A humanoid robot learning to dribble, shield and tackle through self-play feels like a glimpse of what robot soccer could eventually become. Skild AI says its S1 model developed these behaviors through self-play reinforcement learning in simulation, after 140 simulated years of