Researchers have developed a new artificial intelligence approach that exposes critical weaknesses in multi-agent reinforcement ...