The eight-bedroom mansion, located in Hillsborough, about halfway between the headquarters for Google and San Francisco-based OpenAI, is called AGI House, focused on what its website calls “the golden age” of AI. It features a koi pond, pool, Zen garden, climate-controlled wine cellar, and custom water well. According to Zillow’s estimates, rent is roughly $45,000 per month.
Traditionally hacker houses — a tech industry rite of passage — referred to cramped quarters shared by start-up aspirants in search of big ideas and cheaper rent. But the money and power flooding into this wave of AI is warping and intensifying the trappings of a typical Silicon Valley gold rush, now set to explode with the launch of GPT-4. (Zillow estimates the value of AGI House at only $18 million, a third of its asking price.)
Tech’s hottest new job: AI whisperer. No coding required.
In recent months, many people, most of them young, have flocked to the Bay Area’s AI scene to be closer to the action of hackathons, meetups, fireside chats and happy hours. Propelled by faith, curiosity, and FOMO, they want to secure their place in an economic future that tech leaders insist will be upended by AI.
“They’re at your house party, they’re everywhere,” said Gloria Felicia, one of the winners of the hackathon at AGI House. The co-founder of Speedify AI and start-up adviser to Spero Studios, she advertised spots in “an AGI House for women,” opening next month in Hayes Valley, an upwardly mobile neighborhood in downtown San Francisco.
Amber Yang, an AI investor with Bloomberg Beta, went viral in January after tweeting that San Franciscans were now calling Hayes Valley “Cerebral Valley” because of the concentration of “AI communities and hacker houses,” amid the trendy restaurants, boutiques, outdoor gym and brick-and-mortar versions of e-commerce brands like Allbirds and Brooklinen.
A combination of Big Tech layoffs, a post-lockdown return to in-person events and lower barriers to entry have inspired a number of new group houses. The promise of easy money helps.
While tech stocks falter, venture capital investors have already funneled $3.6 billion into 269 AI deals in the United States from January through mid-March, according to the investment analytics firm PitchBook, which found that nearly half of the $40.5 billion in AI start-up funding in the country last year was concentrated in Bay Area companies.
In a sign of the times, the Hillsborough mansion recently changed its name from Neogenesis to AGI House. AGI is short for “artificial general intelligence,” a phrase popularized by OpenAI to describe the idea of AI that is smarter than a human. OpenAI argues that tools like ChatGPT, which can instantly answer questions or generate text like software code and college essays, or the text-to-image generator DALL-E, can respond to a user’s natural language prompt, as steppingstones toward superhuman AI. The term “AGI” has become a watchword for proponents who share the belief that this technological wave of AI will transform the internet.
“It’s sort of become the new crypto,” said Moritz Wallawitsch, a 24-year-old German start-up founder who moved to the Bay Area in November. Unlike crypto, he quickly added, AI is generating a lot of value because it’s “automating jobs.”
Hacker houses are “symptomatic of when people are just all in on building a company and when people are like trying to immerse themselves and learn from others,” said investor Sarah Guo, founder of the early-stage venture capital firm Conviction, who recently attended a dinner at AGI House.
“Obviously it doesn’t work for people at every life stage in every lifestyle,” Guo continued. But co-living is working for the employees of one of her portfolio companies, Harvey, which is building AI models for law firms, because they’re excited about growing their company quickly. “They don’t really do anything else right now. They just work,” she said.
The cost of renting a room in a hacker house works much like having roommates. Openings for rooms in Bay Area communal housing for April run from $650 per month for a shared room up to $3,000, according to a Google doc of different types of collectives. Some houses have an additional monthly fee for shared groceries or cleaning.
The application process is more unusual. Some houses have websites with links to Google forms for would-be members and consider where you work and technical prowess. Others just post a photo of the digs on social media and encourage interested parties to message them on Twitter.
Group houses that mix work-related decisions and career networking with personal lives can be hard for women in tech and quickly turn toxic, said venture capitalist Brianne Kimmel, who has invested in the company behind Stable Diffusion, an open source DALL-E competitor, and other AI tools. As a female investor, Kimmel added, she gets invited only to the 8 a.m. breakfast meetings, not the parties.
Felicia, one of the hackathon winners, said she is trying to cultivate a different vibe with the AGI house for women, one focused on safety, inclusion, mindfulness, and wellness, with no alcohol or drugs.
