Highlights from an interview with AI pioneer Fei-Fei Li: What's the biggest bottleneck for robots right now? In the a16z podcast featuring Fei-Fei Li, co-founder of World Labs, along with Justin Johnson and Ben Mildenhall, Martin Casado asked: What’s the connection between world models like Atlas and robotics? Fei-Fei Li’s response was straightforward: The biggest bottleneck for robots right now is data. In the future, chips might become a limiting factor, but for now, the most critical need is training data that covers real-world environments and unexpected scenarios. Cables bend in different ways, boxes come in various sizes, colors, and positions. Robots can’t just memorize one type of environment. The value of Atlas lies in transforming real-world scenarios into a simulated world that can continuously evolve and be repeatedly trained, speeding up the cycle of 'reality to simulation, simulation back to reality.'