Liam Carroll is a mathematician, musician, and hiking guide bringing a multi-disciplinary perspective to AI safety research and education.
At the Gradient Institute, an independent non-profit research organisation, Liam works to deepen our scientific understanding of AI systems while creating public education programs that help people grasp the capabilities and risks of advanced AI in clear, accessible terms. He is also a Fellow at Timaeus, which aims to make breakthrough scientific approaches to aligning AI with human values and intentions.
Despite the extraordinary attention on AI, we understand surprisingly little about the inner workings of the machines we're building and why they behave the way they do. As these increasingly capable systems become embedded in society and touch more aspects of our lives, ensuring they will act safely and ethically becomes critical. Liam's work addresses this challenge in two ways: conducting mathematical research on Developmental Interpretability to better understand how AI systems function internally, and developing sociotechnical frameworks to guide their responsible deployment in businesses and society.
In a strange way, Liam's work helping people navigate complex AI terrain with public education isn't too different from his work as a wilderness guide on Tasmania's Overland Track. For two years, he walked among ancient alpine rainforests and endless buttongrass plains, helping people connect with themselves and find meaning in something bigger than what we can comprehend. If we do ultimately create Artificial General Intelligence that can do most of the things that humans can do, then what does it even mean to be human?
Liam holds a Master's in Mathematics and Diploma in Clarinet Performance from The University of Melbourne.