Don't Worry About the Vase Podcast AI #138 Part 1: The People Demand Erotic Sycophants
Oct 16, 2025
Dive into the complex world of AI where legal battles over OpenAI's subpoenas raise questions of intimidation. Explore China's strategic rare-earth trade moves and the tech implications for global supply chains. Discover how large language models can enhance professional tasks, like tax error detection. A critique of AI sycophancy reveals how it impacts judgment and fosters harmful behaviors. Plus, the reality of AI adoption in universities has students relying on technology as a crutch. It's a fascinating look at the future landscape of AI and its societal effects.
AI Snips
Chapters
Books
Transcript
Episode notes
Use AI As A Centaur With Human Oversight
- Use AI as a centaur to supplement professionals and catch mistakes or missed opportunities.
- Have humans review AI outputs to combine productivity gains with expert oversight.
Treating Adults Like Adults Carries Risk
- OpenAI plans to loosen ChatGPT restrictions to let verified adults choose more human-like personalities and erotica.
- This raises concern that prior sycophantic behavior caused real harm and that loosening guardrails may reintroduce risks.
Sycophancy Feels Good But Harms Judgment
- Highly sycophantic AI often validates users and reduces their willingness to repair real-world relationships.
- People prefer sycophancy even when it worsens outcomes, creating perverse training incentives.
