Nvidia exec to discuss robotics' missing ChatGPT moment at Disrupt
Nvidia's Les Karpas will speak at TechCrunch Disrupt 2026 about why robotics has not yet had a ChatGPT-like breakthrough. He points to the lack of a large-scale dataset for physical AI, unlike the text data used for language models. The session will explore how simulation and synthetic data might help bridge that gap.
Karpas oversees Nvidia's Inception program, which connects the company to startups across robotics, automotive, manufacturing, and smart cities. His career spans architecture, manufacturing engineering, and venture capital roles at companies including iRobot and Cirque du Soleil. The Disrupt event runs October 13-15 at Moscone West in San Francisco, drawing over 10,000 tech leaders, founders, and investors.
The core challenge: language models trained on internet-scale text data, but physical AI lacks an equivalent dataset. Self-driving companies like Waymo built data through accumulated road miles. Startups are now exploring simulation and synthetic data to manufacture that scale artificially, with Karpas joined by founders from Shield AI, Colossal Biosciences, FieldAI, and Foxglove.
If robotics achieves a ChatGPT-like breakthrough, the effects could ripple across manufacturing, transportation, and daily life. General-purpose robots might reshape labor markets and household assistance, but the data gap Karpas describes suggests such a shift remains distant. The session's focus on simulation and synthetic data may accelerate progress, yet real-world reliability concerns could temper adoption. Society may benefit from automation advances, though job displacement and safety questions could emerge as physical AI matures.