‘Everything is Going to Be Robotic’ Nvidia Promises, as AI Gets More Real

In this talk

1. 0:45

   Robots and AI inside Nvidia

   Huang predicts that physical AI will bring robots into factories, where machines will build products that are themselves robotic. He also says Nvidia uses AI to explore chip designs, optimize software, and identify code bugs, and wants the company to use AI throughout its operations.

2. 2:56

   AI advances in robotics and simulation

   Nvidia presents multimodal models that help robots learn from demonstrations, while Jason Ma says AI can also improve low-level control by running many experiments in simulation. The examples include a robot dog learning to balance on a moving yoga ball. The segment also shows digital humans with realistic speech and lip syncing, and Earth-2 simulations for forecasting local wind and weather effects around buildings.

3. 8:24

   Automation and employment risks

   A graphic designer says AI now recreates designs from templates he developed over six years, reducing work that took half an hour to roughly 30 seconds. Anthropic chief of staff Avital Balwit predicts that employment could change sharply within three to five years. Carsten Young cautions that outcomes depend on policy and organizational choices, and warns that productivity gains may widen wage and wealth gaps.

4. 12:37

   AI limits and uncertainty

   An OpenAI report says disinformation campaigns using its models did not gain significant audience reach, in part because their spam content was ineffective. The host points to Scale AI benchmarks, where GPT-4o is not far ahead of other models, and recommends judging systems against specific use cases. The segment closes by noting that AI research is expanding across academic fields, while its effects on work and society remain difficult to predict.