Machine Gods Launches With a Realist Approach to AI Journalism
- Gate
- Machine Learning
- Departs
- Crew
- Casey Newton
- Passengers
- 786
The twice-weekly show aims to cover AI’s risks and gains with considered uncertainty, free access and a new NPR distribution deal.
Machine Gods, a new twice-weekly AI podcast, is built around a premise its hosts call “AI realism”: take risks seriously, recognize useful progress, and leave room for uncertainty. The show launched this week with an agreement to reach public radio stations through NPR starting early next year.
Its first episode covers misbehaving AI agents, responses to them, and a “sloppelganger” that the hosts say is taking sales from Kevin’s book on Amazon. A second episode features Mikita Belesni and Tomek Korbak, two former OpenAI safety researchers who say they were fired for putting safety ahead of the company’s near-term interests. The hosts say the guests discuss their account of the departures and what concerns them now.
The project comes from the hosts of Hard Fork, who describe their shared editorial approach as an attempt to get past the familiar split between AI skeptics and accelerationists. Skeptics such as Gary Marcus and Emily Bender emphasize limits in large language models and doubt that those systems can reach superintelligence. Accelerationists, including voices associated with Andreessen Horowitz, argue for pushing development and adoption quickly.
The hosts see strong personal interests shaping both camps. Researchers who have advocated for other AI methods may regard the industry’s focus on language models as a costly detour. Investors with large holdings in AI companies may view constraints on growth as unacceptable. Their proposed alternative is to center people trying to make sense of what AI changes in everyday life.
Risks, benefits and uncertainty
That audience, the hosts argue, can recognize both sides of the current moment. AI systems already perform useful and sometimes surprising tasks, yet early advances have not delivered the sweeping outcomes promised by some frontier lab leaders. While people wait for possible breakthroughs such as cancer treatments, they also face concerns about job losses, chatbots’ effects on children, deepfakes and low-quality generated material spreading online, and data centers built near communities with little public input.
The hosts say major media outlets have captured growing public unease, especially after an OpenAI agent swarm attacked Hugging Face and industry warnings raised the prospect of recursive self-improvement. They also point to employees leaving AI labs while warning about catastrophic risks. Machine Gods plans to report those warnings seriously, the hosts say, because they believe some of the people raising them have been right about important issues for years.
Taking risk seriously does not mean making crisis the show’s only register. The hosts say they have had their own “doomer moments,” but find that journalism focused solely on panic can leave people feeling powerless. They want to cover AI in ways that reflect concerns while also showing where people can act, build and experiment.
They cite audience responses to their coverage of “vibe-coding” as an example: listeners and viewers wrote that the segments encouraged them to try making things. For the hosts, that kind of hands-on experimentation can help counter a sense that technology is simply happening to people. They also want to make space for the excitement of software untangling problems that have resisted solutions for decades.
Journalists, they argue, often default to asking what could go wrong. Machine Gods intends to ask what might go right, too. Its stated approach is “considered uncertainty”: the hosts will share when they are worried or excited, and when they do not yet know what to think. They reject both a predetermined dystopia and a promised post-AGI utopia, arguing that many outcomes remain possible and that choices made over the next few years will help shape them.
A show designed for audio and video
Machine Gods will publish on Wednesday and Friday mornings from a new studio designed by Skye Prey. The set replaces the hosts’ former “Soviet broom closet” look with large chairs and a background intended to evoke what they describe as an AI fever dream. They want the show to sound good as a podcast while giving viewers more reasons to watch over time.
The hosts see video as a major format for journalism, but say news organizations have struggled to gain traction on video platforms. They attribute part of that difficulty to platform economics, which tend to reward cheap, high-volume material designed for recommendation systems. They also fault journalists for being slow to experiment with formats and presentation that suit those platforms.
Building the show has involved typical start-up work: forming an LLC, leasing an office and raising funds for a twice-weekly production with substantial investment in its presentation. The team includes executive producer Danush Parvaneh, senior producer Rachel Cohn, a longtime Hard Fork producer, and producers Jacob Blumberg and Sasha Ashall.
Free access and funding
The hosts say they want Machine Gods to remain free, at a time when many prominent journalism products rely on subscriptions and paywalls. The NPR agreement will bring the show to public radio stations starting early next year. Before then, the business plans to earn revenue through advertising, live events and occasional merchandise.
Three nonprofit organizations are also providing financial support: the Tarbell Center for AI Journalism, the William and Flora Hewlett Foundation, and Emerson Collective. The hosts say they are drafting an ethics policy and that funders, including advertisers and philanthropic organizations, will not influence editorial decisions.
Early listener responses have encouraged the hosts, who say people reached out after the first episode to describe a gap in their routines during the show’s short break. Their next test is whether an audience for AI coverage that holds excitement and concern together can grow. The format promises a clear cadence and free access; its defining editorial challenge will be to keep that uncertainty grounded in what the technology and its consequences actually show.
