is my point, and you can disagree, most commercial artists are not producing art, they are designing propaganda for corporations.
Tbf I think the artists most threatened are the ones doing fetish or furry commissions for individual clients paying $100-$300.
Corporations need to be able to make revisions to a style or have artists advise them on where to make changes. AI isn't good at that. (Getting better though.)
> The art masters cannot be replaced, you cannot predict what they’re going to do next because they change the rules entirely. This is how we know a great artist from a mediocre artist.
I agree that in principle the best working artists are those who know how to create, modulate and discard artistic languages over time -- creating and abandoning rules for sequences of work -- but AI systems can clearly be unethically trained to mimic art in those styles, and there are many living artists (or estates of recently departed artists, supporting their families) whose incomes are defined by some periods of their lives where they worked a particular set of rules in a way that became definitive.
So it's not enough to say that great artists won't be affected because they change the rules all the time. Most of them will still be ruined by this, because as soon as there's enough work in the new style it can be stolen.
> because as soon as there's enough work in the new style it can be stolen.
This is where the oft-cited comparison to previous advances like the camera breaks down - AI has an ongoing parasitic relationship with the mediums it is displacing. The camera doesn't need a thousand photorealistic paintings of a new subject to be made before it can capture that subject going forward. AI is an unprecedented situation, as far as I can tell, where a new artistic medium greatly de-values prior ones while also being reliant on those other mediums continuing to be used at scale in order to avoid stagnating.
This week I completed a sculpture which I intentionally made to be non-photographable by choosing lighting and materials which showed off high reflectivity and high chroma so that screens would not be able to reproduce it. To the human senses it is something that really steals your attention, perhaps because we have becomes so used to the constraints of our reproduction technology and this is not that. The Prussian blue pigment sucks you into the astral realm and the gold is hypermodern reference to the Apollo program. It started a lot of conversations.
The work is a critique of photography, prints and screens and all of postmodernism. It also is also a warning about being satisfied with mediocrity.
I always find it philosophically entertaining when I see a material in real life that can't be photographed with a phone like this. (Of course that includes gold foil, mirrors, blue LED concert lights, the sun, the moon… so, pretty common.)
But try making one that can't be videoed or 3D scanned with Polycam.
Do we want art to be highly valued? Doesn't that make it out of reach for the masses? I think making art more approachable is a good thing. I probably would not have gotten into visual art if I could not use a camera, and had to paint or even dance. The camera de-valued painting by saturating the visual art ecosystem. Image generation does the same once again.
Consider: generative art technologies (as all good technologies) raise the skill floor, without lowering the skill ceiling.
As a photographer I am shocked to see other photographers who are AI proponents making the comparison to the beginning of the photography era. (Though it's usually the very bad photographers or those who haven't actually done any reading about the impact of photography on art who make this point)
Art's one hope to avoid this parasitism is to move back into the realms of the purely physical, I think (and as a photographer that includes me).
The fact is that it is much like that, because no one then knew what was going to happen and it seemed to them like an existential threat. The mechanism of how it works or its relationship with our previous technology is irrelevant. The comparison is about everyone freaking out because no one knows how this is going to play out.
But history has shown time and again that carving out a niche in art by using technology is always fleeting and any security you think you had was illusory. Just looking at Hollywood, you have: (a) transition to sound (b) emergence of televisions (c) transition to color (d) effects revolution (e) digital replacing practical effects (f) digital replacing film
Each one of those transitions obsoleted a generation of specialists with niches, yet Hollywood and the people it employs are still doing just fine.
The majority of people complaining are complaining that their “specialness” is being commodified
What actually happened was they commodified themselves and their work, in order to gat paid and maintain their position in the world as an “artist”
They are now pretending that that what they were doing had more value than it actually does simply because it takes skill and craft and they define themselves by it
Guess what, making metal fenders by hand takes a shit load of skill and craft, same with exterior plaster artists same with molding and crowning artistry - 100% of those kinds of workers were replaced perfectly by Machines.
