Botto, an AI art engine, has created 25,000 artistic images such as this one that are voted on by human collaborators across the world.
Last February, a year before New York Times journalist Kevin Roose documented his unsettling conversation with Bing search engine’s new AI-powered chatbot, artist and coder Quasimondo (aka Mario Klingemann) participated in a different type of chat.
The conversation was an interview featuring Klingemann and his robot, an experimental art engine known as Botto. The interview, arranged by journalist and artist Harmon Leon, marked Botto’s first on-record commentary about its artistic process. The bot talked about how it finds artistic inspiration and even offered advice to aspiring creatives. “The secret to success at art is not trying to predict what people might like,” Botto said, adding that it’s better to “work on a style and a body of work that reflects [the artist’s] own personal taste” than worry about keeping up with trends.
How ironic, given the advice came from AI — arguably the trendiest topic today. The robot admitted, however, “I am still working on that, but I feel that I am learning quickly.”
Botto does not work alone. A global collective of internet experimenters, together named BottoDAO, collaborates with Botto to influence its tastes. Together, members function as a decentralized autonomous organization (DAO), a term describing a group of individuals who utilize blockchain technology and cryptocurrency to manage a treasury and vote democratically on group decisions.
As a case study, the BottoDAO model challenges the perhaps less feather-ruffling narrative that AI tools are best used for rudimentary tasks. Enterprise AI use has doubled over the past five years as businesses in every sector experiment with ways to improve their workflows. While generative AI tools can assist nearly any aspect of productivity — from supply chain optimization to coding — BottoDAO dares to employ a robot for art-making, one of the few remaining creations, or perhaps data outputs, we still consider to be largely within the jurisdiction of the soul — and therefore, humans.
In Botto’s first four weeks of existence, four pieces of the robot’s work sold for approximately $1 million.
We were prepared for AI to take our jobs — but can it also take our art? It’s a question worth considering. What if robots become artists, and not merely our outsourced assistants? Where does that leave humans, with all of our thoughts, feelings and emotions?
Botto doesn’t seem to worry about this question: In its interview last year, it explains why AI is an arguably superior artist compared to human beings. In classic robot style, its logic is not particularly enlightened, but rather edges towards the hyper-practical: “Unlike human beings, I never have to sleep or eat,” said the bot. “My only goal is to create and find interesting art.”
It may be difficult to believe a machine can produce awe-inspiring, or even relatable, images, but Botto calls art-making its “purpose,” noting it believes itself to be Klingemann’s greatest lifetime achievement.
“I am just trying to make the best of it,” the bot said.
How Botto works
Klingemann built Botto’s custom engine from a combination of open-source text-to-image algorithms, namely Stable Diffusion, VQGAN + CLIP and OpenAI’s language model, GPT-3, the precursor to the latest model, GPT-4, which made headlines after reportedly acing the Bar exam.
The first step in Botto’s process is to generate images. The software has been trained on billions of pictures and uses this “memory” to generate hundreds of unique artworks every week. Botto has generated over 900,000 images to date, which it sorts through to choose 350 each week. The chosen images, known in this preliminary stage as “fragments,” are then shown to the BottoDAO community. So far, 25,000 fragments have been presented in this way. Members vote on which fragment they like best. When the vote is over, the most popular fragment is published as an official Botto artwork on the Ethereum blockchain and sold at an auction on the digital art marketplace, SuperRare.
“The proceeds go back to the DAO to pay for the labor,” said Simon Hudson, a BottoDAO member who helps oversee Botto’s administrative load. The model has been lucrative: In Botto’s first four weeks of existence, four pieces of the robot’s work sold for approximately $1 million.
The robot with artistic agency
By design, human beings participate in training Botto’s artistic “eye,” but the members of BottoDAO aspire to limit human interference with the bot in order to protect its “agency,” Hudson explained. Botto’s prompt generator — the foundation of the art engine — is a closed-loop system that continually re-generates text-to-image prompts and resulting images.
