When Technology Follows Art
resobscura.substack.com
resobscura.substack.com
> the Nvidia GPUs [...] were originally inspired by [...] the needs of computer graphics artists in the gaming industry.
I think this passage demonstrates a misread of the quoted section[1] from the interview with Huang. The RIVA 128 was a 3D accelerator / video card for gamers, not game artists. Further, it was pretty clear by 1997 that 3D graphics for games was not a fad nor by that time was it a radical new approach -- Nvidia successfully entered an established market.
[1] Huang liked video games and thought that there was a market for better graphics chips. Instead of drawing pixels by hand, artists were starting to assemble three-dimensional polygons out of shapes known as “primitives,” saving time and effort but requiring new chips.
Early 90s: SGI invented OpenGL to make realtime 3D graphics practical, initially for CAD/CAM and other scientific/engineering pursuits, and started shipping expensive workstations with 3d accelerated graphics. Some game artists used these workstations to prerender 3d graphics for game consoles. Note that 2D CAD/CAM accelerators had already been in market for nearly a decade, as had game consoles with varying degrees of 2D acceleration.
Mid-90s: Arcades and consoles starting using SGI chips and/or chip designs to render 3d games in real time. 3DFx, founded by ex-SGI engineers, created the Voodoo accelerator to bring the technology down market to the PC for PC games, which was a rapidly growing market.
Late 90s: NVIDIA entered the already existing and growing market for OpenGL accelerators for 3D PC gaming. This was a fast-follow technical play. They competed with 3DFx on performance and won after 3DFx fell behind and made serious strategy mistakes.
Later 90s: NVIDIA created the “GPU” branding to draw attention to their addition of hardware texture and lighting support, which 3DFX didn’t have. Really this was more of an incremental improvement in gaming capability.
Early 00s: NVIDIA nearly lost their lead to ATI with the switch to the shader model and DirectX 9, and had to redesign their architecture. ATI is now part of AMD and continues to compete with NVIDIA.
Mid 00s: NVIDIA releases CUDA, which adapts shaders to general purpose computation, completing the circle in a sense and making NVIDIA GPUs more useful for scientific work like the original SGI workstations. This later enabled the crypto boom and now generative AI.
Of course, along the way, OpenGL and GPUs have been used a lot for art, including art in games, but at no point did anybody say "hey, a lot of artists are trying to make 3D art, we should make graphics hardware for artists". Graphics hardware was made to render games faster with higher fidelity.
That said, starting in the early 1990s is missing the whole first half of the story, no? Searching Google Books with a 1980-1990 date range for things like "3d graphics" "art" or "3d graphics" "special effects" yields a lot of primary sources that indicate that creative applications were driving demand for chips and workstations that focused on graphics. For instance this is from a trade journal for TV producers in 1987: "Perhaps the greatest dilemma facing the industrial producer today is what to do about digital graphics... because special effects, 2d painting, and 3d animation all rely on basically the same kind of hardware, it should be possible to design a 'graphics computer' that can handle several different kinds of functions." [https://www.google.com/books/edition/E_ITV/0JRYAAAAYAAJ?hl=e...]
It's not hard to find more examples like this from the 1985-1989 period.
Of course graphics hardware was also used for more creative purposes including desktop publishing, special effects for TV, and digital art, so you will find some people in those communities vaguely wishing for something better, but artistic creation, even for commercial purpose, was never the market driver of 3D acceleration. Games were. The hardware was designed for gamers first, game programmers second, game artists a distant third, and for nobody else.
The closest thing to an "art computer" around that time was the Amiga which targeted the design/audio/video production markets.
It was mostly gamers. As a gamer from that time, the hardware was marketed to gamers, hard. I don't doubt that artists had an impact, but the world had many, many more gamers, than artists and gamers spend money for the best/mostest/etc.
I mainly know this from living through the CGA/EGA/VGA/SVGA/3D add-on card/3D era.
Thank you for taking the time to delve into this. While I may not agree with your conclusions, I respect your work, and the effort put in. :)
Edit: this discussion is interesting because I have always just taken it for granted that video games are a form of art. Clearly others don't see it that way, which is fair! Nevertheless, I think a strong case can be made: https://en.wikipedia.org/wiki/Video_games_as_an_art_form
Gamers aren’t primarily spending time or money for the art and neither was NVIDIA. I will grant that the hardware improvements did make the visual aspects more lifelike and detailed and that allowed for increased artistic range, but production costs generally increased accordingly.
I was in IT when we started the transition from 2D cards to 2D cards with a additional 3D card, to the final form of 1 graphics card.
These 3D cards were NOT marketed to professionals, they were marketed to gamers, because gamers drove the tech.
- art has always been about taking existing images and ideas and tweaking them to make them your own. As Grayson Perry says "I believe in the Chinese whispers approach to art, I see something, copy it and then I make it a little better"
- very renowned artists such as Peter Doig use 'found images' i.e. a photo of people in a boat as the prime subjects in his paintings. Yet if I generate 100 images with my own prompts then that is somehow cheating?
