Explore the challenges and potential of foundation models in sequential decision-making through this insightful talk by DARPA program manager Alvaro Velasquez. Delve into the differences between natural language processing and reinforcement learning applications of foundation models. Examine the unique obstacles faced when developing foundation models for planning and decision-making tasks. Gain perspective on the critical issue of transfer learning in sequential decision-making contexts. Consider the future research directions and investments needed to advance foundation models beyond language processing. Learn from Velasquez's expertise in neuro-symbolic AI and his experience overseeing machine intelligence initiatives at the Air Force Research Laboratory. Understand the implications of these technological developments for embodied autonomy and artificial intelligence more broadly.