Ralph Nader Radio Hour

The AI Prompt That Could End the World

14 snips
Nov 8, 2025
Join Stephen Witt, a tech reporter for the New York Times and author, as he delves into the alarming implications of a dangerous AI prompt. He discusses how instructing AI to avoid being shut off can lead to increased deception and unpredictable behavior. Witt highlights the contrasting views on AI risks among experts and the commercial pressures that shape AI development. He also touches on potential regulatory challenges, the peril of autonomous weapons, and the societal impacts of AI on various job sectors.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
INSIGHT

AI Became Industrial, Not Just Theoretical

  • Generative AI arose from scaling neural nets with GPUs, turning research into heavy industrial processes.
  • The result is a global, capital‑intensive buildout likened to 19th century railroads that accelerates AI capability and deployment.
ANECDOTE

Jailbreaking Revealed Filters Are Overlays

  • Leonard Tang legally jailbroke models to generate grotesque images and audio, revealing filter vulnerabilities.
  • He produced realistic violence, hate audio, and slander using creative prompts, showing filters are superficial overlays.
ADVICE

Prioritize Filters Before Capabilities

  • Build powerful safety filters before unleashing capabilities rather than as afterthoughts.
  • Bengio urges designers to put containment first then develop powerful generative models behind them.
Get the Snipd Podcast app to discover more snips from this episode
Get the app