- as the article says, artists have always used whatever tools they can find to make better art. If a trained random number generator helps me produce the image I want to produce then how is that any different?
I don't think that AI art is fundamentally uninteresting (I am a machine learning researcher, so I think it's rad), or that it can't be used as a tool to make great art. I just think that the final image is not the point of art. The point is that a person practiced for hundreds/thousands of hours to develop a creative skill and then spent several more hours painstakingly applying that skill just to create the image. If AI is incorporated into a high-effort, high-skill process, I don't have an issue with it.
I know Matisse kept several works in his studio for many many years pondering and struggling with what to do with them and the actual mark-making was very rapid e.g. The Red Studio painting
For example, a beginner painter will take much longer to make the same painting as a master; so judged on effort, it should be better? Or if I make a minature cathedral out of matchsticks, then it might take me decades - is that therefore more interesting artistic output than a sketch that Picasso (say) made in one smooth line that took minutes?
Of course, a reasonable answer is that it is some combination of effort, aesthetics, personal choice, and fashion. Maybe it's like a recipe where the balance of these ingredients has to be right.
In fact, since AI is trained on a much bigger data set than humans, AI art is in fact higher-effort than human-made art (and therefore, according to the GP, more interesting).
We loosely apply this as a premise for intellectual property rights, but it's contingent on the state being the referee - and many cases have ended in rights being taken away from inventors because someone else worked the legal system to take it away.
In the case of generative training data, it's largely being used as a commons resource. The more you are inclined towards labor theory of property, the more it "feels like" theft, because the original laborer would be credited as owner - the labor of the machine is derivative of that labor, and largely non-transformative since it results in imagery resembling the original, versus metadata, statistics, etc.
Labor theory of property has many philosophical issues, as outlined with the "monkey selfie" case:
https://en.m.wikipedia.org/wiki/Monkey_selfie_copyright_disp...
But that is what webmaven was proposing. "your argument is mitigated by considering the cumulative effort of learning to produce art"
So do you count the training and practice a human and the AI took in order to be able to produce the art, or do you not count either of them?
Do you subscribe to the view that effort is what makes art valuable?
A master painter paints an amazing portrait with a few brush strokes in 10 minutes. A teenager creates a matchstick cathedral by working extremely diligently for 2 years. Do you think that the matchstick cathedral is more artistic than the portrait?
If you say yes to the first, and no to the second then there is a seeming contradiction. Because clearly more effort went into the years to cut matchsticks than the minutes to draw the portrait. To resolve this seeming contradiction webmaven proposed that we should count the hours spent acquiring the skill. The master spent a lifetime drawing and painting until he got good enough to capture the essence of someone with a few brush strokes. Are you agreeing with that argument? If yes, why does it not apply to an AI? If not how do you resolve the contradiction?
> I don't think either of those count because they're not art-specific.
The training of a neural network is very art specific. It looks at art. It creates, gets criticised, creates again on repeat.
Don't get me wrong. I don't think the AI is more artistic just because it encapsulates more effort. I think the mistake is that effort is not what makes art valuable. It is not a pineapple pooping competition. Emotional impact is what makes art art in my opinion. (Which is a different question from what makes art have monetary value.)
Edit: I decided to add my two cents about using time spent as a measure of effort. I think this can be misleading and I have a counterexample, although admittedly it isn't from art: think about an athlete running 100 meters. This doesn't require effort just in the training before, but also in the act itself. And the less time the athele takes to run 100 meters, the more is the effort required.
As for my own opinion about art, I think that anything could be art: what counts is the idea behind it. How this translates to this specific problem of AI generated images I'm unsure though, as it would, at least in part, depend on how much of your idea you can convey through a ML model.
Not as a hard rule, no, although there is absolutely art where the effort to produce it is an inherent part of the value of it.
> Do you think that the matchstick cathedral is more artistic than the portrait?
No, not because of effort involved, anyway. If both things are trying to convey something about the human experience, they're both equally "artistic".
I subscribe to the view that what makes art valuable is that it communicates something human that can't be communicated well otherwise[1].
This is why if there existed an AI that produced art all by itself, I would consider the work produced to be pretty pictures, not art. An artist using AI as a tool to produce human communication, though? Totally art.
> (Which is a different question from what makes art have monetary value.)
I agree. The price tag of a thing says nothing about its artistic value. It only speaks to commercial value.
[1] This is also why I don't think there is any such thing as "bad art". Art can be successful or unsuccessful, but "bad" and "good" are subjective aesthetic judgements. When people say art is "bad", what they mean is it's art they dislike.
