All art is conceptual art.docx
It's only been a week since OpenAI released an update to their flagship language model - GPT-4o that can generate images. And already the world of creative possibilities - and possibly the future of art - has changed beyond all recognition. For those of us who had graphical ideas but no skill to bring them into visual existence, art has become conceptual art. With this technology we can now speak concepts into existence. Just like with photography, digital photo editing, smartphone cameras, desktop publishing and many others - technology has expanded the scope of creative possibilities.
As a mild dysgraphic (meaning I can't even draw a stick figure), I always considered drawing let alone painting a magical ability - kind of like flying. Even when I try to follow the steps of some fool-proof beginner guide to drawing, I end up with a mess. As a result, I focused my creative efforts into language (and to some degree music). Art was so far outside of my area of interest, I never even had any ideas about what I'd like to see created that is not there in a way that resonates with some sort of a vision I have. I've dabbled with photography but that does not seem to be in the same area for me.
But since last Tuesday, I've been frantically creating one image after another. My efforts definitely trend towards the humorous but they certainly enabled me to express something I never had any inkling I had a desire to express. Ideas and concepts in my head. For all intents and purposes I had a drawing disability and now technology gave me the tools to overcome it.
![[Writing/Writing on AI/_attachments/All art is conceptual art-Reflections on the aftershocks of the image revolution/ATTACH - {$originalname}-1.png]]
I have an idea, I express it the way I know how - through words - I get a result and then refine it by speaking through the technological tool at my disposal - ChatGPT. It is sort of like a conceptual Photoshop. Ideas matter more than physical dexterity - perhaps a motto for human progress. Learning is involved. But it is exploratory learning rather than skill learning (full of repetition) I definitely have to learn to use the tool and because it comes without a manual other than "Just tell me what you want", I have to learn about its range of possibilities - what it responds to and how - more like having to learn what an artist I'm commissioning to create something for me expects to hear. When it's worth telling it to change something and when it's better to start over.
This learning is a never ending process. And it is also collective learning - the Internet (or at least my feeds) has exploded with people exploring the space of creative possibilities. Copying, refining, vibing off each others' ideas - discovering new capabilities and pushing what can be done with them. And it has only been a week.
We've seen a similar churn of creative endeavour when Midjourney was first released in 2022 and later that year again when ChatGPT burst onto the scene. In a way, this is only an evolution of those early breakthroughs - image and text creation from ideas expressed in words - but this slight evolutionary shift has pushed us into a revolutionary moment - a temporary puncture in the equilibrium.
What enabled this shift were two technological features and one policy decision by OpenAI. The first feature is object persistence. This enables the creator to iterate on an idea after the first draft. Until now, image generation without persistence effectively meant that even a small tweak was tantamount to starting over. The object persistence even works on transferring objects from photos.
The second feature is accurate text. The early image models like Midjourney or DALL-E could not do text at all. Over the last three years, diffusion models have improved to the point where they can get a few words right but it is in a haphazard and frustrating way. The new more language-like approach to image generation that GPT-4o takes, enables almost any amount of text in any shape (with the occasional imperfection) which enables a whole new genre of both utilitarian art such as advertising or even educational content such as mindmaps but also the creation of comics and memes - that spread quickly.
The last enabling element was a policy decision or rather two policy decisions by OpenAI. The key decision was to allow imitation of style. The new model can take almost any recognisable style and make images in it. This makes it possible to communicate in more resonant ways. And what's more, it can not only imitate style, it can even imitate people's likeness (with some restrictions). You can ask for an image of a moderately notable person or upload a photo of someone and ask to include them in the image.
This has proved less controversial - although not zero-controversial - than one might have expected. Over the last week, the focus has been almost entirely on celebrating the joy of creation. When people asked for a picture of themselves in the style of studio Ghibli, the response was rarely "why don't you create something original" but more often "cool" - this is something that brings joy and resonates with me, too. That is because studio Ghibli's style is not a private possession - it is part of the world. ![[Writing/Writing on AI/_attachments/All art is conceptual art-Reflections on the aftershocks of the image revolution/ATTACH - {$originalname}-3.png]]
This goes to the very core of creativity: all creation is imitation. Studio Ghibli's style, after all, is not fully original (to somebody like me who has never seen the films it looks more like standard animation). Nor is every drawing created by a single creative genius. Animated films themselves are creations of teams imitating a style set by somebody with a conceptual vision. Genres work because of resonant imitation. Creativity and originality are the result of imperfection of this imitation - unique individual takes on existing ideas and forms - not completely novel, never seen before, creations. (Even if the language we use to describe these relatively slight deviations from the expected form makes it seem like they are so.)
There is a strand to art appreciation that values craft - the skill and effort required to generate the artifact. And as technology reshapes the landscape of what skill is required to produce the artifacts, people both claim that "it's not the same", "not as good", not of the "same value" and it will destroy the human element in art. What's worse, what will be the point of creation, when it's available just at the touch of the button? But if the history of art and technology is anything to go by, this has yet to happen. If anything, it has always been artists who push technological innovation forward rather than the technologists. And often the artist and the technologist are one and the same. This does not diminish art or reduce it to something less human - humanity has been part of the process the whole way through. Not does it diminish the underlying craft. Just because we have photography, people haven't stopped painting. Just because we have digital camera, people haven't stopped using film. ![[Writing/Writing on AI/_attachments/All art is conceptual art-Reflections on the aftershocks of the image revolution/ATTACH - {$originalname}-2.png]]
Of course, what changes is the role of the craft. In both the practical and commercial sense. This change will mean a gradual realignment of the commercial value on the utilitarian side of the craft. People who want to have paintings made or by original art will not switch to asking ChatGPT to make them for them. But people who want a simple design or a graphic that they may have paid for on 99designs or Fiverr, will be less likely to do so. It may even undermine the continued need for tools like Canva - assuming the technology will continue improving. It is too early to tell.