I want to talk about this US Copyright Office opinion letter (PDF) about the AI-assisted comic, Zarya of the Dawn. But there are a lot of really big topics here, so I will just have to grapple with it piece by piece, pell-mell (there’s a word I never use–I had to look up how to spell it).
There are sections in the document which reading them cemented my own views around creativity and “authorship”–often because I strongly disagreed with the USCO’s characterization. I lean naturally more towards the UK’s copyright protections for 50 years of computer-generated works, while also admitting the whole thing is fraught.
But the copyright part is just the jumping off point for me. I don’t actually want to talk through all those particulars in this post. Instead, I’ll try to capture a few of the evocative snippets that lead me deeper down this road of the actual “art object” at play here not being any single or set of images, but the fundamental underlying “hypercanvas” of latent art, if you will…
Anyway, one of the things that started to spark this intuition about the hypercanvas concept was this, by the USCO:
The fact that Midjourney’s specific output cannot be predicted by users makes Midjourney different for copyright purposes than other tools used by artists. See Kashtanova Letter at 11 (arguing that the process of using Midjourney is similar to using other “computer-based tools” such as Adobe Photoshop). Like the photographer in Burrow-Giles, when artists use editing or other assistive tools, they select what visual material to modify, choose which tools to use and what changes to make, and take specific steps to control the final image such that it amounts to the artist’s “own original mental conception, to which [they] gave visible form.”15 Burrow-Giles, 111 U.S. at 60 (explaining that the photographer’s creative choices made the photograph “the product of [his] intellectual invention”). Users of Midjourney do not have comparable control over the initial image generated, or any final image
First, this is putting aside the new generative fill or whatever its called in Photoshop, and the art in question was made via Midjourney.
One thing I’m seeing in common in this letter & a good bit of the critique I saw of my AI art books is this assumption that somehow the creative process is absent when one works with AI. But as an artist, for me that’s deeply wrong. Where does it go exactly? Does it happen as soon as you open Discord, or when you type your prompt in, or…?
It’s an assumption (usually claimed as fact by the asserter) that doesn’t match at all my personal lived experience. I am deeply deeply embedded in the creative process when I get on a really good tear with Midjourney or another AI tool. It’s absolutely a creative flow state, completely experientially indistinguishable from that experienced during any other non-AI creative activity.
Much of the USCO letter revolves around “authorship” though, which is different from creativity. I’ll get into that some other time, I’m already getting distracted.
This is tangent to my main point, but wanted to capture it for later:
Because Midjourney starts with randomly generated noise that evolves into a final image, there is no guarantee that a particular prompt will generate any particular visual output.
This “predictability” argument is preposterous. When one sits down to write a novel, have you already perfectly predicted how it will all go, and you’re merely dictating what you wrote in your mind? I highly doubt it. Or something like a Jackson Pollack painting. It’s a work that evolves in conversation with the tools, materials, and the moment, and is embedded in the artist’s life, time, and culture. Prediction is totally a red herring here.
The line immediately following (sorry, I’m still not yet arriving at hypercanvas, but I’ll get there gradually):
Instead, prompts function closer to suggestions than orders, similar to the situation of a client who hires an artist to create an image with general directions as to its contents. If Ms. Kashtanova had commissioned a visual artist to produce an image containing “a holographic elderly white woman named Raya,” where “[R]aya is having curly hair and she is inside a spaceship,” with directions that the image have a similar mood or style to a “Star Trek spaceship,” “a hologram,” an “octane render,” “unreal engine,” and be“cinematic” and “hyper detailed,” Ms. Kashtanova would not be the author of that image. See id.at 8 (text of prompt provided to Midjourney). Absent the legal requirements for the work to qualify as a work made for hire,17 the author would be the visual artist who received those instructions and determined how best to express them.
It’s confusing they use this case of a commissioned piece of art, then criticize their own thought experiment for not properly engaging a work for hire contract. They could have just as easily framed the above as:
If the author commissioned another artist under work for hire (with explicit agreement they were buying copyright), then the copyright would be owned by the author who commissioned it, not the artist who made it under contract.
But they didn’t say that, because recognizing that would undermine their legal theory. Where, in my alternative reading of the situation, Midjourney is the “work for hire” artist/tool, under the direction of the human who arranges the execution of what to do with the tool.
I didn’t even get to hypercanvas yet though, did I? Or didn’t I?
Before I get dragged into the forest of weeds again, I’ll just try to express in plain language what I mean by hypercanvas.
