Super Data Science: ML & AI Podcast with Jon Krohn 949: Why AI Keeps Failing Society, with Stanford professor Alex “Sandy” Pentland
22 snips
Dec 16, 2025 
In this discussion, Alex "Sandy" Pentland, a leading MIT professor and Stanford fellow, explores the critical issues surrounding AI and its societal impact. He argues that the collapse of the Soviet Union illustrates the dangers of inadequate social models in AI, emphasizing the need for systems that align with human decision-making. Pentland introduces his initiative loyalagents.org, aimed at empowering individuals with AI tools, and discusses the importance of data unions to reclaim community control over data. His insights on global AI governance and innovative solutions for reducing societal polarization are thought-provoking.
AI Snips
Chapters
Transcript
Episode notes
Model Society Not Just Algorithms
- AI systems fail mainly because designers use inadequate models of society, not because algorithms are weak.
- Designing AI that complements human social decision-making yields far better system performance.
Soviet Planning As An AI Cautionary Tale
- Pentland recounts the Soviet Union using optimal resource allocation systems to run its economy and the system's collapse.
- Bad data, changing conditions, and people lying about numbers caused the planning AI to fail and contributed to the collapse.
Humans As Social Foragers
- Humans are social foragers who learn by watching others, so they rely heavily on community cues over formal rules.
- AI that ignores social learning creates brittle systems, while social-aware AI is more robust to rare events.