Episodes (Page 4)
✨
Joe Carlsmith examines whether we can trust power structures and techno-capital in AGI development
✨
Patrick McKenzie describes how small Discord-based team circumvented broken government incentives to vaccinate thousands of Americans
✨
Tony Blair reflects on lessons from Lee Kuan Yew and intelligence agency performance on Iraq and Ukraine assessments
✨
Discussion on why large AI models struggle with simple puzzles.
✨
Leopold Aschenbrenner predicts AGI by 2027 and discusses US/China race.
✨
John Schulman explains how post-training and reinforcement learning tame base model capabilities, discussing the nature of progress toward AGI
✨
Llama 3 represents Meta's push toward open-sourcing models on the path to AGI; open source releases create dual-use risks including bioweapons and security concerns
Mark Zuckerberg
✨
LLMs learn through associations rather than explicit reasoning; intelligence emerges from recognizing patterns in training data rather than deliberate computational steps
✨
Scaling is an artform balancing compute, data, and architecture; adding search, planning, and AlphaZero-style training on top of LLMs unlocks new capabilities
✨
Patrick Collison advises talented 20-30 year-olds to leave Silicon Valley and pursue founding or domain expertise; building multi-decade APIs, companies, and relationships compounds long-term value
✨
Tyler Cowen analyzes greatest economists (Keynes, Hayek, Smith, Mill, Coase, Schelling, George) and their relevance to AI, growth, and alignment challenges
✨
Narration of blog post analyzing Robert Caro's Years of Lyndon Johnson biography and lessons about power, persuasion, and political effectiveness
✨
Blog post examines whether AI scaling will continue producing capability gains or hit fundamental limits
✨
Jung Chang lived through China's Cultural Revolution as daughter of a denounced official; witnessed CCP totalitarianism subjugating a billion people through systematic persecution
Andrew Roberts — Why Hitler lost WWII, Churchill as applied historian, & Napoleon as startup founder
✨
Andrew Roberts examines why Nazi ideology cost Hitler WW2 and how Churchill functioned as an applied historian using historical lessons to inform decisions
✨
Dominic Cummings details catastrophic failures in Western government revealed by COVID response; civil service incompetence and institutional dysfunction enabled preventable crises
✨
Paul Christiano has modest AGI timelines: 40% by 2040, 15% by 2030; addresses whether RLHF invention was regrettable and whether alignment is necessarily dual-use
✨
Shane Legg expects AGI around 2028; aligning superhuman models requires different approaches than current RL methods and new architectural innovations beyond transformers
✨
Grant Sanderson argues advanced mathematics doesn't require AGI; mathematically talented students should pursue foundational research, teaching, or AI alignment rather than finance
✨
Continental vs maritime power mentalities explain Xi and Putin's strategic errors; dictators consistently misread adversary resolve and overestimate military advantages leading to catastrophic misc...