You cannot replace demand for art qua art - because you are selling a moment in time of an artist expressing a fully felt emotional release. You can’t buy that because it cannot be created via money. It is a unique and temporal experience that can only be done by one person at one time in the world.
you can however, replace demand for art in service to commerce and it is a function of capitalism that that will be replaced because by function, it is in service of promoting a commercial behavior, not the expression of an artist in communicating themselves
At a certain point, art sells because of the artist not the art. Picasso's sell because they are Picasso's. An artist starts by creating a style that separates them from the pack but eventually turn into a brand that sells because they artwork is from that artist. For people that successfully are able to turn into a brand, great, but I feel new artist are going to be heavily kneecapped by AI.
I’m no artist, but I’m not sure I agree. As an example, Cubism (Picasso) or stuff by Pollack (arguably another ‘great’) seems quite distinct and therefore reproducible – at least visually if not physically (paint texture, etc.).
Goodluck replicating and replacing a leyendecker or a Gibson or a Rockwell. Getting to that level of technical proficiency is more than just settling into a style.
Please is there anything like this for writing? I am serious, I do write and want to experiment. Please would you have any suggestions for me? Thanks for your time.
Writing is the simplest of all. Really trivial actually. I create new styles almost every other day. See the latest entry on my blog for a style i created last week, and this video [1].
Different styles can be created by taking advantage of the context window.
1) Take some random style of a writer, a blogger, or even a website.
2) Generate some examples you may like, and annotate them, something like: Example Chapters.
3) Put the machine to describe the text of Example Chapters.
4) Select some keyword descriptions, and ask to imitate the writer's/blogger's style plus the keywords.
5) Generate some more examples, and delete the previous ones. Now it will start converging a lot better.
6) Ask the machine to use the Example Chapters as a reference, and DO NOT REPEAT examples. Do not repeat has to be written exactly like that.
7) Generate 1 or 2 chapters of your desired text, by generating 10 different drafts, and stitch them together by editing them a lot by hand. That's plenty difficult and time consuming.
8) As soon as you have 2 good pages of the text you like, delete the examples you do not need them anymore.
9) Ask the machine to continue the 2 Chapters it has written, but it doesn't actually need to continue something. Just ask to to write the story, but in the prompt it has to be asked to continue.
10) It is best the 2 Chapters of your story to have the same characters with exactly the same names as the story you want to write.
11) When you ask it to continue, obviously you continue using the same style of writer/blogger plus keywords.
Infinite potential for sure. You can instruct it to be more terse, poetic, avoid pronouns when possible, dial up the humor and so on. Take also a look at David Shapiro's use of Claude and GPT for writing [1].
My process however is different than anyone else. By putting the machine to describe the writing, i use the keywords, remove one i may not like, and add one keyword of my choosing. Then i generate some more Example Paragraphs or Example Chapters and do the same again.
I slowly push the machine to generate examples until i am satisfied with the result. Then i go on and add my characters, and at last use the last bunch of examples to generate my story. It is not easy at first, but once i got the hang of it, now it is trivial.
The challenge I see is that (in the US at least), "style" is not copyrightable in the arts. IMHO this is a good thing because artists learn and copy from one another, and if it were copyrightable that would lead to no end of lawsuits between media companies.
So my concern is trying to apply copyright to training AI models can easily slippery slope into more draconian copyright rules for human artists.
I don't think there is a slippery slope in the foreseeable future, unless you buy into sci-fi views of AI being like a human. We need to update the laws around copyright in response to these machines. It's similar to why copyright laws exist in the first place: the concept was developed in response to the printing press.
At some point there will be a real I, Robot problem about an AI artist that actually understands what its drawing and doesn't depend on interpolated plagiarism of inhuman amounts of data. But we aren't even close to that yet.
Style is hard to objectify. It’s a lot easier to determine whether a 3D model depicts Super Mario than whether it’s in the “modern Nintendo” style.
Style is also very broad. It’s even harder to determine whether an 8-bit NES sprite is in the “old Nintendo” style, because 8-bit sprites don’t have much flexibility to distinguish themselves.
