Victor Florez, “The Use of Generative AI as a Supporting Tool in Research in Applied and Interdisciplinary Mathematics”
Mentor: Peter Hinow, Mathematical Sciences
Poster #145

Generative Artificial Intelligence (GenAI) will impact teaching and research at universities worldwide for many years to come. It is therefore important to discover the utility of GenAI in the areas of which it will have the greatest impact. In this work we explore the use of GenAI as a tool to draw connections between mathematical papers and the creation of programming code from simple prompts. The underlying research topic is the spatial distribution of space debris in Low Earth Orbit (LEO). We have selected four different Large Language Models (LLM), namely OpenAI (ChatGPT-4), Google (Gemini), Perplexity (Perplexity Pro), and Anthropic (Claude 3 Opus). These LLMs are selected due to their differences in model structure and application. The varying model architecture and application selection is to determine which LLM type is most suitable to which task. To these LLMS we feed a scientific paper and ask the same set of questions from a prepared list. We rate the responses and determine which LLM is most suitable for which aspect of the research. These chat logs are saved and stored in a GitHub repository. This ensures the chats are securely saved for reference in the future. After the preliminary questions to all LLMs, we explore deeper individual and adapted conversations. One early promising result is a research idea that was not contained in the original paper that was fed to the ChatGPT-4, namely the possibility of exploiting orbital resonances for the passive removal of space debris. This early result is impactful as it shows an LLM can give a researcher a new area to develop ideas.