Alberto Cazzaniga
researcher MAASAI team Inria Côte d’Azur and Area Science Park, lecturer at 3IA Institute Université Côte d’Azur
Alberto Cazzaniga works to make advanced artificial intelligence systems more understandable and reliable. He is a researcher in the MAASAI team at Inria Sophia Antipolis, within the 3IA Côte d’Azur Institute at Université Côte d’Azur, currently on leave from his position at Area Science Park in Trieste, where he directed the Laboratory of Data Engineering. With a background in mathematics, his research focuses on the interpretability of large language and multimodal models, bridging geometric and mechanistic approaches. He studies how information is organized within these models, which internal components drive their behaviour, and how targeted interventions can improve their robustness. His work also includes selected applications in science and biomedicine. His broader aim is to help make AI systems more transparent, controllable, and reliable.
Past talks
Sat. 26 September
at 11:30 - 12:45
11:30 - 12:45
Saturday 26 September
Salone di Rappresentanza
Ingresso da Piazza Unità d’Italia 1, Trieste, TS
Ingresso da Piazza Unità d’Italia 1, Trieste, TS
TRIESTE
Artificial intelligence is everywhere – from chatbots to medical diagnosis – but one question remains: do these systems really think, or do they simply simulate thought? The panel brings together two perspectives. An expert in the cognitive science of AI asks whether machines can understand and reason as we do. A neuroscientist of artificial systems, working on mechanistic interpretability, takes us inside the ‘black box’ of these models. A conversation across philosophy, cognitive science and advanced research about what it really means to ‘think’.
Location: Salone di Rappresentanza - Ingresso da Piazza Unità d’Italia 1, Trieste
speakers