Broadness: imagine if whoever first came up with the “low-poly 3D” or “flat material” or “voxel” aesthetics could copyright them and prevent anyone else from selling anything in those styles. What defines a style as narrow enough that it can be copyrighted? And what if that definition changes, e.g. if a a specific voxel style gets copyrighted, then someone else discovers a brand new way to render voxels super efficiently in only that style?
Objectivity and similarity: an artist can make a concrete object or character which is very similar to a copyrighted one but also clearly distinct. This is very important, because if “similar” objects could violate copyright, where is the line when something is dissimilar enough? Ultimately it would be very far for small artists, who can’t afford to risk lawsuits; vast swaths of clearly not similar characters and objects would be blocked off from them, because in the eyes of the law and without good representation, they’re no longer “clearly” not similar. In fact, it may be hard for an artist to even come up with an object or character that doesn’t risk a copyright lawsuit, since there are more copyrights that anyone could fully know. (At least to my knowledge, with copyrightable characters and objects this hasn’t been a frequent issue; but if it is, copyrightable style will make it worse, so for the sake of the argument...)
Copyrighting style is basically copyrighting the “similar” works. There’s a fine enough line between whether a character or object is “similar to” or “the same as” another (again to the best of my knowledge). But there’s no fine line with style. If one tries to define a style with objective criteria like making their “style” a specific stroke thickness and color scheme, generative AI users will just create art which falls right outside of this criteria. If one tries to use an AI classifier (ironically) to deduce whether something is “the same” or “similar but distinct”, it will be foiled by AI-adversarial manipulation and its effectiveness will be endlessly disputed in court. And if one defines their style with very subjective judgements, that leads to the issue above.
One reason for that is that, like with ideas, society benefits, from styles intermixing. So making them "copyrightable" would be to the detriment of society.
But the other reason might be that it has just not been feasible to actually measure things like “style“. And if you cannot, in your lawsuit, put into words what you claim is, and where its boundaries are, it’s impossible to enforce it.
It’s very much possible that people will forget the former reason and jump on the latter, now that one might be able to use AI to actually quantify, measure, and map things like style. In fact, I wouldn’t be surprised to see companies like OpenAI play Good Samaritan and “helping legislators“ solve the problems generated by AI. Only after they have secured the own datasets, of course.
The work of a good artist is much more than a style. You can express ideas, tell stories, generalize, convey something in a specific manner. If style can be reproduced with ease, the other qualities of what makes art great are harder to reproduce by AI. So if your art is more than just a style, you shouldn't be worried about AI.
I only have a Ryzen 5 3600x and a 7650xt, and downloaded AMD's LLM tool. It runs, and chats. I have thought about training it on the 4500 pages of LJ entries I saved to PDF (I know I'd have to extract the text, and then spell check that puppy. or not. Maybe I should just leave my misspells if it's a doppleganger).
But it seems it'd take a few weeks of running nonstop to train? Everywhere I'm reading says it's hard to determine the actual time required.
I would think 4500 pages of text would be a large enough volume to train. (assuming 500 words per page? If not more)
What's a good minimum number of blog entries/words to train a thing. I would think the more the merrier, but I ain't spending non-stop processing for weeks at a time.
I hate AI, I have no idea why I even want to try this, but... The data is there and it's local to my machine not shared with upstream, so I figure it'd be fine to mess with. Is my estimate in the ballpark?
Depending on the finetuning tool you're using, you can just start the training run, and then it shows you how long it'll take. Like give it 5 mins to stabilise, then see the estimated duration.
Axolotl is a good finetuning tool if you need one.
I should clarify - I'm ambivalent on a lot of the current structures (political, economic) that this is being created in. It's less about "AI" and more about how it's being shoved down our throats. I have ethics concerns (even though I understand and am slightly sympathetic to the arguments that "training" an AI on data is like a human studying previous works - though I don't think that's what's really happening). I have concerns about this being done "at scale" (especially seeing what Nvidia is doing to build giant massive data centers - though - as with most tech "it gets cheaper/cleaner/more efficient" so I can understand over time this could go down).
Being local reduces some of these issues... But even then I'm not sure I trust the big actors the renege on the original deals and start hoovering what they can.
