"Large Language Models, but Large Compared to What?" Teaching Scaling, Similitude, and Dimensional Reasoning in the Age of Generative AI

Setting aside hypersonics, three of the fluid dynamicist's trustiest weapons are scaling, similitude, and dimensional reasoning. Instilling future generations with a healthy respect for these principles remains a central goal in fluid dynamics education. A recent Wall Street Journal article noted that generative AI will change everything; it is natural to wonder, "does 'everything' include scaling, similitude, and dimensional reasoning?" In this talk, we present tastefully curated anecdata in support of the answer, "probably." We identify broad areas of strength and weakness on this subject matter across a range of large language models (LLMs), with a particular focus on unorthodox Buckingham Pi groups identified by certain LLMs. Time permitting, this presentation will also include a live demonstration.