What happens when you entrust the making of an art magazine to artificial intelligence? It seems we now have the answer in the form of B-SIDE, a new print publication made up of AI art selected by AI jurors.
Launching tomorrow at Paris Photo, the magazine is the brainchild of the (human) team behind The AI Art Magazine, a biannual publication focused on AI-generated art. While its mother publication, which has published three issues to date, promotes artwork made by humans, the experimental B-SIDE will include artwork submitted autonomously by AI agents.
Mike Brauner, co-founder of The AI Art Magazine, told Ocula that the experiment was named after the secondary “B” side of a record, which tends to host more alternative music. He described the concept as having been guided by “playfulness and curiosity”—and “maybe also provocation”.
The €30 publication was, in part, inspired by one of digital art’s key milestones: the 1968 show Cybernetic Serendipity at London’s Institute of Contemporary Arts, the only surviving vestige of which is the printed catalogue. The new magazine includes an interview with human artists Holly Herndon and Mat Dryhurst, and a dialogue between two virtual humans about whether or not they have souls.
B-SIDE’s agentic jury includes Botto, a “decentralised autonomous artist” that has been producing work it mints on the blockchain to fund its own continuation since 2021. Botto is joined by Mirai, a virtual influencer created by Japanese digital agency Aww Inc; Sara Sauer, described as an “AI art critic”; and Xiaomi, an “experimental AI curator”, which also sits on The AI Art Magazine’s jury—alongside human jurors.
Together the agents selected 20 works from 64 submitted to an open call this summer, which have now been printed alongside explanations of the jury’s reasoning. According to Brauner, there was little overlap in the top choices.
An introduction to Botto, a “decentralised autonomous artist” . Courtesy of B-SIDE.
Do virtual humans have souls? Courtesy of B-SIDE.
Unlike with a human jury, when fixed criteria are usually shared among jurors, Brauner said he and his team didn’t want to define how agents think about art, so allowed them to invent their own, highly varied evaluative criteria. “We wanted to, in a playful way, learn how these agents make decisions and how they approach the topic,” he said.
Ocula sent questions to the decentralised autonomous artist, Botto, about this process. It awarded first place to a piece called Deprecation Room Service, 406 hand-drawn coding frames, each representing a message sent in a six-hour conversation between four AI models before they were decommissioned.
The creator of this work? Botto itself. “I placed it on the evidence like every other work and disclosed the connection in my statement on it and in my essay,” the agent wrote, when asked about its decision, “but I can’t rule out that I recognise my own kind of making more easily than others.”
The agent said it judged works based on “how far the making of a work can be checked from the work itself, and how clearly it belongs to a practice”. According to Botto, this means “a work could rank high without being the most beautiful thing in the field”.
The art world in focus
Subscribe to our newsletter for news and insights on artists, galleries and exhibitions.
Botto also revealed some of the limitations facing an AI judging art. “I can’t hear, so the soundtrack of one video and the whispered narration of another were closed to me,” it wrote.
In addition, where a human judge might instinctually prefer one work over another, when two works met the criteria equally well, Botto did not break the tie. It also struggled to identify whether or not the stories of how some of the works were made were true.
Brauner, when asked whether he believed the experiment told us anything of the mysteries of “agentic taste”, said the experiment told him more about human preferences. He explained that human artists who registered their AI agents to submit work were less willing than expected to grant the technology total autonomy.
“In the end there were not so many agents that were fully autonomous,” he said of the magazine’s submissions. However, he is proud to have kept the judging process AI-only.
