The recent advances in Generative Artificial Intelligence have led to the widespread adoption of Large Language Models (LLMs) and their integration into our lives through a broad array of applications. Despite these achievements and their remarkable performance, foundational LLMs are prone to certain limitations such as generating often non-factual but fluent and plausible content (hallucinations) or biased content, inconsistency, misunderstanding human instructions, limited knowledge of the underlying reality, non-access to real-time (recent) information, toxicity and limited reasoning.

Open Science could benefit from the AI landscape by unleashing the power of AI and revolutionizing the way we interact and engage with Science, by harnessing the volume, velocity and variety of scientific data and the vast amount of research artifacts (publications, datasets, software, methods, tools, problems, claims etc). In this talk, we discuss several steerable use cases and controllable scenarios on this synergy between AI and Open Science while considering the epistemological and normative aspects.