AXRP - the AI X-risk Research Podcast 38.1 - Alan Chan on Agent Infrastructure
Nov 16, 2024
Alan Chan, a research fellow at the Center for the Governance of AI and a PhD student at Mila, delves into the fascinating world of agent infrastructure. He highlights parallels with road safety, discussing how similar interventions can prevent negative outcomes from AI agents. The conversation covers the evolution of intelligent agents, the necessity of understanding threat models, and a trichotomy of approaches to manage AI risks. Chan also emphasizes the importance of distinct communication channels for AI to enhance decision-making and promote safe interactions.
AI Snips
Chapters
Transcript
Episode notes
Agent Infrastructure Research
- Alan Chan researches interventions for mitigating risks from AI agents.
- He focuses on solutions applicable across various threat models, from spam to cyberattacks.
Traffic Safety Analogy
- Alan Chan uses traffic safety as an analogy for AI safety.
- He suggests that similar layered interventions, like roundabouts or driver training, are needed.
Trichotomy of Agent Infrastructure
- Agent infrastructure interventions can be categorized into three types: legibility, physical prevention, and identification.
- Legibility improves agent understanding, prevention limits unsafe actions, and identification enables accountability.