Abstract: With the accelerated iterations of artificial intelligence (AI) technology, the level of AI continues to evolve toward higher tiers of intelligence, gradually transcending its traditional role as a technological tool to become an "AI agent" with a certain degree of innovative capability. The evolution of human-machine relationships has emerged as the core driving force behind innovation activities, giving rise to a new paradigm distinct from traditional technological and digital innovation—AI innovation. Based on relevant domestic and international literature, AI innovation can be comprehensively defined from three dimensions: technological foundation, innovation evolution, and innovation outcomes. It is a process in which AI systems, within the framework of human-defined objectives, continuously generate new combinations of elements and facilitate their value transformation through autonomous evolution and human-machine collaboration. This process inevitably triggers a transformation in the innovation paradigm. Under the core logic of aligning AI technology with application scenarios, the human-machine relationship continuously evolves, ultimately shaping three key characteristics of AI-driven innovation: subject reconstruction, dynamic evolution, and systemic transformation. These give rise to three types of AI innovation: scenario-supported, collaborative-enhanced, and autonomous-optimized, which further contribute to disruptive changes across the entire innovation process. In light of China's specific contextual needs in promoting the comprehensive transformation, upgrading, and high-quality development of the manufacturing industry, research on the development pathways, management, and governance of AI-driven innovation in manufacturing can be further deepened.