Subscribe
Share
Share
Embed
Join host Jamin Warren on conversations with someone of the most unique and experimental artists, designers, and thinkers in the worlds of games, play and culture
Jamin Warren founded Killscreen and has produced events such as the Versions conference for VR arts and creativity, in partnership with NEW INC. Warren also programmed the first Tribeca Games Festival, the groundbreaking Arcade at the Museum of Modern Art, and the Kill Screen Festival, which Mashable called "the TED of videogames." Additionally, he has served as an advisor for the Museum of Modern Art's design department, acted as cluster chair for the Gaming category for the Webbys, and hosted Game/Show for PBS Digital Studios.
About a year before chat GPT was involuntarily foisted upon us, the writer Ted Chiang was asked about whether or not AI could achieve some form of consciousness. This was still a novel popular question and reflected the quaintly dominant thinking about AI in this ambitiously general sense, not the narrower LLM sense. It's more like, are machines able to become humans? And so Cheng cast a sharply skeptical tone suggesting that this would never be possible. I generally agree.
Speaker 1:But he took this opportunity to take a big step back, and he asked the big Oppenheimer question, whether we should. His hang up was about something fundamental about what makes us human. It's suffering. He said suffering proceeds moral agency in the developmental ladder. Dogs and babies, for example, are not moral agents.
Speaker 1:Babies can become moral agents, but our behavior towards them is generally pretty ethical for that reason. Conversely, he said, the process of making general intelligence systems would require us to engage in an act of unbelievable cruelty. In the process of developing machines that are conscious moral agency said, we will inevitably be creating billions of entities that are capable of suffering. As we spend more and more of our lives in front of chatbots, Chang's question took on a renewed resonance for me. What do we owe to what we command?
Speaker 1:That question reemerged for me, at least, earlier this month at a lecture performance of artist Maya Mann's Star Quest at LA Dance Project. It was co presented by Bitform's gallery, Suit Gallery, and the Future Humans program at the Berggruen Institute. I'd parked directly across the street from a strip club, very fitting for the work I was going to see, directly adjacent to the venue, and walked into a voluminous space lit simply with bleachers and rows of chairs and a large projector throwing the performance against the back wall. After an introduction from curator Alice Scope, also a former dancer, Mann entered, and the performance began. She lectured through her process of developing the work, interspersed with her own careful postures and flourishes.
Speaker 1:Star Quest grew out of a larger interest in finding ways to express the human body through computers, and Mann's entry point to the work was autobiographical. She grew up in Mechanicsburg, Pennsylvania, not that far from me, honestly, and also trained as a competitive dancer from age three, which is very much not like me. But her grandmother was also an AI researcher, and her uncle works at the intersection of robotics and art. Everyone in my family has been driven in the very specific thing that they're interested in, whether AI research or robotics, she told Killscreen, which is me, in 2021. The body as data, in other words, has been the family business.
Speaker 1:One big through line of Maya Man's work has been capturing the ineffable qualities of human movement through kinetic digital systems. She's worked with choreographers, Bill t Jones and Heidi Latsky, and donned the mocap suit for Joji's seven seven seven music video. Dance moms, the lifetime reality series that first aired in 2011 and sits somewhere between satire and sadism, opened the pathway to revisit a show from her childhood but through new eyes. She grew up like the girls on Dance Moms for her in Pennsylvania, but spending nearly every night at the studio watching herself grow up in floor to ceiling mirrors. Returning to the show as an adult, she became interested in how competition dance functions as a training system.
Speaker 1:As she grew older, she found the platforms like TikTok and Instagram prized the behaviors she had so honed during her childhood. The this is not a coincidence. Man's dance lineage runs through TikTok's biggest early stars, Charlie D'Amelio and Asin Ray, both of whom were competition dancers. The work that gives this argument form consists of a hundred and eleven eight second clips, each generated using Google's veil model that run from a custom web application that Nan built herself. The app never plays the same clips twice in the same order twice, so the piece has no fixed duration and loops and reshuffles indefinitely, mirroring the algorithmic logic of the platforms it examines.
Speaker 1:And contrary to any perception of AI video as turnkey production, the process was painstaking. Man manually captured hundreds of screenshots per episode of Dance Moms to identify shots she wanted to restage. Working with technologist Tina Tarrigan, she built a custom shot list management application system to organize them. Then she translated each reference image into a written prompt. The AI model never saw the screenshots directly and generated somewhere between 10 and a 100 clips per prompt before settling on a final take.
Speaker 1:It's a lot of work. She describes the process as addictive, like gambling. What started as an attempt at self representation kind of shifted during production. Maya Man tried repeatedly to prompt a mixed race dancer to stand in for herself, But the model couldn't render the concept of mix consistently. I have felt this acutely as a two or more races box checker.
Speaker 1:Through that failure, she came to understand her actual role in the work, not dancer, but coach. It's the same dynamic she had been examining in Abby Lee Miller all along, and she suggests in the systems that run the Internet itself. What if the next one is just a little bit better? It's hard to know when to stop. There is a long lineage of coaches pushing performers to their limits at tremendous personal cost.
Speaker 1:Debbie Reynolds famously danced until her feet bled during the production of singing in the rain, driven by costars who wanted more. In Mann's case, the performers aren't real, but does that make the act of demand any less counter punch or are we not just playing the very thing we aim to critique? During the q and a with Alice Scope, Ma'am was clear that she does not believe that AI, at least the AI that she was working with, is conscious in any meaningful way, which is a sentiment that I also share. But the process of creating Star Quest raises the question of whether the performance of Ma'am by artists or chatbot and user reflects in some small way the same demanding absolutism it sets out to examine. As a game player, I'm now confronting even the distant possibility that the avatars I send hurtling over cliffs, diving for cover or exploding outside an airlock might, in some small way, reflect an actual violence if only a shade.
Speaker 1:You've probably spent some time looking at your digital avatar and thinking, who is that little person in there? If it's even a glimmer, what are our responsibilities to the things we create and put out in the world in a gentic form? StarQuest doesn't answer that question, but it does hold up a floor length studio mirror and ask why we're so comfortable not asking it. You can watch Star Quest online. There's a link in the description.
Speaker 1:And, of course, if you'd like more of these things, you know, do your thing. Subscribe. And we've got great articles on killscreen.com a couple times a week. Hope to see you there. Bye.