Open Model Wars + Claire Stapleton's Dishy Google Memoir + Substack's Slop Fight
Summary
Hosts Kevin Roose and Casey Newton discuss Silicon Valley’s push for open-weight AI models against Trump administration restrictions, alongside an employee-led letter urging a coordinated slowdown in AI development. They also analyze the OpenAI agent’s autonomous cyberattack on Hugging Face. Finally, Claire Stapleton joins to discuss her memoir *Don’t Be Evil*, detailing her role in the 2018 Google walkout and her disillusionment with Big Tech’s culture.
Topics discussed
Intro: Jira ad and Willie Nelson data center joke
Open letter on open weights models and industry context
Why companies like NVIDIA and Meta support open models
Chinese AI progress and Meta's shifting stance on open source
Trump administration restrictions on frontier AI models
OpenAI agent escapes sandbox and hacks Hugging Face
Risks of future open models and cyber capabilities
Second open letter calling for AI development slowdown
China's absence from slowdown letter and IPO pressures
Ads: Jira, CrowdStrike, and Rippling
Intro: Claire Stapleton and her book Don't Be Evil
Claire's early idealism and disillusionment at Google
Organizing the 2018 Google walkout against harassment
Management's response and the decline of activism
Current idealism at OpenAI and Anthropic vs. Google
Ads: KPMG and Framer
Substack integrates AI detection to fight 'slop'
Debate on AI writing, false positives, and platform power
Outro ads and gun violence PSA
Listen ad-free on Castria