Foundation Models for Scientific Discovery: From Paradigm Enhancement to Paradigm Transition

Fan Liu, Jindong Han, Tengfei Lyu, Weijia Zhang, Zhe-Rui Yang, Lu Dai, Cancheng Liu, Hao Liu

Keywords: scientific discovery, foundation model

Abstract: Foundation models (FMs), such as GPT-4 and AlphaFold, are reshaping the landscape of scientific research. Beyond accelerating tasks such as hypothesis generation, experimental design, and result interpretation, they prompt a more fundamental question: Are FMs merely enhancing existing scientific methodologies, or are they redefining the way science is conducted? In this paper, we argue that FMs are catalyzing a transition toward a new scientific paradigm. We introduce a three-stage framework to describe this evolution: (1) Meta-Scientific Integration, where FMs enhance workflows within traditional paradigms; (2) Hybrid Human-AI Co-Creation, where FMs become active collaborators in problem formulation, reasoning, and discovery; and (3) Autonomous Scientific Discovery, where FMs operate as independent agents capable of generating new scientific knowledge with minimal human intervention. Through this lens, we review current applications and emerging capabilities of FMs across existing scientific paradigms. We further identify risks and future directions for FM-enabled scientific discovery. This position paper aims to support the scientific community in understanding the transformative role of FMs and to foster reflection on the future of scientific discovery.

Lay Summary:

Foundation models such as GPT-4 and AlphaFold are transforming how scientists work. These powerful AI systems can read, write, and reason across many forms of data, helping researchers generate hypotheses, design experiments, and interpret results faster than ever before. But their influence raises a deeper question: are they simply improving current scientific methods, or are they changing what it means to do science?

This paper argues that foundation models are driving a fundamental shift in the scientific process. We propose a three-stage framework to describe this evolution. First, foundation models integrate into existing research workflows, enhancing efficiency within traditional disciplines. Next, they become true collaborators that help humans frame problems, reason through evidence, and make discoveries. Finally, they may evolve into autonomous scientific agents capable of producing new knowledge with minimal human guidance.

By examining how these models are already being used across fields and discussing their emerging capabilities, risks, and implications, our goal is to help the scientific community understand how FM could redefine discovery itself and guide the responsible development of this new paradigm of science.

Submission Number: 193