Manifestes 5: Design Under Artifice

At the beginning of the 2020s, so-called ‘intelligent’ systems such as GPT-3, OpenAI, Dall-E, Midjourney or Disco Diffusion made it possible to generate images thanks to textual commands (‘prompts’) associated with large data sets available online. As part of machine learning technologies, these systems found their place in art and the ‘creative industries’ (fashion design, graphic design, product design, architecture, etc. ), profoundly reconfiguring these professions in the same way desktop publishing and design (DTP and CAD) had in the 1980s. Although these productions prove rather easily stereotyped – merely remixing existing content – debates have focused on the possible replacement of designers by artificial intelligence (AI), shielding the essential question: what is the spectrum of risk and opportunity when it comes to machine learning for design practices?In order to better understand how these AIs participate in what we propose to call a ‘design under artifice’ – that is to say, an insidious subversion of design’s historical principles under the influence of cognitive-behavioural approaches – it is first necessary to establish a more detailed understanding of the psychological theories specific to machine learning. Participating in a neurocognitivist approach that assimilates the human psyche to a circuit switch, machine learning is part of the (already) long history of creative software, which aims both to democratise access to computers and to standardise creative practices. Just as these programs have automated a certain number of tasks usually assigned to designers, machine learning technologies displace and redefine the notions of creation and subjectivity. Although they run the risk of homogenising the sensitive world, they also open up new forms of cooperation with machines, of which the contemporary design studios interviewed for this essay offer a glimpse. In this book, Anthony Masure reconnects the challenges of artificial intelligence and design, while proposing a series of avenues to guide this ideal of automation on a small scale, in a controlled and ''tailor-made'' way, so that machine learning can foster invention and curiosity.