by Manu SharmaSep 27, 2022
If you have ever read Sapiens by Yuval Noah Harari, you might remember the part where he talks about artificial intelligence taking over all our jobs. Harari makes the argument that in time, as the sophistication of the technology increases, AI could even take on the role of musicians and artists. The Israeli historian points out that musicians may face this threat before visual artists, as music can be mathematically mapped and therefore easily derived into a smart algorithm. However, I recently stumbled across a project which points out to me that visual art is not any less vulnerable to the looming impact of AI. 'Botto' is an artist, but 'Botto' is also an algorithm. 'Botto' is an autonomous entity of its own, however 'Botto' is also a community-driven, decentralised creator… Are you curious and confused at the same time? So was I, until I caught up with AI artist Mario Klingemann who conceptualised 'Botto', a generative and decentralised art hybrid, and conceived it with the support of leading engineers.
Klingemann’s artistic exploration of artificial intelligence was sparked at the early age of 20 when he read The Society of Mind by Mavin Minsky in 1990. The artist says, “His ideas on how a machine could become intelligent fascinated me, but at that time they were all of a purely theoretical nature. It took me more than 15 years until I could start employing machine learning methods in my art practice. My first bigger project in which a machine tried to evolve art and used learning methods to improve the process was Sketchmaker which I wrote in 2006.” Klingemann went on to work with code, neural networks and algorithms in his creative practice, and in many ways spearheading the AI art movement.
The German artist tells us how 'Botto' was brought to life. “The underlying idea for 'Botto' was one that I had been carrying around with me since a very long time - once you start writing generative art algorithms, the thought that the whole process could be automated and become self-sufficient seems obvious and I am surely not the only person who had that idea. The difficulty with a simple idea as this is the execution and being able to make it actually happen,” he says. This simple thought was in fact also the first to strike my mind when I first heard of 'Botto', and the one to eventually lead me to the understanding that while 'Botto' is very much a live project, its ultimate form will be when it transitions fully from algorithm to artist, community-driven creator to autonomous entity.
Klingemann continues, “Before the arrival of the blockchain and the crypto art market, a project like this would probably have been doable, but it would likely only have been a kind of conceptual piece as you see them on art festivals - looking great in the white cube but not really viable outside in the ‘real world’. What I wanted was the real thing and for that it needed several elements: AI art models that are able to show a lot of versatility as well as have aesthetic qualities, a market to sell the art and some patrons who were willing to take the risk of financing the development of this project. I was very lucky to get to the point where all of these came together.”
'Botto' today is both a conceptual wonder and also a very practical litmus test to understand the art market as well as popular aesthetic inclinations. Klingemann shares his recognition of the disparity between his own preferences in visual arts language, versus that of the community which drives 'Botto'. He says, “The community has definitely a quite different taste than myself - at the moment it appears that most popular are those fragments that remind in their style of old paintings, but we also already saw a few times in the weekly leaderboard a clear divide among the voter community between those who prefer the calmer and more familiar motifs and those who like the more grotesque or surrealist style. So far I would say that apart from a few exceptions the ‘safer’ choices have made it to the auction.” As an onlooker from afar, neither the creator nor really the audience, I wonder what intangible insights 'Botto' might be able to provide us with when examining a diverse dataset of communities.
Klingemann gets down to brass tacks, explaining the delicate balancing act that is played out by the algorithmic, artistic entity. He tells us step by step, “'Botto' currently creates around 300 fragments every day by exploring ‘prompt space’ - in practice this means that it first generates a random sentence from which an image generator that is driven my OpenAI's CLIP model makes an image that tries to represent the meaning of the sentence or words. In this phase 'Botto' does not try to force certain outcomes but rather to cover as many variations as possible. From the constantly growing pool of generated fragments 'Botto' selects 350 fragments to be presented to the community so that they should be voted upon. This selection process is controlled by 'Botto’s ‘taste’ model. This model tries to learn to reproduce the voting behaviour of the community, and by this predict which of the newly generated fragments might be popular and successful in the vote. At the same time, it also classifies its own creations by originality and mixes the possibly popular with other fragments that look different to what the audience has seen so far. That way I hope that 'Botto' will not fall too quickly into becoming too repetitive or just trying to please.” The community which drives 'Botto' is made up of 5,000 people, and continues to grow.
Klingemann continues to look after the behaviour of 'Botto', much like a parent to their child. He tells us, “Right now 'Botto' is like a toddler and I am its guardian that has to guide its first steps and make sure it does not hurt itself. Over time it will gain new capabilities - initially that will likely be novel or improved generative models and classifiers, but further on it will start communicating and maybe even try out other art forms, like music or literature. Also, whenever technological progress will allow to give 'Botto' more autonomy in making its own responsible decisions it will be given more freedom - but of course only after sufficient testing until the team and I can trust it with those powers.” Rest assured, we are not about to witness a world takeover by an artistic AI.
So far, five artworks as NFT by 'Botto' have been auctioned on SuperRare for up to $325,00 (almost 80 Ethereum coins), making over a million dollars from its first five auctions. Apart from his artistic practice, Klingemann also co-curates a space in Munich with Alexandra Lukaschewitz, called Dog & Pony.