A deep dive into a high-tech table tennis robot: 17 distinct shots, adaptive strategy, and real-time exploit of opponents' weaknesses. We trace its training—from millions of simulated plays to seven cycles of iterative grounding—and how researchers bridged the sim-to-real gap with domain randomization. Against 29 human players of varied skill, the robot wins some rounds and exposes limits, revealing what it takes for machines to learn, adapt, and compete in a sport that still rewards human intuition.Note: This podcast was AI-generated, and sometimes AI can make mistakes. Please double-check any critical information. Sponsored by Embersilk LLC

Ver todos los episodios