![]() The more technical challenge when adding more diversity of representation to the list was whether the artists would have enough data in the dataset for Midjourney to produce anything resembling both the artist and their work. So solving for diversity creates a deeper issue of appropriation. So while I added names of artists whose work I could easily remember and love that list skewed white and male.īut when I tried to diversify that (human made) list, a seemingly worthy endeavour, it adds people to the list from cultures who have been historically underrepresented, which is good, but also where there has been great cultural appropriation, which is bad in this context. An embedded issue within the larger issue of appropriation is that what we understand as 'recognizable' is a product of a system and cultural context that has been historically dominated by white men. But lesser known artists also lead to fairly 'accurate' results which speaks to the vast depth of the data in use. The strange tension is that more 'notable' artists return more 'accurate' results because there is more data to scrape. The first list was established by me based on my own limited knowledge of the art world but with a particular attempt to find a list that balanced notability with diversity. The second task was to prompt Midjourney for each artist. The first task was to establish the list. ![]() This experiment explores this question with well established artists to raise awareness of the problem. But there is still limited understanding of the issue in the general public and by artists themselves. It's a fine line, especially given that many artists create self-portraits as core to their bodies of work.Ĭlass-action lawsuits have started that rightly claim that these products "infringe the rights of artists and other creative individuals under the guise of 'artificial intelligence'". But, on the other hand, it's an attempt to draw attention to the issue of appropriation without engaging in the act itself. The idea of creating portraits of the artists rather than just images in the style of the artists was, on the one hand, a way of clearly associating the individual artist with their style. The platform was simply prompted based on artist names and generated images from data it already had. To be clear, this experiment did not add anything to Midjourney's data. The experiment here is: What if we ask Midjourney to create portraits of artists in the style of those artists? What's to stop others (like me) from simply generating 'new' works in the style of artists without asking their consent? That information includes imagery from artists who did not grant their permission. That means programs like Midjourney use vast amounts of information to produce imagery. Within the creative industry there is much debate about the ethics of creating new work based on the work of current artists. We're hearing a lot about generative AI and how it draws on existing data pools to create.
0 Comments
Leave a Reply. |