After a rather scrumptious meal of Siu Mai and Gyoza dimsums at Dumpling Time in San Francisco, we take a walk around the Warriors Chase Center square, home to Golden State Warriors, the beloved San Franciscan basketball team, discussing when we will be able to see the iconic basketball player Stephen Curry play live.
It’s a cloudy Saturday afternoon in February. I’ve just finished an intense interview and want to wander and breathe in the city. But a sudden shower ruins the mood. We find shelter in a glass building across from Chase Center. And as serendipity and San Francisco would have it, find ourselves in Misalignment Museum, with art pieces that use AI models to make nerdy, artistic points.
I stop at the entrance on the left, where on a table, rows upon rows of recycled tins of canned meat sit together in the artwork Spambots by Neil Mendoza, who uses software, electronics, and robotics to create surreal and wacky art. Each can have two tiny arms and controls four keys of the keyboard. Run on a Raspberry Pi, the Spambots coordinately click and clack, generating a novel on the screen in front of them. Not just any novel. It’s a work plagiarised — or perhaps rephrased or reimagined is a proper term for AI — by a large language model trained on author Aldous Huxley’s Brave New World.
The Mismanagement, as the museum is called, has an eclectic collection of AI-themed and machine-learning-coded art pieces that are, according to its founder and curator Audrey Kim, left behind by a misaligned and apologetic artificial intelligence that unintentionally wiped out humanity. “This museum is a space where people can learn nuances of AI and understand what it is,” she says. It’s also a space where people from the AI industry can geek out on terms. Misalignment, for example, is a pun. It’s used to describe an AI system that is not pursuing its intended objectives (and killing humanity instead).
The temporary museum first sprang up in San Francisco’s Mission District last year, quickly going viral with its whimsical works that veer from an apocalyptic imagination of the future to a mild apology for how technology is ruthlessly taking over, to sheer fun. A broomstick whizzes by me. It’s the Broomba, a broomstick attached to the top of the popular automated vacuum cleaner Roomba, doing an absurdist dance within the thin Saturday crowds, as if to remind us that AI has been around and automating things we use. “All pieces teach you something about AI, show you nuances and are fun,” says Kim who has co-created most of the art in the museum. In the past, Kim worked in Google and Cruise, before leaving the corporate world behind to introduce the public to the nuances and possibilities of AI.
Since we’re in the geek city, which attracts the world’s most ambitious people to build exactly these models, there’s also art that plays on technical terms, sort of an insider joke only an AI nerd would understand. I stand in front of Shoggoth with a Smiley Face — a rather squiggly drawing of a green, many-eyed monster, and find it is an AI meme from 2022, almost vintage by internet standards. The monster hides behind a smile and represents how AI companies are training their language models to be politically correct, polite, and friendly. I laugh, thinking of the latest backlash on Google’s Gemini extra woke model, which when asked to show popes, Nazis or knights, produced images only of people of colour. Google retracted its model after a backlash from white Americans.
One of the strongest works of art, the sculpture Paperclip Embrace is based on another arcane AI term. It shows two humans in each other’s clutches but is made of paperclips, a reference to Oxford philosopher Nick Bostrom’s paperclip maximizer problem about a super AI that was given the goal to manufacture as many paperclips as possible. A modern parable for a task given to an AI, it’s reminiscent of many dystopian novels and short stories where the machines keep working long after the humans they serve have died.
This sliver of darkness pervades most pieces, making you laugh, but also ponder on how AI may change your life in the next decade. It is heightened by large signage that looms high up in the wall, almost an industry apology: “Sorry for killing most of humanity.”
Just below it is an artwork that educates the audience on just one, very real, possibility of a deep fake. A Lesson In Echos plays on the voice cloning scams where people get doom AI calls in the voices of their loved ones. The work asks you to read a spell in a mic in front of you, and in a second, an AI model clones your voice and the marionette puppet sitting across the mic starts to talk like you: “Help! I’ve been cloned.” We know deepfakes will fundamentally change our trust in digital spaces, making us more susceptible to crime and scams, but hearing your voice, saying things you don’t want to, shakes you to the core.
Towards the end of the exhibit, there is Gates to Hell Selfie Spot that warns people in neon cursive font to “Abandon All Hope” while sitting on a fluffy armchair made of stuffed toy animals (The artwork called the Alpine Chair is there for aesthetics). A group of college students walk to the armchair taking selfies. A camera follows them, putting them on a video screen, and adding tags as they walk by. “Person” it announces, reminiscent of CCTVs that abound in our world today and will become omnipresent in the future.
Across the hall from this rather dark corner is a vintage payphone booth, which uses the same deepfake technology. On this phone, you can have a conversation with an AI who mimics the diction and speech of Fred Rogers, a popular American host. It’s one of the most popular art pieces in the museum. Perhaps that shows how resilient we, or maybe San Franciscans are when it comes to adapting to technology. Just like the technology, this museum is a work in progress: Half-finished pieces stand next to the ones I’ve engaged with, waiting for chips, or programs to make them run and introduce possibilities of what AI can do.
As I leave the museum, I come across Sonosynthesis, a self-playing piano which plays a ditty by an AI music composer. The AI algorithm is coded to create short songs based on biological patterns that are generated through the incubation of microorganisms on Petri dishes. The piece is smart and educational on what AI can do, and the piano plays soothing elevator music. However, the calculated automation behind the construction of music makes me feel a bit let down. Isn’t music, fiction, or art created by humans who stare into space? To touch your soul, to change you forever? Can an AI creator have the same effect? Knowing the music was built by an algorithm, makes it less…magical.
Shweta Taneja is an author and journalist based in the Bay Area. Her fortnightly column will reflect on how emerging tech and science are reshaping society in Silicon Valley and beyond. Find her online with @shwetawrites. The views expressed are personal