In the pre-AI world, scientific progress was incremental, with gradual discoveries leaving many questions unanswered. Is rapidly advancing AI positioned to solve all the remaining questions within the next several decades? Only two years ago, writing an original essay seemed uniquely human. Then ChatGPT emerged, demonstrating that AI can produce human-like essays. Now, we wonder: How far will this go? What will AI be able to accomplish in the future? This perspective article discusses a fundamental and imminent shift in the way science is conducted, as research is set to become the first intellectual activity largely transitioned from humans to machines. This is because all intellectual activities require reasoning, creativity and imagination, while science uniquely demands a comprehensive knowledge of the vast and ever-growing body of scientific data as a foundation for effective experiment planning. Humans, however, are inherently limited in their capacity to acquire and retain such extensive knowledge. This is precisely where AI is expected to gain a decisive advantage over humans, as it can rapidly assimilate information. However, full advantage of the future AI ability to brainstorm through the entire scientific knowledge can be reached only when its reasoning abilities at least reach human intellect. Such convergence of comprehensive scientific knowledge and more advanced reasoning is projected to create a superior AI scientist, likely by the end of this decade, when AI is expected to match human-level reasoning. Moreover, AI is even expected to surpass human reasoning ability in the following decade, further emphasizing its superiority in scientific discovery and innovation. As AI is poised to rapidly permeate scientific research, this article not only presents a point-of-view but also introduces the fundamental concepts of artificial intelligence that all scientists, regardless of their field, urgently need to master.
Marko Radulović (Mon,) studied this question.