Article URL: https://www.thenewcritic.com/p/manifest-man Comments URL: https://news.ycombinator.com/item?id=48899888 Points: 6 # Comments: 3

*THIS SATURDAY, July 18th, to celebrate our readers, fellow editors, and many esteemed contributors, The New Critic will party in New York on from 8 p.m. to 1 a.m. If you’re not yet a paid subscriber, you can become one below. For $30 a year, paid subscribers get access to Postscript, our interview series, Contra, our criticism section, and exclusive New Critic parties like July 18th’s all-night open bar.* On a cool June evening in Berkeley, I watched a man named Victor try to shoot a beam of light out of his penis. Minutes earlier, he had raised his hand because he found the volunteer onstage before him attractive. And now here he was, facing her in front of a crowd of perhaps 60 people, having been instructed to do precisely this by a woman named Aella, a pseudonymous camgirl and self-described sex researcher. She had promised the men in the room that the exercise, if undertaken in earnest, might actually result in hookups, real ones. Now she told her first subject to imagine himself gathering all of the energy in his body into his pelvis and then sending that energy outward as a beam of light, straight through his trousers and into the room. Victor pointed out that this seemed a little absurd, which it manifestly was. But he did it anyway, standing there and trying — and you could see him trying — to imagine himself emitting light from the tip of his penis. Aella reported that Victor wasn’t shooting enough light, and he readjusted. She and the volunteer watched. The rest of us did, too. And then the volunteer delivered her verdict, which was that Victor had come up short. She had, in fact, “felt more” from him a moment earlier, before he’d begun to imagine light streaming out of his phallus. This was Manifest, and Victor’s ordeal belonged to its official programming. The conference — this was its fourth iteration — convenes each summer at Lighthaven, a verdant complex of gardens and halls in a converted Berkeley hotel that rationalists keep as a gathering place. At Manifest, rationalism has little to do with Descartes; it’s a Bay Area subculture built around the site LessWrong, a “community blog” preoccupied with mathematics, philosophy, genetics, and the apocalyptic risks (and opportunities) presented by artificial intelligence. Around 800 people attend Manifest each year. It takes its name, and its nominal subject, from Manifold, a prediction market. At the opening assembly, Austin Chen, co-founder of Manifold, asked everyone who had ever bet on it to stand, and very nearly the whole audience got to its feet. Officially, then, this is a conference about forecasting. But “a conference about prediction markets” describes Manifest about as well as “a conference about effigies” would describe Burning Man. Prediction markets are what the attendees have in common, but they are not, mostly, what they came to discuss. Any attendee could propose a session on any topic throughout the weekend, and several hundred sessions took place over the three days, a dozen or so underway in any given hour. At one session, attendees proposed political systems that might replace representative democracy (among them “futarchy,” a governance framework in which voters set overarching goals and prediction markets determine the specific policies used to achieve them). At another, a married couple gave a practical guide to IQ-screening one’s unborn children, drawn from their own experience doing the same. At the very moment that I watched Victor try and fail to emit photons from his manhood, a few doors down, a mathematician was explaining his PhD thesis on the topology of three-dimensional shapes. Aella’s session was titled Get Hotter With Aella (Reinforcement Learning: Hot Girl Feedback). Reinforcement learning with human feedback is a technique for teaching a machine learning model to do what a human wants it to do. The model produces an output, a human rewards or penalizes it, and the cycle repeats, each round nudging the machine’s behavior closer to what the human prefers. Here, the model was Victor, and Aella, the human in the loop, was teaching him to be desired.