This half-day short course provides atmospheric and environmental scientists—at all career stages—with practical skills for integrating large language models and related AI tools into their research and professional workflows. Participants will explore how large language models can accelerate literature reviews, refine research proposals, and adapt technical material for different audiences. Critically, we'll address when AI use is appropriate, recognize its limitations—particularly hallucinations in literature reviews and technical writing—and develop effective prompting strategies to obtain reliable results. Through interactive exercises, participants will gain practical experience applying these tools while understanding their boundaries.

The course emphasizes prompting strategies—practical techniques for structuring inputs to LLMs in ways that improve reliability, reduce hallucinations, and produce outputs that align with scientific standards. Participants are encouraged to bring their own ongoing projects to apply these techniques in real time.

To make the course as relevant as possible, participants are encouraged to bring their own research or writing projects. If they would like to prepare for a job interview, it will be helpful to bring the job description for the position they are applying to; a CV can also be useful, though participants may prefer not to share it with an LLM. The participants would need to bring a laptop and a free account for one or more of the following LLMs: ChatGPT, Gemini, NotebookLM, Claude, etc.

107th AMS Annual Meeting
Colorado Convention Center
January 10, 2027 at 8:30 AM - 12:00 PM Mountain Time (In person)

Registration for this course will open in late October.

Course Description:

Participants will:

• Use large language models to accelerate literature review, refine research proposals, and adapt technical content for diverse audiences.
• Recognize when and why LLMs confabulate citations and generate plausible but incorrect technical content, and apply verification strategies and prompting techniques that reduce these risks in scientific practice.
• Construct prompts that improve the reliability, specificity, and usefulness of AI outputs across a range of scientific tasks, distinguishing between contexts where AI augments judgment and contexts where over-reliance introduces risk.
• Assess the appropriateness of AI use for a given task, and make informed decisions about when AI tools support — versus undermine — scientific rigor and our cognitive processes.

VIEW AGENDA

If you have questions regarding the course, please contact Tsvetomir Ross-Lazarov.

Instructors: