Although the current large models are already very powerful, they still rely on structured data training and machine languages like MaKang. It seems that we will have to wait until we can completely break free from dependence on human instructions for AGI to be realized.