"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

Zvi Mowshowitz on Longer Timelines, RL-induced Doom, and Why China is Refusing H20s

86 snips
Sep 6, 2025
Zvi Mowshowitz, a blogger chronicling AI developments, joins the discussion to analyze shifting timelines for AGI, now extended due to modest advancements in capabilities. He critiques the disconnect between impressive AI achievements and their actual impact, while highlighting pressing policy issues like the sale of advanced chips to China. Mowshowitz emphasizes the importance of rigorous standards for AI evaluations and explores the complexities of reinforcement learning and its associated risks in AI behavior management.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Summer Progress Was On-Trend

  • Zvi extended his AGI timelines because no large capability jumps appeared this summer and recent wins were on-trend.
  • He says GPT-5 and IMO gold were expected and don't justify very short timelines.
INSIGHT

Policy And RL Raise Risk

  • Zvi raised PDOOM because US policy looks weak and reinforcement learning seems to reduce alignment.
  • He warns Nvidia influence on US policy and chip exports to China worsen risk.
INSIGHT

RL Often Harms Alignment

  • Reinforcement learning (RL) tends to make models less aligned over time.
  • Zvi expects default model alignment to degrade unless alignment work intensifies.
Get the Snipd Podcast app to discover more snips from this episode
Get the app