This issue of Nature Computational Science explores the transformative impact of large language models (LLMs) across various scientific fields while highlighting their potential challenges. LLMs are revolutionizing research practices, enhancing literature synthesis, hypothesis generation, and experimental design in disciplines such as life sciences, chemistry, and engineering. Expert perspectives in this issue discuss LLMs’ roles in cognitive science, urban planning, and the humanities, showcasing their ability to enable new research inquiries and streamline workflows. However, significant risks accompany these advancements, including unclear research goals, potential biases, and the energy-intensive nature of LLM training. Experts emphasize the importance of ethical frameworks, validation strategies, and innovative hardware solutions, such as in-memory computing, to ensure that LLMs promote sustainable and equitable scientific exploration. This Focus aims to underline LLMs as essential catalysts for interdisciplinary innovation while addressing the prevailing challenges in their development and application.
