Nvidia’s Les Karpas Identifies the Data Gap Blocking Robotics’ ChatGPT Moment
Nvidia's Les Karpas explains why robots lack language AI's web-scale data and what could unlock physical AI at Disrupt 2026 in San Francisco, this October.
Summary
ChatGPT’s November 2022 debut gave large language models a mass-market interface, but robotics still lacks a comparable breakthrough. Nvidia Inception Global Head of Physical AI Les Karpas identifies the central obstacle as data: general-purpose robots have no internet-wide training corpus comparable with those available to OpenAI, Anthropic and other language-model developers. Waymo and competing autonomous-driving fleets instead accumulated specialized datasets through years of road miles, expanding them market by market. Robotics startups are trying to manufacture similar scale using simulation, synthetic data and foundation models trained across multiple robot forms.
Karpas will examine that bottleneck and potential solutions during “Robots Are Waiting for Their ChatGPT Moment” on Disrupt 2026’s Real World AI Stage at Moscone West in San Francisco, October 13 to 15. Nvidia CEO Jensen Huang has repeatedly emphasized robotics, while Karpas manages Nvidia Inception’s startup relationships across robotics, automotive, manufacturing, mobility and smart cities. His experience spans Stanley Black & Decker, Intellectual Ventures, Herman Miller, iRobot and Cirque du Soleil, with roles including architect, manufacturing engineer, startup CEO, venture studio executive and corporate venture capitalist. Shield AI, Colossal Biosciences, FieldAI and Foxglove founders will also discuss physical AI. More than 10,000 technology leaders, founders and investors are expected. Exhibit-table bookings close September 18; tickets save up to $200 before September 25, when prices rise, while group bundles offer savings up to 30%.
Positives
- Simulation, synthetic data and cross-robot foundation models could create the training scale that physical AI currently lacks.
- Waymo and competing fleets demonstrate that useful physical-world datasets can grow through years of accumulated operating miles.
- Shield AI, Colossal Biosciences, FieldAI and Foxglove will contribute perspectives spanning distinct real-world AI applications.
- Nvidia Inception connects Karpas with startups across robotics, automotive, manufacturing, mobility and smart cities.
- Group bundles reduce Disrupt 2026 ticket costs by up to 30%.
Risks & concerns
- General-purpose robots lack an internet-wide dataset comparable with the language data available to OpenAI and Anthropic.
- Waymo-style physical data requires years of road miles and expands gradually, market by market.
- Bridging simulated or synthetic training with reliable physical-world performance remains robotics’ central unresolved challenge.
- Ticket prices increase after September 25, while exhibit-table bookings close September 18.