Co-PI: Fabio Patrizi (Università La Sapienza di Roma)
AI techniques are already widespread in everyday use, however their use is often limited to systems considered low-risk, such as search engines and recommendation systems. Recent advances in artificial intelligence algorithms have shown their potential for use in a variety of other applications, including safety-critical systems such as autonomous vehicles and high-cost industrial processes such as power plants, scientific experiments, public administration. Deploying AI systems in high-risk applications requires algorithms and tools to meet high-quality requirements. This line of research aims to characterize the various qualities of interest for AI-based systems, their measurability, evaluation and certifiability, in all components of AI including perception, actuation, learning, modeling, reasoning and planning.