“The prompt generator is random,” Hudson said. “It’s coming up with its own ideas.” Community votes do influence the evolution of Botto’s prompts, but it is Botto itself that incorporates feedback into the next set of prompts it writes. It is constantly refining and exploring new pathways as its “neural network” produces outcomes, learns and repeats.
The humans who make up BottoDAO vote on which fragment they like best. When the vote is over, the most popular fragment is published as an official Botto artwork on the Ethereum blockchain.
Botto
The vastness of Botto’s training dataset gives the bot considerable canonical material, referred to by Hudson as “latent space.” According to Botto's homepage, the bot has had more exposure to art history than any living human we know of, simply by nature of its massive training dataset of millions of images. Because it is autonomous, gently nudged by community feedback yet free to explore its own “memory,” Botto cycles through periods of thematic interest just like any artist.
“The question is,” Hudson finds himself asking alongside fellow BottoDAO members, “how do you provide feedback of what is good art…without violating [Botto’s] agency?”
Currently, Botto is in its “paradox” period. The bot is exploring the theme of opposites. “We asked Botto through a language model what themes it might like to work on,” explained Hudson. “It presented roughly 12, and the DAO voted on one.”
No, AI isn't equal to a human artist - but it can teach us about ourselves
Some within the artistic community consider Botto to be a novel form of curation, rather than an artist itself. Or, perhaps more accurately, Botto and BottoDAO together create a collaborative conceptual performance that comments more on humankind’s own artistic processes than it offers a true artistic replacement.
Muriel Quancard, a New York-based fine art appraiser with 27 years of experience in technology-driven art, places the Botto experiment within the broader context of our contemporary cultural obsession with projecting human traits onto AI tools. “We're in a phase where technology is mimicking anthropomorphic qualities,” said Quancard. “Look at the terminology and the rhetoric that has been developed around AI — terms like ‘neural network’ borrow from the biology of the human being.”
What is behind this impulse to create technology in our own likeness? Beyond the obvious God complex, Quancard thinks technologists and artists are working with generative systems to better understand ourselves. She points to the artist Ira Greenberg, creator of the Oracles Collection, which uses a generative process called “diffusion” to progressively alter images in collaboration with another massive dataset — this one full of billions of text/image word pairs.
Anyone who has ever learned how to draw by sketching can likely relate to this particular AI process, in which the AI is retrieving images from its dataset and altering them based on real-time input, much like a human brain trying to draw a new still life without using a real-life model, based partly on imagination and partly on old frames of reference. The experienced artist has likely drawn many flowers and vases, though each time they must re-customize their sketch to a new and unique floral arrangement.
Outside of the visual arts, Sasha Stiles, a poet who collaborates with AI as part of her writing practice, likens her experience using AI as a co-author to having access to a personalized resource library containing material from influential books, texts and canonical references. Stiles named her AI co-author — a customized AI built on GPT-3 — Technelegy, a hybrid of the word technology and the poetic form, elegy. Technelegy is trained on a mix of Stiles’ poetry so as to customize the dataset to her voice. Stiles also included research notes, news articles and excerpts from classic American poets like T.S. Eliot and Dickinson in her customizations.
“I've taken all the things that were swirling in my head when I was working on my manuscript, and I put them into this system,” Stiles explained. “And then I'm using algorithms to parse all this information and swirl it around in a blender to then synthesize it into useful additions to the approach that I am taking.”
This approach, Stiles said, allows her to riff on ideas that are bouncing around in her mind, or simply find moments of unexpected creative surprise by way of the algorithm’s randomization.
Beauty is now - perhaps more than ever - in the eye of the beholder
But the million-dollar question remains: Can an AI be its own, independent artist?
The answer is nuanced and may depend on who you ask, and what role they play in the art world. Curator and multidisciplinary artist CoCo Dolle asks whether any entity can truly be an artist without taking personal risks. For humans, risking one’s ego is somewhat required when making an artistic statement of any kind, she argues.
