The Lawfare Podcast Lawfare Daily: Jonathan Zittrain on Controlling AI Agents
4 snips
Oct 17, 2024 
Jonathan Zittrain, Faculty Director of the Berkman Klein Center at Harvard Law, dives into the critical need for controlling AI agents. He distinguishes these agents from current generative AI, raising alarm about their autonomy and decision-making risks. Zittrain calls for urgent regulatory measures, drawing parallels to historical tech incidents. He humorously suggests that even simple tasks like ordering pizza might require a license to ensure accountability in AI use. The conversation emphasizes the balance between leveraging AI benefits and addressing ethical concerns.
AI Snips
Chapters
Transcript
Episode notes
Past Algorithmic Disruptions
- In 2010, algorithms caused a flash crash, highlighting the risks of automated systems.
- A used book on Amazon reached a $2.4 million price due to interacting algorithms.
Intellectual Debt
- Unforeseen interactions between simple algorithms can create systemic issues, similar to technical debt.
- This 'intellectual debt' makes it difficult to predict and control outcomes.
Agent Responsibilities
- Treat AI agents as legal/social agents with duties to their users.
- Establish clear responsibilities and prevent conflicts of interest in their design.