In this talk, Richard Tomsett, Senior Applied Scientist at Onfido, will explain the concept of robustness in machine learning (ML), looking at what ML does well, what it isn’t so good at, and the weird and unexpected ways it can fail. He will go into some details about the work being done to improve robustness, including uncertainty estimation and adversarial methods. Finally, Richard will discuss this work within the context of ML and, more broadly, AI, as they are used in the industry. He will introduce the concept of AI governance and how robustness is just part of a whole range of governance processes that are crucial in real-world AI applications.
