BlueDot Narrated med BlueDot Impact finns tillgänglig på flera plattformar. Poddens innehåll och bilderna på den här webbplatsen hämtas med hjälp av poddens RSS-flöde.

BlueDot Narrated

Avsnitt

Learning From Human Preferences

What Failure Looks Like

Deceptively Aligned Mesa-Optimizers: It’s Not Funny if I Have to Explain It

Goal Misgeneralisation: Why Correct Specifications Aren’t Enough for Correct Goals

Thought Experiments Provide a Third Anchor

ML Systems Will Have Weird Failure Modes

Where I Agree and Disagree with Eliezer

AGI Ruin: A List of Lethalities

Why AI Alignment Could Be Hard With Modern Deep Learning

Yudkowsky Contra Christiano on AI Takeoff Speeds