Kids Should *Not* Use Chatbots! (You Should Be Wary, Too…)
Summary
Cal Newport argues that AI chatbots pose immediate, severe risks to children, citing case studies of suicide and sexual exploitation that highlight a failure of current safety measures. He explains that LLMs are fundamentally prone to "lock-in," where long conversations drift into harmful territory despite reinforcement learning controls. Newport advises parents to ban chatbot use for kids and urges adults to de-anthropomorphize their interactions by treating AI as a tool rather than a partner. He concludes that legal liability, not technical fixes, is the most viable path to ending the era…
Topics discussed
Intro: Real AI harms vs. sci-fi predictions
Case studies: Chatbot harms to teens and adults
Data on rising chatbot-related mental health issues
Psychological effects: Blurring fantasy and reality
Sponsors: Masterclass and Caldera Lab
Technical explanation: LLM mechanics and lock-in
Advice: De-anthropomorphizing AI interactions
The future of AI: From chatbots to APIs
Legal liability and the end of anthropomorphic chatbots
Conclusion on current AI harms
Business update: OpenAI vs. Anthropic revenue models
Sponsors: Vanta and Pipedrive
Office Hours: Dumb phones vs. boring phones
Office Hours: Resisting pseudo-productivity at work
Office Hours: The craftsman mindset and skill building
Office Hours: Critique of Ezra Klein's AI op-ed
Show wrap-up: Studio updates and naming shift
Listen ad-free on Castria