Prompt Engineering 101 and Leveraging LLMs

Daniel Schlesinger, a MOS Surgery and cosmetic Surgery fellow at Northwestern University, presents on prompt engineering and large language models (LLMs), highlighting their rapid evolution and applications. He defines key terms like LLMs, natural language processing (NLP), and generative AI, mentioning their utility in creating diverse content beyond text, including images and code. Schlesinger notes that the effectiveness of LLMs greatly depends on the prompts provided, emphasizing the importance of asking well-crafted questions to obtain useful responses. He compares different LLMs, such as ChatGPT and Google’s Gemini, revealing discrepancies in their outputs and capabilities, particularly regarding real-time data access and the ability to provide source citations. The presentation underscores the common issue of 'hallucinations,' where LLMs may produce inaccurate or fabricated information. Schlesinger also illustrates the potential of LLMs in various professional tasks, from data analysis to creative content generation, and advises users to consider privacy and confidentiality issues, especially within healthcare contexts. He concludes by stressing the value of utilizing multiple AI tools for improved results and being mindful of how personal data is handled.