Most of the applicants for Felicia’s AGI house for women were more interested in a community space than co-living she said. They were really worried about security and the possibility of sexual assault by “strangers and, well, sometimes some of the tech bros,” she explained. “They don’t know the boundaries. They don’t understand social cues as well.”
Co-living has been key to building the Bay Area’s tightknit AI scene over the past decade. It’s part of the origin stories for OpenAI and Anthropic, two of the wealthiest AI start-ups, both based in San Francisco. Socializing at those houses became a fast track to hearing about jobs at a top research lab, access to start-up investors, and even insight on technology from the people who develop it. During the pandemic, those networks deepened.
The allure of events at AGI House is the possibility of meeting Silicon Valley elite. The weekend hackathon was sponsored in part by Hugging Face, the open-source AI company valued at $2 billion, with welcome remarks from tech luminary Sebastian Thrun, the self-described godfather of self-driving cars.
The previous incarnation of the house, Neogenesis, was started by OpenAI’s Andrej Karpathy, Tesla’s former head of AI, and was known for throwing lavish parties where tech titans like Google co-founder Sergey Brin might stop by.
Karpathy and AGI House did not respond to requests for comment.
Wallawitsch, the German start-up founder, first arrived in the city on a tourist visa in 2021, eager to explore the Silicon Valley he knew only from podcasts and blogs. For a couple of months, he lived in the San Francisco precursor to Neogenesis, called Genesis House. About half of the residents were AI researchers at the time, including some from OpenAI, and Wallawitsch ended up joining a reading group for keeping up with research papers and learned about the latest breakthroughs in AI.
“When you live with someone, you’re definitely more at the edge of what’s happening,” he said. Serendipitous interactions, like being automatically introduced to a roommate’s interesting friends, “isn’t something you necessarily get at an event,” said Wallawitsch, who introduced a former housemate to one of the investors in his company RemNote, an app for studying and organizing information.
Wallawitsch launched his own group house in San Francisco’s Panhandle this month. Applicants didn’t necessarily have to work in AI, but he looked for candidates interested in machine learning and human-computer interaction. He explored incorporating generative AI into his company but decided it wasn’t the right fit and good investors would know better. Forty people inquired about a spot in the six-bedroom house.
For people in their early 20s, who have never worked in a non-covid environment, orienting their lives around this technological shift can be appealing. “They’re like, I would willingly work seven days a week in person and live with these people,” said Yang, the AI investor who popularized the term Cerebral Valley.
Some tech insiders winced at the nickname, which had not been in use. Others bristled at the idea that San Francisco was the center of the AI universe. But in the weeks since, the moniker keeps popping up.
There’s Cerebral Valley AI, a community-building initiative that hosts co-working sessions in start-up offices and a handy Google Doc of events that are not always widely promoted. The upcoming Cerebral Valley AI Summit, an invite-only event in “the heart of the AI boom,” will take place at the end of the month, hosted by Substack writer Eric Newcomer. On Partiful, the tech crowd’s favorite tool for online invitations, AI event listings also increasingly advertise their location as “Cerebral Valley.”
“I’m 25 years old, still in student loan debt. I need to be as relevant as possible,” said Aqeel Ali, a former operations manager who helps organize Cerebral Valley AI. Ali said he barely left his bedroom for two weeks after ChatGPT was released in late November because it was clear the technology could do the work of “eight junior employees.”
Within a decade, he reasoned, this technology would soon be able to handle on-the-ground operations in the real world — not just writing up plans for an event, say, but ordering catering from DoorDash.
But Ali says adopting this type of AI will lead to new jobs, not just take old ones. He pointed to Anthropic’s help wanted listing for a prompt engineer (a nontechnical role for people good at talking to AI models) that advertised a salary of at least $250,000.
“You want to know why those jobs exist,” Ali said. “I just want to find one I’m excited about that I’m good at and that’s needed.”
Ali, who recently started his own Hayes Valley group house called Luminance, said there has been a cultural shift since the pandemic back to “community.” Event organizers have even opted against a Zoom conference option for talks. “We’re trying to maintain a high-fidelity, high-quality experience.”
Venture capitalists have invested $3.6 billion into 269 AI deals in the U.S. from January to mid-March, according to data from PitchBook. A previous version of this story incorrectly said the money was invested into 269 AI start-ups.