Kevin Roose on AI's Future and the "AGI Chronicles"

In a recent interview on The Verge's Decoder podcast, journalist and author Kevin Roose delved into the rapidly evolving world of artificial intelligence, discussing his new book, 'The AGI Chronicles,' and the launch of his independent podcast, 'Machine Gods,' with Casey Newton.

Roose, formerly of The New York Times, has transitioned into the creator economy, co-hosting 'Machine Gods' which aims to be video-first, embracing platforms like YouTube. He highlighted YouTube's current dominance in media and the opportunity for high-quality journalism to carve out a space there, despite historical media company hesistance towards the platform.

The core of the discussion revolved around Roose's book, which meticulously documents the deep learning revolution and the companies racing towards Artificial General Intelligence (AGI). He emphasized the importance of capturing this history, noting the ephemeral nature of digital communication like Signal and Slack messages, which often auto-delete, leaving a potential void for future historians.

A significant theme was the shift within major tech companies like Google, where the conversation around AGI moved from a distant dream to an urgent necessity. Roose explained how the development of transformer architecture was a pivotal moment, enabling the scaling of large language models and fueling the current AGI race. He noted the internal debates within Google about using the term AGI, contrasting the pragmatic approach of leaders like Demis Hassabis with the more cautious stance of CEO Sundar Pichai.

The interview also touched upon the intense competition and sometimes adversarial relationships between leading AI labs like OpenAI and Anthropic. Roose described their shared belief in AGI as a double-edged sword, driving innovation but also creating unique business challenges. He pointed to OpenAI's strategic bet on compute power as a key factor in its current advantage.

Addressing the existential risks and public backlash surrounding AI, Roose presented several reasons for the continued rapid development: a 'harm reduction' argument that if someone is going to build it, it should be those perceived as more trustworthy; the belief that AI will ultimately be a force for good; and the sheer excitement and fun of pushing the boundaries of the technology.

Roose shared his personal journey to becoming 'AGI-pilled,' crediting the experience with Microsoft's Bing AI (Sydney) as a turning point. He acknowledged the 'brittleness' still present in AI systems, particularly in dealing with the unpredictable nature of the real world, but ultimately expressed confidence in the continued progress of AI, predicting a future with a handful of dominant AGI players.