AI Safety
We develop rigorous techniques for building safe and trustworthy AI systems, ensuring confidence in their behaviour and robustness, and facilitating their successful adoption in society.
Developing AI systems that perform reliably under diverse and unexpected conditions, including outlier detection, adversarial robustness, and anomaly detection.
Ensuring AI systems operate effectively at large scales by employing efficient algorithms, resource management strategies, and distributed AI technologies.
Creating AI systems that are understandable and interpretable through Explainable AI techniques, model interpretability, and transparency mechanisms.
Evaluating societal impact of our research to anticipate and mitigate potential harms, contributing to the development of AI technologies that benefit individuals and communities worldwide.
Collaborating with industry, academia, and research organizations to leverage diverse expertise and drive innovation in AI safety research.
Addressing ethical challenges in AI research, including bias mitigation, fairness, and privacy preservation, to ensure the responsible deployment of AI technologies.