Will World Models Solve Physical AI? — with Christian Rupprecht & Tolga Kart
World models are being lauded as the next major unlock for physical AI — but what are they, and what do they actually do? This fireside chat digs into the real challenges holding back functional, mainstream robotics development and adoption, and asks whether world models are truly the key to solving them, or just the latest promising piece of a much bigger puzzle.Just as LLMs defined the last wave of AI progress, we believe robotics represent the next frontier.
Yet despite record levels of investment and new technologies, teams are still hitting significant hurdles in training, manufacturing, and deploying autonomous robots. Understanding why this space remains so hard, and what the industry needs to do differently, is critical to turning our current momentum into real, sustained progress.
Developing autonomous robots is a genuinely hard problem, and despite this roundtable's title, there's likely no single technology that "solves" it all. But if training robots is really a core bottleneck blocking mainstream deployment, are world models actually the right way to break through this challenge?
In addition, world models are so nascent that there isn’t an agreed upon definition of what world models are or the best way to build them. So is the technology even there yet, or are we still betting on what’s ahead?
The main issues we’re talking through should be extremely relevant to those building in robotics, who likely are encountering these challenges themselves. The topic should also be interesting to investors, as we’ll dig into how this impacts the growth of the physical AI industry.
Not really sure yet, but we could imagine maybe taking some polls throughout the discussion, like whether audience members have heard about “world models” (it’s likely they have, as it’s such a buzzy term these days), but for those who have, whether they truly understand what they are?
Speakers:
Christian Rupprecht — Chief Research Officer, Third Dimension AI / University of Oxford. Computer vision researcher affiliated with Oxford, now leading research at Third Dimension AI.
Tolga Kart — Third Dimension AI.