That's why I'm willing to play with this locally on my own end just for funs while not necessarily buying into the hype, or trusting the players/creators that they have the best intent besides more exploitation of natural, human, and data resources.
Sometimes it's ok to have complex opinions, and admittedly I was a bit strong in my condemnation above, but also brevity in comments is a virtue, better to underexplain at first than overexplain and clarify if needed. (He says as he spends multiple paragraphs over-explaining in a followup)
AI is to the point where it can create it's own style. We talk about these styles as if it's some kind of sacred thing. What you don't realize is that we're past this point.
There's a latent space of different art within a certain style category, but there's the latent space of different categories itself. AI can already traverse the space between art in a single style category AND it can traverse the gradient space BETWEEN categories. It can even go beyond that.
AI has turned everything into a freaking best fit curve.
Copywrite is such a mundane question. It's just a fight against the inevitable truth. That human creativity are just points on that curve. Nothing more, nothing less.
I view this as a BAD thing. Copywrite is good, it's great. It's a law that implies what humans create has meaning. But we have to face the truth. That the output we produce is in actuality just one part of a high level pattern. That's all we are.
Human creativity turns into a point on that curve. One mission of art is to find a different dimension, out of the curves reach, until eventually it becomes more common and the curve can be fitted to it again. AI cannot think outside of the box because it cannot think at all, there is no meaning behind what it does.
>Human creativity turns into a point on that curve.
All human creativity are points ON the SAME Curve.
It doesn't matter what turns what into what into some point on the curve or finding a different "dimension"
If you come up with an algorithm that can traverse that curve you've found the algorithm for human creativity.
We are close, deadly close, to the end.
Especially given the fact that these AI algorithms literally treat the problem as a best fit curve from a mathematical perspective. Like the analogy I made is not even really an analogy, it's the reality of how these algorithms actually work.
> All human creativity are points ON the SAME Curve.
That's what they said about mathematical proofs.
> If you come up with an algorithm that can traverse that curve you've found the algorithm for human creativity.
And that's what they said about programs that take finite time to prove whether or not other programs halt.
Neither of those turned out to be true though. (In particular, because the curve you're talking about is infinitely large and so you cannot compute on it in finite time.)
Also, real world things cannot be reduced to their bit descriptions because they have metadata even if their descriptions are identical: https://ansuz.sooke.bc.ca/entry/23
Surveys of humans strongly demonstrate that they are rather impressed with their mountain climbing accomplishments thus far. The main shortcoming to humanity is only the actions of those other people.
Whether or not there are copyright issues depends on the laws of specific countries. In general though it's a lot more legal than currently legal things like Google Image Search, because it's much more transformative than making thumbnails of other people's images is.
But if you don't want it to be legal then just pass a law saying it's not.
Maybe? If they show you started with their code and abstracted your code from their's mechanistically, then it seems reasonable to say that you copied from their code.
If you, a person, read their code and imitated the style - assuming it to be definable as Google's style and not easily confused with the style of others (which would suggest it was not distinctive) - then yes, why not? You would have directly abstracted the essence of Google's work from them.
They should be entitled to the copyright for at least 7 years, maybe up to at most 14!
Could you expand on what you mean by "its style of coding"?
Impressionism has scope. Most people can tell a [famous] Renoir from a Monet and would not think Renoir copied Monet (in general). The devil is in the detail, courts have too decide what is too close and what isn't, but it's very much 'I know it when I see it', I feel. Allowances to be made for distinctiveness.
I'm not a movie buff, I know a few films from Spielberg's oevre, but I've never seen anything that I thought was a copy of his style ... perhaps you have examples you feel are so close that 'a blind man on a horse' would see (sic) that they are copying Spielberg's style?
I like the change of tack though, would appreciate some more push back.
I'd like to stick with Impressionism. Perhaps you think Renoir is distinct, and sure, he is. Now what about the hundreds of non-famous artists that are not Renoir and are instead just trying to make novel art in exactly the style of Monet? You think their original works are illegal? You think they are more illegal than an ai generated monet painting that uses a different color palette to be distinct?
I'm not very interested in debating this further because I think we just disagree. I'm glad this is not the way copyright law actually works.
