RL in the Real World

​Reinforcement learning has enabled remarkable breakthroughs in AI, from superhuman game-playing agents to increasingly capable robotic systems. But bringing RL out of simulation and into the physical world remains one of robotics' hardest challenges.

​Unlike digital environments, the real world is messy, unpredictable, and unforgiving. Robots must learn to navigate uncertain environments, adapt to unfamiliar objects, and recover from failure—all while operating under the constraints of physical hardware. As advances in robot learning accelerate, a critical question emerges: What will it take to make reinforcement learning work reliably at scale in the real world?

​Join Savant for an evening with leading researchers and builders at the frontier of Physical AI as we explore the breakthroughs, bottlenecks, and emerging approaches shaping real-world reinforcement learning.

​Topics will include:

• ​RL Beyond Simulation: Bridging the sim-to-real gap and understanding what it takes to transfer learned policies to physical hardware.

• ​Learning from Real-World Experience: How robots can improve through trial and error, online learning, and continuous interaction with their environments.

• ​Scaling Robot Learning: Tackling the challenges of sample efficiency, reward design, data collection, and the cost of training on real robots.

• ​RL Meets Foundation Models: How reinforcement learning can complement imitation learning, vision-language-action models, and world models to unlock more generalizable robot capabilities.

• ​From Research to Deployment: What separates impressive lab demonstrations from robust, reliable autonomy in real-world settings.

• ​The Road Ahead: Which technical breakthroughs could make RL a cornerstone of general-purpose robot intelligence?

​This event is designed for robotics researchers, Physical AI founders, reinforcement learning practitioners, and frontier lab builders working to bring intelligent machines into the physical world.

​Join us as we explore what happens when reinforcement learning meets reality.

​Thursday, Feb. 25th, 2027

​Agenda:

• ​5:30 PM — Doors Open + Light Bites

• ​6:15 PM — Welcome Remarks

• ​6:30 PM — Fireside Chat

• ​7:00 PM — Audience Q&A

• ​7:25 PM — Closing Remarks

• ​7:30 PM — Networking

• ​9:00 PM — End of event

​About the Host:
Savant is a community of technologists embarking on good quests, ideas that are hard, consequential and civilizational-defining. We support founders navigate this brutal -1 to 0 phase with a rare concentration of exceptional peers, hardware lab & office space, office hours, access to our network and more — often before we ever write a check. Together, let's move civilization forward!