Exploring the Impacts of ChatGPT on Future Scientific Work

Abstract

Large Language Models (LLMs) such as ChatGPT look set to transform many aspects of knowledge work, including scientific research. Anticipating the full impacts of LLMs on future science requires studying their use within the work context, motivating our study to explore scientist experiences and perceptions. We surveyed scientists specialising in AI and technology innovation, who had used ChatGPT in their work, on their perceived implications of LLMs for future science. Our findings indicated scientists were, on balance, optimistic about the potential for ChatGPT in science. ChatGPT’s high quality writing and coding output, versatility and interactivity were identified as key strengths applicable to the science domain. Participants believed LLMs would increase scientists’ productivity and their ability to access new knowledge and skills. Nevertheless, most were also concerned about the risk that ChatGPT’s persuasive hallucinations, particularly on technical subjects, could compromise the quality of scientific output. The findings provide insight into future skillsets required by scientists and ways LLMs can be harnessed to improve the efficiency, innovation, and impact in science.

Ähnliche Arbeiten