This article briefly explores the evolution of scientific paradigms, tracing the transition from traditional inductive Scientific Discovery. It first provides a detailed introduction to recent advancements in the field of AI for Science, highlighting significant breakthroughs in physical science, life science, and earth science. Then, the article introduces the concept of Autonomous Scientific Knowledge Discovery Platform, detailing its system architecture across three layers: the foundational platform, the multi-agent knowledge discovery core, and the human-machine collaborative research engine. Finally, it discusses the challenges in building such platforms and emphasizes the importance of creating open, large-scale, and interdisciplinary platforms for advancing technological innovation.