"Novel art" in exactly the style of Monet, is not entirely novel. I think it would be reasonable to allow Monet to have sole right to produce his style for 7 years or so, and then it enter the public domain. That rewards coming up with a new artistic style, which I feel is at least as valuable to reward as coming up with a new piece of artwork in an established style.
Again, my art history is weak, but when Picasso came up with his first cubist work, the cubism part of that work seems as valuable, and worthy of reward, as the actual instance of cubism that the work embodies?
There's a generally derivative style that follows the spirit of the age, and then there's a clearly copied style from a specific maker/creator/artist. It seems poor to protect the individual instance of an inspiring new style, but then not reward the creator of that style, not even with an acknowledgement.
FWIW, I'm not sold on either side of the argument. To me that's not how online discussion works. Also, I absolutely would not want copyright terms of the length we have to protect anything like this (nor anything really, they're obscenely long).
I would not of course want to inhibit the natural progression of styles, Monet's landscapes bore some similarity to JMW Turner, and a Chinese style called Lingnan, but this is a different thing to taking Monet's pallette used for water lilies, then painting a similar subject in the exact same style. If you do that immediately after Monet first presented his style for water lilies then I'd at minimum want you to say 'in the style of Monet' as a subtitle, and preferably have to pay him a little of your profit for the first few years.
Thanks for your interesting, thoughtful responses.
Copyrights only apply when there is copying; independent production is allowed. As it's tort (mostly!), you only have to prove they copied on balance of probability (in jurisdictions I'm familiar with), but you still have to show they copied.
Fwiw, copyright is named for "rights", legally allowed/restricted activities. "Copywrite" is the verb meaning to write content for a publication, content being referred to as "copy" in this context.
It's just we will no longer be able to differentiate it from machine generated content.
Machine generated content will become more superior and more prolific such that it will cheapen human meaning.
You'll find people will begin clinging onto the last vestiges of human created things like real painted art or vintage sculptures like they do vinyl records and non-digital books: An exercise in irrationality in attempt to reverse a war already lost.
Perhaps that is what will save human meaning in the end. Self delusion. We will deliberately ignore superior and better content and tell ourselves that human works are and always will be the best. That's a little of what's happening in this thread already.
Human art and appreciation thereof is already by and large an exercise in irrationality, so I don't see why this is necessarily a qualitative change. If people want to imbue "human made" with a special meaning, that's really no different than thousands of other ways we pretend things into being every day.
I think, in using terms like "rational" the grandparent is confusing
AI with science and presuming some kind of objectivity. One cannot
make any sensible claim about what is "more superior" (the "more" is
redundant by the way) since quality of art and meaning is not arranged
on any cardinal scale. Humans get to define what is better, and that's
the final judgement. And should they choose to define AI as
"meaningless", then it's meaningless.
Now as to whether it will be "more prolific", that's a another matter.
Cockroaches are already far more prolific than humans.
I think they are arguing that by already established criteria, AI art can easily exceed human art in quality. So we can of course rejig our criteria, but in doing so we cannot help but acknowledge that the original ones no longer cut it.
Yeah. The difference is at one point in time we didn't pretend, now we do. And therefore the lie we tell ourselves is more obvious. Many people will see through the lie.
lol sometimes I wonder if the reason so many tech bros make so many grandiose statements about how AI has already trancended the bounds of human creativity is because they've never really had actual experience with art and literature and music growing up. They say things like this and then you go and look for an example of these incredible works of artificial art and it is the biggest pile of shit you have ever seen.
Just like how they think that these glorified typing assistants function identically to the human brain. I mean maybe they really do replicate the level of creativity and intelligence of the people who believe these things, but maybe the people with decades of research in linguistics and neuroscience are right to be skeptical.
Its kinda like when people talk about research mathematics like its all adding really big numbers together or calculating really tricky integrals.
If you need intelligent things to act exactly like a human being I think you're going to look bad in the future where we have aliens or uplifted ravens and this looks racist/speciesist.
Being trapped in a computer with no environmental interaction is obviously a big difference though.
