The article reports on the successful experiment of training a 42-year-old bonobo named Kanzi to play Minecraft, revealing that his learning methods are similar to those used in training GPT-4 agents. Through techniques such as contextual reinforcement learning, RLHF, imitation learning, and curriculum learning, Kanzi has mastered skills such as moving, collecting rewards, and breaking obstacles. This experiment demonstrates the advantages of animals in perceiving and adapting to new environments, surpassing the capabilities of AI agents.