Art, AI and Pollock’s denial of accident (Part I)
Jackson Pollock (1912-1956), was an American painter and key figure in the notorious abstract expressionist movement. He is renowned for his drip painting technique that produced works that, at first glance, might appear chaotic and unintentional. Pollock’s process however wasn't at all random, quite the opposite.
Pollock worked in a highly controlled manner, carefully manipulating the flow of paint, his body movement and the interaction between paint and the canvas. He famously said, "I don't use the accident. I deny the accident,". It's a statement that underscores his belief that true artistry comes from mastery and intentionality, not chance.
Full Fathom Five, 1947 by Jackson Pollock
What does he mean by 'denial of accident'? It's the idea that as an artist he was the sovereign creator of his pieces, fully in command of every aspect of his work. In that sense every drip an every splash came from him and not through him by an outside force. For Pollock, there was only the appearance of randomness that masked the depth of his control. In traditional views of artistry, the value of a work is often linked to the artist’s ability to exercise control and demonstrate skill.
So we can say that for Pollock, denial of the accident was a way to assert his authority over his medium. So art could not be left to fate or randomness. Every splatter, splash and drip was a deliberate act.
The inevitability of accident in AI-generated Art
In contrast to Pollock’s view, AI-generated art is inherently tied to the concept of accident. AI models, particularly those used for generative tasks operate on principles of probability and randomness. These models, such as GANs (Generative Adversarial Networks), start with a random noise vector and use it to generate outputs. This called Random Initialization and it's not the only factor.
The probabilistic nature of AI models means they operate on probability distributions to determine what comes next in a sequence (for instance the placement of pixels, colors, or forms in a generated image). Probabilistic decisions introduce variability, which can result in elements that are not entirely predictable or intentional.
The complex interactions between different layers and components of the AI model can lead to emergent behaviors that are difficult or impossible to foresee. This is a well known charateristic of generative AI - even with the same input, slight variations in processing lead to different results.
Finally there's our own human perception of accident. The result of an "accident" or unexpected outcome within the model's operations might be interpreted by human viewers as meaningful or intentional, even if they were not planned.
Jackson Pollock, photographed by Hans Namuth.
Randomness is not just a side effect of generatve AI models, it’s a core feature of how they operate. The probabilistic nature of AI models introduces a form of "accident" that is intrinsic to the medium. The AI doesn’t "intend" to create these accidents—they are a byproduct of its computational processes
An artist working with AI can try guiding the process by choosing inputs, adjusting parameters, and curating outputs, but they cannot fully control the final result in the same way Pollock could.
But an argument can be made for the artist wanting to keep and embrace or not certain accidents. This happens in the artistic process all the time, right?
Unfortunately, those choices are not possible for someone using AI to generated art because you fundamentally cannot know what is accident and what is not.
In part 2 (read it here), I will explore the role of ‘intentionality’ and the impact of this key element for art, in the context of generative AI.
