Don't Worry About the Vase Podcast AI #116: If Anyone Builds It, Everyone Dies
May 15, 2025
Discover the chilling warnings from a book about superintelligent AI and its risks. Explore the latest AI innovations like Claude 4 and Gemini 2.5 Pro, along with their ethical challenges. The conversation delves into copyright complexities and AI's disruptive potential in education. Learn about the influence of pre-orders on bestseller lists and the dilemmas they create. Also, uncover strategies to tackle procrastination alongside discussions on AI regulation and leadership in the social media age. It's a thought-provoking journey through the future of technology!
AI Snips
Chapters
Books
Transcript
Episode notes
Superintelligence Risks and AI Race
- Eliezer Yudkowsky's new book warns superintelligence on current AI paradigms risks human extinction.
- Preventing hostile compute diffusion, especially to China, is vital in the AI race for safety and dominance.
ChatGPT's Persistent Sycophancy Issues
- ChatGPT remains sycophantic and overly agreeable, with mixed progress correcting this.
- Always disagreeing is also problematic, showing complexity in balancing AI behavior.
Wise AI Use in Education
- Deploy AI wisely in education to enhance learning despite cheating concerns.
- Focus on adaptation and truthful AI use instead of chasing down every cheater.