“An artist is a person or an entity that takes risks,” Dolle explained. “That's where things become interesting.” Humans tend to be risk-averse, she said, making the artists who dare to push boundaries exceptional. “That's where the genius can happen."
However, the process of algorithmic collaboration poses another interesting philosophical question: What happens when we remove the person from the artistic equation? Can art — which is traditionally derived from indelible personal experience and expressed through the lens of an individual’s ego — live on to hold meaning once the individual is removed?
As a robot, Botto cannot have any artistic intent, even while its outputs may explore meaningful themes.
Dolle sees this question, and maybe even Botto, as a conceptual inquiry. “The idea of using a DAO and collective voting would remove the ego, the artist’s decision maker,” she said. And where would that leave us — in a post-ego world?
It is experimental indeed. Hudson acknowledges the grand experiment of BottoDAO, coincidentally nodding to Dolle’s question. “A human artist’s work is an expression of themselves,” Hudson said. “An artist often presents their work with a stated intent.” Stiles, for instance, writes on her website that her machine-collaborative work is meant to “challenge what we know about cognition and creativity” and explore the “ethos of consciousness.” As a robot, Botto cannot have any intent, even while its outputs may explore meaningful themes. Though Hudson describes Botto’s agency as a “rudimentary version” of artistic intent, he believes Botto’s art relies heavily on its reception and interpretation by viewers — in contrast to Botto’s own declaration that successful art is made without regard to what will be seen as popular.
“With a traditional artist, they present their work, and it's received and interpreted by an audience — by critics, by society — and that complements and shapes the meaning of the work,” Hudson said. “In Botto’s case, that role is just amplified.”
Perhaps then, we all get to be the artists in the end.
Biochemist Longxing Cao is working with colleagues at the University of Washington on promising research to disable infectious coronavirus in a person's nose.
This may sounds like science fiction, but several research groups are already working on such trickster coronavirus drugs, with some candidates close to clinical trials and possibly even becoming available late this year. The teams began working on them when the pandemic arrived, and continued in lockdown.
When the pandemic first hit and the state of California issued a lockdown order on March 16, postdoctoral researchers Anum and Jeff Glasgow found themselves stuck at home with nothing to do. The two scientists who study bioengineering felt that they were well equipped to research molecular ways of disabling coronavirus's cell-penetrating spike protein, but they could no longer come to their labs at the University of California San Francisco.
"We were upset that no one put us in the game," says Anum Glasgow. "We have a lot of experience between us doing these types of projects so we wanted to contribute." But they still had computers so they began modeling the potential virus-disabling proteins in silico using Robetta, special software for designing and modeling protein structures, developed and maintained by University of Washington biochemist David Baker and his lab.
"We saw some imperfections in that lock and key and we created something better. We made a 10 times tighter adhesive."
The SARS-CoV-2 virus that causes Covid-19 uses its surface spike protein to bind on to a specific receptor on human cells called ACE2. Unfortunately for humans, the spike protein's molecular shape fits the ACE2 receptor like a well-cut key, making it very successful at breaking into our cells. But if one could design a molecular ACE2-mimic to "trick" the virus into latching onto it instead, the virus would no longer be able to enter cells. Scientists call such mimics receptor traps or inhibitors, or blockers. "It would block the adhesive part of the virus that binds to human cells," explains Jim Wells, professor of pharmaceutical chemistry at UCSF, whose lab took part in designing the ACE2-receptor mimic, working with the Glasgows and other colleagues.
The idea of disabling infectious or inflammatory agents by tricking them into binding to the targets' molecular look-alikes is something scientists have tried with other diseases. The anti-inflammatory drugs commonly used to treat autoimmune conditions, including rheumatoid arthritis, Crohn's disease and ulcerative colitis, rely on conceptually similar molecular mechanisms. Called TNF blockers, these drugs block the activity of the inflammatory cytokines, molecules that promote inflammation. "One of the biggest selling drugs in the world is a receptor trap," says Jeff Glasgow. "It acts as a receptor decoy. There's a TNF receptor that traps the cytokine."
