DOC • The algorithmic atelier

Articles<br>Series<br>Search

About Search

DOC<br>Content<br>Articles<br>Series<br>Search

DOC<br>About<br>Contact<br>RSS

Community<br>Instagram<br>Twitter<br>LinkedIn

UX Collective<br>Newsletter<br>Medium<br>Essays

Typography<br>Neue Haas Grotesk<br>Druk Medium<br>Times New Roman

The algorithmic atelier<br>Generative AI and the long history of artistic automation

Written by<br>Joshua Leigh

Image generated with Google Nano Banana

The discourse surrounding generative AI in the creative arts is frequently characterised by a sense of historical rupture, a seismic shift unlike any that has come before. Critics often frame the emergence of Large Language Models and diffusion-based image generators as an unprecedented existential threat to the ‘soul’ of human expression, a black swan event that may signal the end of human creative sovereignty. However, when we look back over the history of Western art and technology, it becomes clearer that this anxiety is not a new phenomenon but rather a recurring structural feature of the relationship between human intent and technical innovation. Whenever a new tool emerges that threatens to lower the barrier to entry or automate a process previously governed by arduous manual labour, the established artistic class has responded with a mixture of moral panic, accusations of ‘soullessness,’ and a retreat into essentialist definitions of what constitutes ‘true’ art.<br>To understand this moment, we must contextualise AI within the history of artistic tools and conceptual shifts, while simultaneously confronting the very real socio-economic and ethical challenges it poses to the creative ecosystem.

Deconstructing the toupee fallacy<br>Early criticism of generative AI focused on the poor quality of its output. GenAI was not yet a threat because it could not compete with human creative output on quality or believability.<br>The visceral reaction against early GenAI perhaps stemmed from a hypersensitivity to anomalies that break the illusion of human authorship. The ‘uncanny valley’ describes a revulsion response to humanoid representations that are almost, but not quite, human — a phenomenon observed in robots, animatronics, and increasingly, AI-generated imagery. When an AI image displays six fingers or background incongruities, the viewer immediately labels it as ‘artificial’ and files it away as evidence that AI is easily detectable and low-quality.

The Will Smith Eating Spaghetti test is an informal benchmark within the artificial intelligence community, used to assess the capabilities of generative video models in rendering realistic human actions and facial expressions. The quality has improved vastly from the 2023 original.

This creates a false impression of the technology’s overall capabilities through several reinforcing psychological biases. The Toupee Fallacy ensures that only the most noticeable (and typically worst) examples are identified, leading to the belief that AI is inherently detectable and of low aesthetic value. Flawless AI-generated images pass by unnoticed and unquestioned, creating a survivorship bias where only the failures are used to judge the entire category. This mental machinery is further fueled by the uncanny valley, where minor glitches like extra fingers are perceived as deep moral or spiritual failures rather than technical ones. Finally, anchoring bias causes us to fixate on these first ‘bad’ impressions, leading to a resistance to recognizing that these models improve exponentially over time.<br>This desire for authenticity can be observed in the ‘Invisible CGI’ phenomenon in the film industry. To guard against a backlash toward CGI in cinema, studios frequently market films by claiming “No CGI was used,” despite employing thousands of digital artists to composite scenes, remove safety rigs, and enhance environments. This practice was the subject of a video essay by Jonas Ussing titled “NO CGI” is really just INVISIBLE CGI.<br>When the CGI is ‘invisible,’ the public credits the ‘practical’ mastery of the director; when it is ‘bad,’ the technology is blamed. Direct experience becomes an unreliable guide to overall quality, as the viewer samples from the bottom of the distribution while remaining oblivious to the top performers.

The mechanization of vision: from Vermeer to the silicon brush<br>The most persistent argument against GenAI is that it removes the skill of the artist, allowing an amateur to produce high-quality work without the arduous training traditionally associated with mastery. Yet, the history of the Old Masters suggests that ‘skill’ has always been a fluid concept, frequently augmented by hidden technical aids. The Hockney-Falco thesis, proposed by artist David Hockney and physicist Charles Falco, argues that the dramatic rise in naturalism in Western art from the early Renaissance was not due solely to improved talent, but to the widespread use of optical devices such as the camera obscura, camera lucida, and curved mirrors.

Illustration of a “portable”...