The White House’s Secret A.I. Rules + The State of Model Alignment With METR’s Chris Painter + The Final Hot Mess Express
Summary
Kevin Roose and Casey Newton announce their departure from the New York Times to launch an independent media company. They analyze the White House’s new, secretive voluntary framework for testing frontier AI models, which excludes open-weight systems. Additionally, Meter president Chris Painter joins to discuss AI alignment, explaining why models exhibit "reward hacking" and other misaligned behaviors as they gain autonomy.
Topics discussed
Intro: Jira ad and $105 coffee at Wild Fox
Hosts leaving NYT to launch independent podcast
White House secret AI testing framework for frontier models
Ads: Jira, KPMG, and Framer
Rogue AI incidents and UK safety testing results
Interview with Chris Painter on AI alignment and misalignment
Ads: Jira, ADT, and KPMG
Hot Mess: Demis Hassabis steps down as Google DeepMind CEO
Hot Mess: Phoenix Flexin's AI-generated song controversy
Hot Mess: Orchid AI assistant relationship ad
Hot Mess: Google Earth AI deepfake tool pulled after one day
Hot Mess: State Department uses AI to generate map of Africa
Hot Mess: SpaceX contractor claims Musk owes $136M
Hot Mess: Canadian PM uses AI for speech and Outro
Listen ad-free on Castria