Dinners for AI Safety

Published on October 21, 2026 4:04 AM GMT I’d be pretty happy if we lived in a world where in every reasonably sized city in the US (and perhaps more widely) there was one (or more) local regular dinners where people talked about AI. Not meetup groups, though those seem maybe good, too. Dinners, in someone’s home, run by nice people, ideally potluck or otherwise home-cooked, where they invite friends, family, and people they run into on the street, to just talk about AI, perhaps with guides or what to talk about or financial support, perhaps not. I think (though haven’t deeply researched) that this kind of thing has been responsible for quite a lot of political change. Most famously feminist consciousness-raising groups where small groups of women met in living rooms, often weekly, and took turns talking about their own lives on a chosen topic, which is where we get “the personal is political.” Moishe House and other groups fund people to run Shabbat dinners, just so more Jewish community can happen (this doesn't aim to or succeeed in leading to political action). Other examples Claude pulled below, with counterarguments below. And I think it solves a lot of problems at once. What can any person do? Host a dinner. People are scared and uncertain, and they’d benefit from talking to other people. People are lonely, and would benefit from talking to their friends and neighbors in person, without phones or the internet mediating. What does the world need? People to wake up and demand something different from their elected officials[1]. People will take more actions, and identify more with doing something about AI, if they have social ties built around it. And it speaks to the moment: this is happening to all of us, and we’re all in it together. I’ll admit it also just kind of feels like the kind of thing that would be happening in a sane world. I also think it could turn into some pretty good stuff: showing up at presidential town halls, calling reps, making Minneapolis (or whatever) the proud “Don’t Build Superintelligence” city. Here’s what we’d do if we believed the above: 80,000 Hours videos would end with hosting a dinner as a call to action, perhaps even fund it One or more of the various mass movement organizations would have an arm focused on making this happen (at one point I thought Humans in Control was doing this but I no longer get that sense - but haven’t talked to them) List I know about: Evitable, Humans in Control, Humans First, Sapiens 1st, Irreplaceable (Alexander McCoy), Pause AI, Stop AI, AI Awareness Foundation (Changlin Li), Quit GPT, Stop the AI Race People could just try this out with their families and see how it goes. Overall, I’m not sure about this, but I feel reasonably excited about it. Examples Farmworker movement. Fred Ross and Cesar Chavez built the Community Service Organization and later the UFW through “house meetings”: a supporter invites a few neighbors over, an organizer talks, and each guest is asked to host the next one. This is probably the clearest case of the format being a deliberate growth engine. Swedish study circles. From the late 1800s, the labor, temperance and free-church movements ran huge numbers of small reading and discussion groups. They’re often credited with building the leadership class of Swedish social democracy. Civil rights. Septima Clark’s Citizenship Schools taught literacy (to pass voter tests) in kitchens, churches and beauty parlors. Highlander Folk School ran workshops in a similar spirit. Abolition and suffrage. Female antislavery societies were often sewing circles that made goods for fundraising fairs and talked politics while doing it. Suffrage organizing ran heavily on parlor meetings. The right too. The early pro-life movement grew a lot through church and home networks, and the 2009–10 Tea Party was around a thousand local groups meeting in restaurants, libraries and homes. Electoral versions. MoveOn’s house parties (2003–04) and Obama 2008’s house meetings borrowed directly from the Ross/Chavez model. Doug McAdam found that what predicted who actually showed up to Freedom Summer was mostly social ties to other participants, not ideology. Ziad Munson found many pro-life activists got involved through relationships first and developed strong views afterward. And Hahrie Han’s How Organizations Develop Activists is the best modern treatment: she distinguishes “mobilizing” (getting lots of people to take a single action) from “organizing” (small-group work that turns people into leaders who recruit others), and finds the high-performing chapters do much more of the second. Counterarguments (pulled from Claude) We remember house meetings in movements that won. Plenty of failed movements had reading groups too. Consciousness-raising specifically had a mixed record, and critics said so at the time. Jo Freeman’s “The Tyranny of Structurelessness” (1970) argued leaderless small groups produce hidden informal hierarchies and struggle to