The Dishcast with Andrew Sullivan Karen Hao On The Overreach Of AI
Oct 24, 2025
Tech journalist Karen Hao, known for her work at MIT Technology Review and Wall Street Journal, delves into the complexities of AI. She explores the environmental impact of AI and its implications for democracy. The conversation also covers the rise of OpenAI, the contrasting leadership styles of Sam Altman and Elon Musk, and the unique threats AI poses to white-collar jobs. Hao warns about misinformation in AI training data and critiques the competitive pressures driving the industry, emphasizing the need for balance between innovation and democratic values.
AI Snips
Chapters
Books
Transcript
Episode notes
Two Roots Of Modern AI
- AI has two main branches: learning-based machine learning and symbolic knowledge-based systems.
- The last decade favoured data-driven machine learning due to cheap data, cheaper compute, and tech company dominance.
Why Scale Became The Game
- Three forces drove the recent ML boom: cheap internet data, cheaper compute, and huge tech-company funding.
- Those forces created a winner-takes-most dynamic favoring firms with vast data and capital.
Scale Can Hurt Accuracy
- Bigger datasets can degrade models if their quality is poor or unknown.
- Scaling without curation produces garbage-in, garbage-out and unknowable training corpora.
