This review paper explores the transformative role of artificial intelligence (AI) in advancing scientific discovery. In the past decades, scientific research has become increasingly complex, requiring interdisciplinary collaboration and vast computational resources to address pressing challenges such as climate change, rare disease treatments, and the development of sustainable technologies. AI has emerged as a powerful tool capable of addressing these bottlenecks by accelerating knowledge synthesis, enabling advanced simulations, and fostering breakthroughs in various fields. Examples like AlphaFold in protein folding and GraphCast in meteorology underscore AI's potential to enhance both the efficiency and scope of research. This paper delves into the evolution of AI's role in science through three conceptual phases: Keplerian (data-driven pattern recognition), Edisonian (autonomous experimentation), and Einsteinian (foundational innovation). While AI has already made significant contributions, the integration of such technologies raises concerns related to ethical considerations, environmental costs, and data privacy. Addressing these concerns requires balancing technological innovation with sustainability and regulation. By reviewing recent advancements, identifying risks, and highlighting future opportunities, we provide a thorough perspective on how AI can propel scientific progress, from hypothesis generation to global problem-solving, while reshaping the landscape of discovery in the future.