Navigating the AI Safety and Governance Labyrinth with Andrew Harrison

Anjon Roy welcomes AI Governance expert Andrew Harrison to Paradigm Shock to discuss the complex challenges of AI Safety and Governance. Harrison shares insights from his multidisciplinary educational and professional journey, emphasizing the importance of synthesizing perspectives from psychology, data science, technology and international relations. The discussion spans emerging risks associated with the wide distribution of agent deployment, model scheming, technical solutions like mechanistic interpretability, the pivotal role of constitutional AI frameworks, and cyber risks from potential uplift scenarios. The episode underscores the transformative potential and risks of AI and the necessity of understanding its impacts on economic systems, international relations, and organizational dynamics. Harrison calls for a nuanced understanding of emergent risks, advocating for a balanced approach that draws from both technical and policy-driven perspectives to address AI's potential and perils.LinksAndrew’s LinkedIn: https://www.linkedin.com/in/andrewphilipharrison/BlueDot Impact AGI Course: https://bluedot.org/courses/agi-strategyContact ShowNew Substack- https://aiuncertainty.substack.com/[email protected]