Physical AI Needs More Data Than OpenAI Ever Had | Ashutosh Saxena, TorqueAGI

One model solving many digital problems works. Doing the same in the physical world is a different order of problem. Book Ian for your event → iankhan.com | booking@iankhan.com SUMMARY Ashutosh Saxena explains what separates digital AI from physical AI: LLMs learn one model that solves many problems, while robots operate in wildly different physical contexts. He argues the approach many companies are taking is infeasible, and that the data required is beyond the scale OpenAI worked with. KEY FINDINGS • LLMs learning one model that solves many digital problems • Physical AI attempting the same thing in the physical world • Robots operating across radically different physical contexts • The approach many companies take being infeasible at scale • Robots learning from each other across deployments • Data requirements beyond the scale OpenAI worked with • How that data generation could be accelerated NOTABLE QUOTES "There is a certain amount of data needed which is off the scale of what OpenAI worked with." — Ashutosh Saxena FEATURING Ashutosh Saxena, Founder & CEO, TorqueAGI FULL TRANSCRIPT iankhan.com/futurist/ashutosh-saxena-torqueagi-physical-ai Recorded at the Humanoids Summit for The Futurist. ──────────────────── ABOUT IAN KHAN — ROBOTICS & AI KEYNOTE SPEAKER Most AI keynotes end with a better understanding. Mine end with a measured readiness score and a ninety-day plan. Your leadership team is assessed before the event, the room sees its own collective result live, and ev