The Multicom Research Group conducts interdisciplinary research on intelligent information processing, representation, and transmission, bridging classical information theory with modern artificial intelligence.
Modern AI systems have achieved remarkable empirical success across many domains. Yet many of their scientific foundations remain poorly understood — particularly how learning systems represent information, build predictive models, resist adversarial perturbations, and protect intellectual property. Understanding these questions is not merely academic: it shapes how reliably and securely AI systems can be deployed.
At the same time, information theory, which revolutionized communications and data compression in the twentieth century, is entering a new stage of development. The emergence of large language models and embedding-based representations offers, for the first time, a concrete path toward extending Shannon's framework to encompass semantic information and intelligent information processing.
Our research pursues this frontier through a deep interaction between information theory and artificial intelligence: using information-theoretic tools to understand and improve AI systems, while using insights from modern AI to inspire new theoretical developments in information theory.
