HAND Industry Day Explores the Future of Intelligent and Capable Real-World Robotics 

HAND’s 2025 Industry and Innovation Day brought together some of the most influential voices in robotics to explore where dexterous, intelligent machines are headed, and, how quickly they may enter real-world use. Across keynotes and panel discussions, a shared message emerged: recent advances in mobility, manipulation, and foundation models are accelerating progress toward robots that can operate safely, intuitively, and reliably in human environments.  

Al Rizzi, CTO of the Robotics and AI Institute (RAI) and former chief scientist at Boston Dynamics, impressed attendees by tracing three decades of progress in mobility, manipulation, and machine intelligence in a mere hour as he led the energizing keynote. Tracing three decades of breakthroughs, Rizzi illustrated the remarkable evolution from early dynamic manipulation to today’s quadrupeds capable of parkour-like agility and whole-body coordination. He highlighted the growing role of reinforcement learning and large vision–language models, which now enable robots to learn from demonstration and generalize beyond explicitly programmed behaviors. Rizzi emphasized that meaningful AI impact will come when systems unite athletic intelligence with cognitive intelligence, providing for robust physical capabilities paired with situational understanding. 

The panel on Humanoid Robotics, moderated by HAND’s Director of Industry and Innovation, Scott Ransom, expanded the conversation from technical capability to the realities of commercial adoption. Panelists Al Rizzi, Rob Ambrose (Texas A&M; former NASA Valkyrie program lead), and Fabio Puglia (CEO, Oversonic Robotics) explored how humanoid development has unfolded in cycles of rapid advance and plateau. Ambrose noted that real-world adoption will require the convergence of reliability, cost, and compelling use cases, the same pattern seen in industrial arms and autonomous vehicles. 

Rather than focusing on morphology, panelists argued that humanoids should be defined by capability: dexterity, semantic understanding, safe human-paced interaction, and intuitive collaboration. Puglia described the tension between public expectation and operational reality, especially in environments like hospitals where both technical precision and social signaling matter. The panel also examined economics and societal attitudes, emphasizing that early deployments will likely focus on simple, high-value support tasks, especially amid persistent labor shortages.  

The day concluded with Russ Tedrake’s (MIT Professor and Senior VP for Large Behavior Models at Toyota Research Institute) keynote at the Industry and Innovation Day dinner. Russ brought the day’s themes together by addressing one of the field’s central questions: does large-scale pretraining meaningfully improve a robot’s ability to learn new tasks? TRI’s research offers a clear answer: YES! While subtle implementation details such as data normalization still influence outcomes, Tedrake showed compelling evidence that broad, diverse pretraining creates more adaptable, capable robotic learners. His keynote underscored why foundation models are poised to play a transformative role in robotics. 

Taken together, these insights painted a coherent picture of a field on the cusp of real-world transformation. As researchers work to unify physical intelligence, dexterous capability, and scalable learning, HAND ERC is uniquely positioned to guide this next era of advancing the technical foundations while shaping how society engages with the future of intelligent robotic systems.