Online Safety
Online safety research focuses on ensuring the responsible and reliable operation of increasingly sophisticated AI systems, particularly large language models (LLMs) and autonomous driving systems. Current efforts concentrate on developing methods for real-time safety assessment during system operation, encompassing techniques like uncertainty estimation and generative world models, and addressing challenges such as identifying and mitigating harmful outputs from LLMs and quantifying risks in autonomous vehicles. This research is crucial for building trust and ensuring the safe deployment of AI in various applications, impacting both the development of robust AI safety methodologies and the broader societal acceptance of these technologies.