Inside the early days of the AI revolution
Summary
Host David Pierce discusses the early history of the AI industry with Kevin Roose, author of The AGI Chronicles. They explore how interpersonal rivalries and the belief in AGI as a successor species drove the rapid development of AI, while also examining Google's bureaucratic failures and OpenAI's decision to launch ChatGPT despite its flaws. The conversation addresses the ethical controversies surrounding training data and the ongoing debate over whether AGI has been achieved.
Topics discussed
Intro: AI history, Anduril laptops, and data center politics
Introduction of Kevin Roose and The AGI Chronicles
Google's DeepMind acquisition and the Musk-Page split
Demis Hassabis's competitiveness and the Singleton hypothesis
The 'Seekers' worldview and the nature of human intelligence
The small AI community and the decision to race
The Transformer architecture and the start of the race
Google's bureaucratic failures and the Marmot project
The launch of ChatGPT and the element of surprise
The fun of conversation and the discovery of scaling laws
The ethics of training data and copyright concerns
Compensating creators and the 'ends justify the means' debate
Reflections on the Sydney incident and public misalignment
Defining AGI and whether we have reached it
Promoting an AI movie creation game
Outro and listener contact information
Listen ad-free on Castria