"It took me four years to paint like Raphael, but a lifetime to paint like a child" - Picasso
If you use AI, very few people, if any, could tell if you or an AI emulating you made a certain painting. I think this is a key source of the rejection, it's saying "Even if you can't tell, you can trust my paintings were made by me and not an AI, I stake my reputation on it."
* Aesthetically pleasing - often there is a traditionalist idea that only landscapes and portraits are 'real' art, and everything else is rubbish
* Financially successful - the market sets the price of a work, and that is the value, no matter what your personal idea of it
* Emotionally affecting - if the work is moving in some way, or you get something from it, then it has value
Unfortunately, people using different definitions of value here will likely never agree when these values clash when considering a work of art.
I guess this is a long-winded way of saying that I reckon some people have just defined generated images (or partly generated) as 'not Art' so will never accept them. I know I have changed how I think about some art/artists (such as single-colour painters like Rothko) from 'this is bad' to 'I don't care for this, I see that others do'.
I suppose I fall into this camp and why I find most AI generated images nothing even close to art. Art is truth, art is a communication between the artist and the viewer, good artists do things for their personal reasons. Generative AI is cold and unfeeling. If anything, artists that use AI should embrace that and stop trying to make so many airbrushed, Penthouse style renderings that try to mimic so much that has happened before. I mean you have the power of Gandalf, so why are you making images that look like something that's been done thousands of times over? Otherwise it's content and illustration, which both have their place in the world.
Looking at the spiral image thumbnails used in the article look like something that an artist would have done 100 years ago. I don't see an artist there. I see a content creator. If they were painted by hand I would say the same thing, but be slightly more impressed by the craft of it.
I find illustrations, amateur or professional, interesting and engaging because people's style heavily reflects what they value in the imagery they create - deliberate choices made in pursuit of depicting or conveying an aesthetic, a theme, a subject-category, an idea, a notion. Paths taken in developing their skill = style.
Current AI image generators are too blunt of an instrument to be a part of that process.
The communication capabilities / concept space of image-gen AI needs a couple more paradigm shifts before it can join the pantheon of tools for "making better art". That means pursuing what they really want to depict, not filling space on the canvas with the AI's vague notions of it. Right now "style" in AI art is just a shorthand for surface-level LORA-ripoffs of existing artists, and attempts to statistically blend them together.
It is not settled law at all anywhere that somehow using copyrighted data as training data is legal. The companies doing this assumed it was and moved quickly as fast as possible. They invented novel legal theories (“it’s just like a human”) to justify what they did, then spread it all over the internet to shut down artists questioning the legality.
If you allow AI to skirt copyright law but humans have to live by it, then human labour becomes meaningless except as training data for the next AI model, and all profit making will become concentrated into AI, while humans make nothing or next to nothing. Any innovation made by a person will be copied by the next model, which being closed source will profit only a small number of people.
Huh? At least from the argument you're presenting you're attempting to create "The Right to Read" world by RMS. Copyright is about output, not input. If I output something too close to your copyrighted content, that's the violation. Not me getting a book and reading it and sharing it with every one of my friends. Not looking by your art and being influenced by in in my artworks.
> then human labour becomes meaningless
You're 150 years late for the storyline of John Henry.
The problem here isn't machines can do everything and make us meaningless, where you're breaking is thinking that capitalism can even begin to work under this paradigm shift.
Unfortunately for me, that is a bleeding edge understanding of reality that hasn’t propagated to the billions of people who live under capitalism.
That’s the tragedy. AI is a dystopian technology under current economics.
If you change the economics AI is utopian.
But we’re electing neoliberal governments, for the most part, in the West. Million $ question: what’s to be done ?
I think that a lot of the rejection about generated art comes from the fear that it would become the dominant artistic form rather than just another kind of art.
I watched a YouTube video the other day that expressed a sentiment that I share and that may apply here as well. He was talking about Photoshop rather than generative art: (Paraphrasing despite the quote marks) "As an artist, I love photoshop for the power and flexibility it gives me. As an art appreciator, though, I would never hang photoshop art in my home because the aesthetic is not appealing to me."
It's an open secret in the art world that modifying existing works and putting your own spin on it is a common technique. But blatant copies with no such 'spin' on the original need to be called out. Also there are places like China who do clean room reverse engineering of existing tech and make new tech in the same likeness, but the engineering is completely different from the original. It's how they avoid IP/copyright suits.
Or let's take your example of Peter Doig, whose art I happen to like. He may use found photos but comparing this to AI renderings is absurd. He takes those and then ACTUALLY PHYSICALLY PAINTS THEM into a work of art. The difference is enormous.
Now if someone incorporated generative AI into a complex composition of mixed media to construct a creative piece of art that required thinking, planning, some modicum of effort at least, I could understand it. Generating 100 copies of a certain style with a few prompts is definitely not that though.
IRL art process is a subtle thing. Analog, fuzzy and mysterious.
Computer art process is crass and pinhole-narrow in comparison.