global-journal-of-computer-science-and-technology-d-neural-ai Global Journal of Computer Science and Technology - D: Neural & AI 0975-4350 0975-4172 Global Journals Publishing Group Incorporated 54742 Leveraging Foundation Models for Scientific Research Productivity AI Foundation Models and Scientific Productivity GruetzemacherRoss UNITED STATES, Wichita State University 08 12 2023 23 D3 27 42

The objective of this work was to elucidate paths for expediting and enhancing scientific research productivity from the emerging AI paradigm of foundation models (e.g., ChatGPT). Faster scientific progress can benefit mankind by speeding up progress toward solutions to shared human problems like cancer, aging, climate change, or water scarcity. Challenges to foundation model adoption in science threaten to slow progress in such research areas. This study attempted to survey decision support systems and expert system literature to provide insights regarding these challenges. We first reviewed extant literature on these topics to try to identify adoption patterns that would be useful for this purpose. However, this attempt, using a bibliometric approach and a very high level traditional literature review, was unsuccessful due to the overly broad scope of the study. We then surveyed the existing scientific software domain, finding there to be a huge breadth in what constitutes scientific software. However, we do glean some lessons from previous patterns of adoption of scientific software by simply looking at historical examples (e.g., the electronic spreadsheet)

 Full Text

he objective of this work was to elucidate paths for expediting and enhancing scientific research productivity from the emerging AI paradigm of foundation models (e.g., ChatGPT). Faster scientific progress can benefit mankind by speeding up progress toward solutions to shared human problems like cancer, aging, climate change, or water scarcity. Challenges to foundation model adoption in science threaten to slow progress in such research areas. This study attempted to survey decision support systems and expert system literature to provide insights regarding these challenges. We first reviewed extant literature on these topics to try to identify adoption patterns that would be useful for this purpose. However, this attempt, using a bibliometric approach and a very high level traditional literature review, was unsuccessful due to the overly broad scope of the study. We then surveyed the existing scientific software domain, finding there to be a huge breadth in what constitutes scientific software. However, we do glean some lessons from previous patterns of adoption of scientific software by simply looking at historical examples (e.g., the electronic spreadsheet).