> but maybe the people with decades of research in linguistics and neuroscience are right to be skeptical.
Like other kinds of academics, many of those people are cranks so they're not necessarily right. For instance, some of them are Chomsky.
(But it's their job to be wrong most of the time so of course this is expected.)
No, you need intelligent things to be intelligent.
And it would probably be good to distinguish between "things that appear superficially intelligent" and "things that have genuine capacity for internal experience" while you're at it, since the popular perception on this forum seems to be to just declare that physicalism/materialism is obviously correct and explanatory and we can just brush the hard problem of conciousness under the rug. People are so eager to endow the statistical model that has been designed to produce responses that are statistically feasable in a given context (a very impressive feat of technology to be clear) as having internal experience.
Chomsky happens to be more right about LLMs than 90% of the popular discourse will ever be able to understand, which isn't that suprising considering the impact if this "crank's" earlier work.
Chomsky's critique of LLMs was somehow not even correct about his own theories (https://x.com/profraha/status/1634311282135318529). And of course has the usual problem of 1. claiming LLMs are random statistical models (they aren't, the sampler is) 2. claiming instead of generating correct text they only "seemingly" generate correct text and not explaining why this is any less impossible.
But people have built explicit grammar knowledge into LLMs (it's in llama.cpp) and nobody seems to need it to make English text.
Of course the main reason Chomsky is a crank is that he's completely convinced he's right about politics despite his unbroken record of endorsing every mass murderer in the world as long as they're anti-American.
I don't think that the opinion "70s intellectuals' love of the genocidal Khmer Rouge is bad" is "my" politics. I think this is a commonly held opinion actually.
Materialism is rather obviously correct when it comes to consciousness just based on what we know brain damage does to humans.
But what does this have to do with problem of consciousness or internal experiences? There's no reason why the latter cannot be a material process.
As for the linguists... people forget these days, but before the AI winter, there were many experiments trying to build an AI by modeling the language. And what they learned from it is that when you have enough compute and just throw it at a neural net, you get better results (there was even a joke about AI progress being inversely proportional to the number of linguists working on it). Which, to put it bluntly, means that our understanding of language is still very lacking, and quite likely to be fundamentally wrong some respects. I don't think those people are in a position to judge at that point; not until they figure out language enough to program something that can talk as well as GPT-4 does.
There is zero understanding of the cause of internal experience. We can't even strictly define what it is. My contention isn't that materialism doesn't appear to be a reasonable canditate for cognitive process, but that we still have no idea HOW any of that happens. We may never know.
People who claim that it is obvious that neural networks and LLMs replicate the functioning of a brain, any brain - let alone human, are just wrong. They are wrong to assume that it is obvious that just making bigger LLMs will somehow generate a being with the capacity of internal experience, whatever that may mean. They are wrong when they act like they have solved the problem of understanding cognition by just forgetting to mention the hard problem of conciousness. They are even wrong to simply assert that neurons and neural networks are "to do" with cognition. Hallucinogens and brain damage affect subjective experience, and both of those things are involved with the vascular system of the brain. Would I be right to say the vascular system produce cognition? Why doesn't anyone argue in favor of that view since it is not too far off from the same reasoning?
Obviously, it's reasonable to assume that neurons are somehow involved with subjective experience, most serious neuroscientists and other researchers would hold that same assumption. But you need to EXPLAIN how, preferably with evidence. It is the burden of the one claiming they have finally figured it all out to present a convincing theory, or at least conjecture the path to get there. Loudly and smugly asserting that your assumption is correcter than the other people who have spent decades working on the problem doesn't make the problem dissolve, but it is especially irritating when the incentive of the loudest people doing so is in the pockets of the shareholders they are indebted to.
There's a long tradition of random art and if you like that, there's dril GPT and Magic card GPT that produced lots of good absurdity. Of course, there's a human curation pass after it to let it be called art.
Yep, you're right. But idk why you phrased that like it's a rebuttal to my criticism. This is another failure of our economic system, for incentivising the owners of art production to churn out literal mindless slop to cut costs, instead of safe guarding artists and creators for the sake of their intrinsic value to society and humanity.