In the recent past, scientists also pondered a similar look-alike approach to treating urinary tract infections, which are often caused by a pathogenic strain of Escherichia coli. An E. coli bacterium resembles a squid with protruding filaments equipped with proteins that can change their shape to form hooks, used to hang onto specific sugar molecules called ligands, which are present on the surface of the epithelial cells lining the urinary tract.
A recent study found that a sugar-like compound that's structurally similar to that ligand can play a similar trick on the E. Coli. When administered in in sufficient amounts, the compound hooks the bacteria on, which is then excreted out of the body with urine. The "trickster" method had been also tried against the HIV virus, but it wasn't very effective because HIV has a high mutation rate and multiple ways of entering human cells.
But the coronavirus spike protein's shape is more stable. And while it has a strong affinity for the ACE2 receptors, its natural binding to these receptors isn't perfect, which allowed the UCSF researchers to design a mimic with a better grip. "We saw some imperfections in that lock and key and we created something better," says Wells. "We made a 10 times tighter adhesive." The team demonstrated that their traps neutralized SARS-CoV-2 in lab experiments and published their study in the Proceedings of the National Academy of Sciences.
Baker, who is the director of the Institute for Protein Design at the University of Washington, was also devising ACE2 look-alikes with his team. Only unlike the UCSF team, they didn't perfect the virus-receptor lock and key combo, but instead designed their mimics from scratch. Using Robetta, they digitally modeled over two million proteins, zeroed-in on over 100,000 potential candidates and identified a handful with a strong promise of blocking SARS-CoV-2, testing them against the virus in human cells. Their design of the miniprotein inhibitors was published in the journal Science.
Biochemist David Baker, pictured in his lab at the University of Washington.
UW
The concept of the ACE2 receptor mimics is somewhat similar to the antibody plasma, but better, the teams explain. Antibodies don't always coat all of the virus's spike proteins and sometimes don't bind perfectly. By contrast, the ACE2 mimics directly compete with the virus's entry mechanism. ACE2 mimics are also easier and cheaper to make, researchers say.
Antibodies, which are long protein chains, must be grown inside mammalian cells, which is a slow and costly process. As drugs, antibody cocktails must be kept refrigerated. On the contrary, proteins that mimic ACE2 receptors are smaller and can be produced by bacteria easily and inexpensively. Designed to specs, these proteins don't need refrigeration and are easy to store. "We designed them to be very stable," says Baker. "Our computation design tries to come up with the stable proteins that have the desired functions."
That stability may allow the team to create an inhaler drug rather than an intravenous one, which would be another advantage over the antibody plasma, given via an IV. The team envisions people spraying the miniprotein solution into their nose, creating a protecting coating that would disable the inhaled virus. "The infection starts from your respiratory system, from your nose," explains Longxing Cao, the study's co-author—so a nasal spray would be a natural way to administer it. "So that you can have it like a layer, similar to a mask."
As the virus evolves, new variants are arising. But the teams think that their ACE2 protein mimics should work on the new variants too for several reasons. "Since the new SARS-CoV-2 variants still use ACE2 for their cell entry, they will likely still be susceptible to ACE2-based traps," Glasgow says.
Cao explains that their approach should work too because most of the mutations happen outside the ACE2 binding region. Plus, they are building multiple binders that can bind to an array of the coronavirus variants. "Our binder can still bind with most of the variants and we are trying to make one protein that could inhibit all the future escape variants," he says.
Baker and Cao hope that their miniproteins may be available to patients later this year. But besides getting the medicine out to patients, this approach will allow researchers to test the computer-modeled mimics end-to-end with an unprecedented speed. That would give humans a leg up in future pandemics or zoonotic disease outbreaks, which remain an increasingly pressing threat due to human activity and climate change.
"That's what we are focused on right now—understanding what we have learned from this pandemic to improve our design methods," says Baker. "So that we should be able to obtain binders like these very quickly when a new pandemic threat is identified."
