Rayna Shaik

Helios Scholar

Usage of artificial intelligence (AI), the capability of machines to perform tasks typically requiring human intelligence, is exploding in scientific fields. Thus, there is a growing need to understand the myriad applications of AI within TGen and the Institute’s readiness to adapt to this technology, particularly within scientific writing. To address this need, we developed comprehensive prompting guides for high-value, TGen-centric writing tasks. To assess the efficacy of these guides, we systematically compared scientific writing outputs from Helios Scholars and experienced scientific writers to those of a popular AI: ChatGPT. This involved summarizing a complex scientific article and prompting ChatGPT to do the same. Results indicated that ChatGPT saved 104.5-155.5 minutes of human work time, on average, with the accuracy of its summaries confirmed by the Science Writing Team, the senior author, and the lead bioinformatician associated with the paper. While ChatGPT performed well overall, it struggled with the methods section, highlighting the need for targeted improvements to prompting techniques. Based on these outcomes, we developed new, comprehensive prompting guides to provide best practices and step-by-step instructions for utilizing AI in summarizing scientific publications. We experimented with various inputs, refining commands based on audience, format, and specificity, optimizing the translation of scientific content for non-specialist audiences (e.g., TGen Foundation pitch pieces). AI offers transformative potential for TGen’s scientific writing, and by expanding these guides to encompass additional tasks, we can enhance efficiency and quality, streamlining project pipelines and ultimately advancing TGen’s research to better serve patients.