Like the USCO is taking this conventional reading of the artistic process of using AI tools, which says the “art object” is the fixed form copyrightable artifact: one or several images. But reading through this and the law firm letter included at the end, made me realize that the art object is actually above all of that. It exists as a canvas or hyper-canvas in latent space. It is “latent art” for lack of a better word, which relates to a kind of active engagement with and exploration of latent media and language spaces. And the actual end products generated during that process are very much secondary to the actual higher-dimensional form the artist is activating…
Let me drill back down into the letter for other examples to expand this hopefully more. This part is from the original lawyer letter which starts toward the end of the document, so this is the law firm asserting their legal theory:
The visual structure of each image, the selection of the poses and points of view, and the juxtaposition of the various visual elements within each picture were consciously chosen. These creative selections are similar to a photographer’s selection of a subject, a time of day, and the angle and framing of an image. In this aspect, Kashtanova’s process in using the Midjourney tool to create the images in the Work was essentially similar to the artistic process of
photographers – and, as detailed below, was more intensive and creative than the effort that goes into many photographs. Even a photographer’s most basic selection process has been found sufficient to make an image copyrightable.
Regarding this visual exploration process, the lawyer letter has a section on that, which I think starts to illustrate what a “hypercanvas” looks like. I’ll reproduce two pages from it here, for educational purposes and for encouragement of political debate, as a matter of Fair Use:
I’ll pick up the threads on that copyright letter another time, but the above is something to slow down and consider.
I took this idea of the hyper canvas, and the “art object” existing in higher dimensional space, and dropped it into both Claude & ChatGPT. Snippets from each that might help fill out our understanding of this concept:
The latent space that generative AI models create could be seen as a new type of artistic medium that artists work within. Just as a traditional painter works on the 2D canvas with paints, an AI artist navigates and creates on this high-dimensional latent canvas.Claude
The cultural impact of AI art comes from how artists embed the latent canvas explorations into specific artifacts, narratives, and meanings. So the latent canvas gets actualized in ways that speak to the human experience.CLAUDE
I thought it might make sense to explore also how this might link up to the concept of hyper-objects, which is something I’m only dimly aware of what it means, but seems related af.
Hyperobjects are phenomena that are massively distributed across space and time, challenging traditional ideas of locality and perception. Latent spaces created by AI could be seen as a type of computational hyperobject – vastly multidimensional spaces that human artists navigate.
Timothy Morton’s conception of hyperobjects emphasizes their nonlinear nature – how they don’t adhere to traditional chronological timelines. The iterative, feedback-driven process of making latent art also has a nonlinear temporality as artists move recursively through latent space.CLAUDE
I like this one in particular:
Works of latent art actualize slices or samples from the broader latent hyperobject, bringing something formless into tangible form… creating localized perceptible manifestations.Claude
The aesthetic qualities of any given latent artwork arise from the particular way it manifests and embodies some of the latent potential. Different artistic choices result in different local manifestations.CLAUDE
Don’t mind me, just dumping a bunch more quotes:
An artist’s body of latent artworks could be viewed holistically as an attempt to map out latent space through different localized perceptual snapshots. Each work provides a glimpse into the broader hyperobject.
Over time, the accumulation of localized manifestations by many artists engaging with the same latent space leads to a gradually increasing (but always partial) understanding of its overall nature.Claude
And let’s give ChatGPT a chance to pontificate here too:
For the “Latent Canvas,” one could conceptualize the art creation process in these steps:
- Iteration: The artist engages with the AI system to explore the latent space. The specific path taken is guided by the artist’s vision and reaction to each output from the AI. It’s a journey of exploration and discovery through a higher-dimensional space.
- Existence: The art piece itself isn’t a single output, but rather the series of transformations that occur within the latent space as a result of the interaction between the artist and the AI. The “canvas” exists in this higher dimensional latent space, and the artwork can be thought of as a trajectory within that space. This introduces a temporal element to the art, with the possibility of revisiting previous ‘states’ of the canvas.
- Representation: The challenge then becomes how to represent this higher-dimensional journey….
In the context of “Latent Canvas,” the hyperobject could be seen as the entirety of the latent space – a multi-dimensional space too vast for one to comprehend in its fullness, but whose characteristics and possibilities are expressed through each unique generation by the AI tool. Just like a hyperobject, the latent space is not directly perceivable. Instead, the artist and the viewers interact with it through localized expressions (the generated outputs), and the trajectory of iterations made by the artist.
In a sense, the “Latent Canvas” captures the essence of hyperobjects by presenting an artistic journey through a latent space that is high-dimensional, non-local, and only known through its particular manifestations.ChatGPT
Okay, I’ll stop there, cause that’s plenty for now.