Abstract: This paper delves into the key pathways for artificial intelligence (AI) to achieve true intelligence, advocating for the deep integration of factual computation and value calculation to transform AI from a mere automated tool into an intelligent system with cognitive and ethical capabilities. While current AI excels in handling objective data, it is not good at understanding and addressing the complex value issues in human society, which may lead to decision-making biases and ethical dilemmas. The paper clearly distinguishes between factual facts and value facts, noting that the latter are based on the former and influence factual cognition. In reinforcement learning, it proposes that the design of reward and punishment functions should balance factual accuracy and value goals. In human-machine-environment systems, a dynamic collaborative framework is constructed to integrate the computability of facts and the determinability of values. It emphasizes the improvement of human-machine interaction through embodied cognition and other technologies to drive AI from functional simulation to mechanism simulation. In multi-agent systems, it analyzes the dynamic interaction of multiple facts and values, highlighting the necessity of communication and self-organization. For AI to transcend instrumental rationality and move towards value rationality, it must achieve a seamless integration of the fact and value.