UC San Diego Launches Center for Principled Agentic Intelligence (today.ucsd.edu)

UC San Diego has announced the launch of the Center for Principled Agentic Intelligence (CPAi), an interdisciplinary initiative aimed at addressing the alignment, evaluation, and oversight of advanced AI systems. The center will integrate expertise from data science, computing, and the arts and humanities, with a focus on ensuring that increasingly autonomous AI agents operate according to human values and intentions. Supported by $1 million over three years and additional computing resources, CPAi will explore critical questions about trust and accountability in AI systems, emphasizing the need for rigorous evaluation methods. CPAi will tackle two main initiatives: "AI Alignment and Society," which investigates how AI training and evaluation processes influence their behavior, and "Agentic AI System Evaluation," aimed at developing an open-source testbed for assessing AI's performance in real-world tasks. The center's work underscores the growing importance of incorporating humanistic perspectives in AI research to foster responsible innovation, balancing technological advancement with societal needs. As AI capabilities expand, CPAi aims to cultivate a framework that not only enhances our understanding of agentic AI but also safeguards its alignment with human goals.