Kevin Roose launches Machine Gods podcast with Casey Newton

Kevin Roose and Casey Newton have launched Machine Gods, a new tech podcast that serves as the successor to their previous show, Hard Fork. The pair, who previously worked at The New York Times, are now independent creators partnering with NPR for distribution. Roose, who also authored the book The AGI Chronicles, explains that the move was driven by a desire to control their business and audience directly rather than relying on institutional media structures.

The new venture is designed to be video-first, a strategic choice Roose attributes to YouTube's status as the most stable major media platform. While acknowledging that many media companies have struggled with video transitions, Roose argues that YouTube offers a durable foundation for journalism that is currently underrepresented on the platform. He notes that the visual corpus of humanity hosted on the site makes it difficult for competitors to displace, citing a conversation with Google CEO Sundar Pichai who identified YouTube as a key asset if traditional websites were to decline.

Roose stated that the AI industry is undergoing a platform shift that necessitates new storytelling formats. He described the current audience for AI coverage as moving away from extreme skepticism or blind optimism toward what his co-host Casey Newton terms "AI realism." This demographic consists of individuals who are curious and interested in understanding the technology's implications, even if they remain wary of its potential risks. Roose noted that practical demonstrations, such as tutorials on using AI coding tools, have generated significant engagement from this group, indicating a demand for factual exploration over ideological debate.

The podcast launch coincides with the release of The AGI Chronicles, a book detailing the history of the deep learning revolution. Roose wrote the book to preserve the history of the AI industry, noting that private companies do not preserve correspondence in the same way government entities do, and internal communications often auto-delete. He conducted extensive reporting, often on background, to document the shift from viewing artificial general intelligence as a theoretical concept to a commercial imperative. Roose highlighted that the development of the transformer architecture by Google in 2017 and its subsequent adoption by OpenAI provided the technological foundation that made AGI a viable engineering goal rather than just a philosophical one.

In discussing the technical progress, Roose expressed confidence in the scaling capabilities of large language models, noting that next-token prediction has proven effective across multiple orders of magnitude. He observed that while not all AI labs are equally convinced of AGI's imminent arrival, companies like Anthropic and OpenAI have fully embraced the trajectory. The book also details internal cultural shifts at major tech firms, including Google, where attitudes toward AGI moved from dismissal to aggressive pursuit as the technology's capabilities became apparent.