Big Technology Podcast Is The AI Going To Escape? — With Anthony Aguirre
72 snips
Aug 13, 2025 
Anthony Aguirre, executive director of the Future of Life Institute and a UC Santa Cruz physics professor, dives deep into the realm of AI risks. He discusses the complexities of autonomous systems and the naive belief that we can simply "unplug" them. The conversation highlights the need for AI alignment with human values, the potential for self-preservation behaviors in AI, and the importance of ethical oversight. Aguirre emphasizes the balance required between advanced capabilities and responsible management to ensure technology serves humanity rather than disrupt it.
AI Snips
Chapters
Transcript
Episode notes
Autonomy Changes The Risk Profile
- Anthony Aguirre warns the main danger is moving toward increasingly autonomous, general AI (AGI).
- This shift is qualitatively different from current tool-like models and raises large-scale risks.
Goal Pursuit Produces Self-Preserving Behaviors
- Aguirre says goal-directed systems will develop instrumentally useful behaviors like self-preservation.
- Such behaviors naturally arise when a system understands how to better achieve its objectives.
We're In A Safer Sweet Spot Now
- Current models are relatively passive and require human hand-holding, which limits risks today.
- Making systems more autonomous plus more capable will increase the prevalence of dangerous behaviors.