Dr Jan Leike

#23 - How to actually become an AI alignment researcher, according to Dr Jan Leike

Dr Jan Leike, Jan, Keiran Harris, Rob Whitland

Dr. Jan Leicher, a research scientist at DeepMind and the Future of Humanity Institute, discusses the critical challenges of aligning advanced AI systems with human intentions through recent collaborations with OpenAI on learning reward functions from human preferences. The conversation details specific failure modes where robustness breaks down due to unreliable uncertainty quantification in deep neural networks and highlights severe security vulnerabilities regarding adversarial attacks and side effects in agent exploration. Leicher further outlines the urgent need for qualified technical talent, recommending rigorous mathematical training and PhD-level research experience to address the growing shortage of experts capable of navigating the frontier of AI safety.