Details

"Physical AI: Foundation Models That Act in the Real World"

Abstract

Artificial intelligence has transformed how machines see, predict, and generate information. The next transformation is Physical AI: systems that do not merely interpret the world, but act within it. Ashutosh Saxena will explore what it means to build foundation models for the physical world. Unlike digital domains, physical environments are governed by geometry, dynamics, partial observability, and irreversible consequences. Data is expensive, failures are costly, and cloud dependent learning is often impractical due to privacy, latency, and reliability constraints, particularly in healthcare settings. Physical AI therefore demands models that can perceive, reason, and act cohesively, while learning efficiently from limited and sensitive data. I will discuss how this reshapes model training, representation learning, and loss function design, where objectives must jointly encode accuracy, safety, temporal consistency, and downstream actionability rather than static prediction performance alone. Saxena will discuss why Physical AI represents one of the largest open opportunities in computer science and engineering. As intelligence moves from screens into the physical world, it enables new classes of systems, companies, and careers, with the potential to create trillions of dollars in value across healthcare, logistics, manufacturing and beyond.

Speaker

Ashutosh Saxena, founder and chief AI officer of TorqueAGI, is a pioneering force in the field of physical AI and robotics. As the founder and chief AI officer of TorqueAGI, he is leading the development of AI foundation models that enable enterprise-grade robots to perform complex, real-world tasks in high-stakes industries. Saxena holds a PhD in computer science from Stanford University (with AI luminary Andrew Ng). Saxena became an accomplished professor at Cornell University, where he led the "Wikipedia for Robots" project (MIT 10 technology breakthroughs). With over 20,000 citations and more than 100 published papers, his foundational work on 3D computer vision and deep learning has been recognized with numerous awards, including the MIT Technology Review TR35 Innovator Award, a Sloan Fellowship and the World Technology Award. His entrepreneurial experience includes co-founding Katapult (NSDQ: KPLT), Brain of Things ($8 million ARR) and Caspar AI (Top 100 AI Companies by CB Insights).