These images were created by an artist known as Botto, which exhibited at Sotheby's in New York this October and has made more than $4 million from sales of its work. In truth though, Botto needs only GPUs to get creative juices flowing.
Botto is a decentralized semi-autonomous artistic agent created in 2021 by the German artist Mario Klingemann; Simon Hudson, a media entrepreneur; and Ziv Epstein, a computer scientist and designer.
Botto contains an AI image generator similar to Dall-E or Midjourney but its output is also shaped by a “taste model” that selects the most pleasing images generated by a prompt. The taste model is tuned to reflect the preferences of a community of Botto enthusiasts who vote on the images Botto produces and posts online here. Botto is also governed by a decentralized autonomous organization, or DAO, meaning enthusiasts can buy $Botto cryptocurrency and influence how the system is managed and developed.
The recent Sotheby’s show is just the latest in a string of successful exhibitions for Botto. The October event alone racked up $350,000 in sales. Botto has made around $4 million in total since 2021, its creators say.
Klingemann and Hudson say those who control Botto through the associated DAO have chosen to add a modified version of Mistral’s biggest open source large language model and a knowledge base that allows it to converse about its artwork, and which will be further fine tuned through interactions with the Botto community. “Through this interaction and through various channels of input, its knowledge will grow and it will develop a personality and interests,” Klingemann suggests.
Klingemann and Hudson hope that this personality will even start to steer the art that Botto creates, perhaps allowing it to use an “unaligned” image generator—meaning one without the guardrails that would prevent it from producing racy or violent imagery—to see if it can develop its own sense of what is artistically acceptable. “Right now we give Botto safe models, but as it gets older you [could] give it things that require greater maturity,” Hudson says, comparing Botto’s maturation to that of a person in human society.
It’s an interesting idea, and it is fun to see the idea of an AI agent explored within the relatively benign realm of artistic expression.
That said, Botto still poses some ethical conundrums. Many working artists rightly worry about the impact AI is having on their profession, as models trained on millions of copyrighted works are used to generate infinite knock-offs on demand.
Perhaps Botto is something altogether different. Klingemann is an early adopter of AI in art, using neural networks as part of the artistic process, and as a kind of performance schtick. His previous creations include a video installation featuring ever-changing AI-generated portraits and a robot dog that poops critiques of visual artworks.
And while Botto generates high-priced images using a model trained on public work, Klingermann does not see this as outright plagiarism. “Image models and LLMs are the new search engines,” he says. “For me, creativity is kind of finding something that already exists in possibility-space, and deciding this is interesting, while making sure it looks [like it] doesn't belong to anybody already.”
The images made by Botto seem aesthetically pleasing but also feel—to my untrained eye, at least—like fairly generic AI image generator offerings.
While the Botto project poses some interesting questions about what constitutes artistic agency, for now I think it only emphasizes the importance of human intelligence and inventiveness. The spark of creativity belongs not to the machine that churns out a never-ending variety of images with feedback from the crowd, but to the artists who came up with the idea in the first place.
What do you think of Botto and its artwork? Is it a worthwhile artistic idea or just another way to make money from generative AI and meme coins? Send a message to hello@wired.com or leave a comment below to let me know.