Since, of course, they must eternally seek to reproduce capital to the detriment of everything else.
Well your argument is admonishing tech bros as if they don't get it, but generally they do get it completely in this context. The goal here is not replacing great art. It's replacing art assets.
The distinction is philosophical. Tech bros consider "art assests" or consumer oriented products as art. In their eyes its capibility as a tool for profit is a downstream effect.
My point was that since their entire purpose in life revolves around developing and refining technology for the reproduction of capital, it reveals their blindness to any deeper aspect of the human condition than what can be sold for a profit. When people post 3d versions of the Mona Lisa or a shitty remix of Monet's water lillies with breathless praise, it's because in their eyes the stylistic reproduction or reinterpetation of the literal shapes and colors on a flat screen is all there is to art, not merely because it demonstrates utility in replacing already exploited sectors of creative workers.
I don't think we disagree about any of this. I think that I'm just making a tangential point to the article in response to a dumb comment, while you are talking about the subject more directly. Otherwise I fail to see what you think is naive, if only because I've seen so many examples of this rot of understanding around the essence of artistic expression from the mouthpieces of silicone valley and our wider, increasingly corporatized, culture.
> My point was that since their entire purpose in life revolves around developing and refining technology for the reproduction of capital, it reveals their blindness to any deeper aspect of the human condition than what can be sold for a profit
And I think that's grossly missing the point. The vast majority of people making art are just trying to make something look aesthetically nice within some sort of provided communication framework. And that's perfectly fine.
Frankly I think elevating artists to the pedestal of the illuminating the human condition is very damaging to their actual interests
Is creating things of aesthetic quality not of value with respect to the human condition? Is not some subset of that aesthetic quality derivative of the intuition and intention of a human mind? Sure, if you want to build some tehchnology to make designing packaging for dog food more effecient go ahead, as long as it isn't the previous dog food packaging creators that are being made to absorb the economic cost of the new technology (which they are).
But I hope at last some subset of human art strives to do more than that, and I think it's a bad thing that the incentive to increase revenue at the expense of quality has now expanded past being a tool to improve efficiency of laborious tasks and is now instead cutting costs via cutting artistic value.
Fast-forward in a few years, perhaps we could claim that if even a basic computer is able to do the same, the originality of the work is low, it may not meet the threshold for being copyrightable ?
imho the whole GenAI and artists and luddites thing is way overblown and mis-characterized. The outputs simply aren't up to snuff and people hates it. The negative emotional outrage comes first, then the logical explanation ex post facto, everyone can see that.
By the way, I have one of numerous back-burner ideas about the GenAI rage problem - why isn't anyone trying to "close the loop", like by attaching a gaze tracking and mood detecting device to a human(like a camera pointed at such tool's developer) and chaos monkeying the image with noise, maybe with a bit of help from generative algorithms, until the human is satisfied? Then operation log of that monkey tool can be put into GAN side of diffusion or made into a new tag for embedding or something.
I think that saves everyone's time a lot. There shouldn't be ethical concerns to it, at least to same degree that running A/B tests on unsuspecting bunch is considered totally ethical and morally acceptable.
It's astonishingly simple to train both LoRAs and Embeddings nowadays -- if you combine it with tech like ControlNet, you can more or less have direct control over the illustration as well (see this example I converted to Hollie Megert's style with barely any effort [1] using the model from [0]) -- but FWIW I was playing with the WhatsApp AI stickers recently and human-made illustrations still have a "je ne sais quoi" that sets them apart, vaguely reminiscent of "realistic" animation vs. "squash and stretch". I would not be worried as an artist*, but I would be worried if I am making assets for corporations in Corporate Memphis [2].
[0] https://waxy.org/2022/11/invasive-diffusion-how-one-unwillin...
[1] https://i.imgur.com/ZwtKug5.jpeg
[2] https://en.wikipedia.org/wiki/Corporate_Memphis
* In fact, I think I'd be excited, it can be a great tool, and people who lack artistic sensibility might find it almost as hard to use as to make art themselves. However, I do think there needs to be some kind of protection from having your 'style' commoditized without consent, using unauthorized images of your own art.