Learning from robotic blunders - how are artificial intelligence can often fail and what we're doing to improve it. Robots can fail at university entrance exams, landing on planets and even navigating daily life. So how do we make them better?