An interview with Professor Ricardo Baeza-Yates on building AI Agents in a responsible way

Professor Ricardo Baeza-Yates discusses the challenges of building responsible AI agents, emphasizing bias amplification through sequential actions, cultural differences in trust and transparency expectations, and context-dependent communication of AI limitations. He warns against deploying autonomous agents in high-stakes domains like justice and recruitment where data inadequately represents reality, advocating for human judgment in complex scenarios. Key recommendations include conducting risk impact assessments, studying AI ethics deeply, forming interdisciplinary teams, and incorporating logical reasoning with fact-checking into language model architectures.