Susan Kelley is a Senior Technical Content Strategist for Medidata AI, a clinical trials data analytics company. Before coming to Medidata Susan was an Information Architect at Broadcom and a professor at Carnegie Mellon University where she taught Rhetoric and Human-Computer Interaction. Her research and teaching focus is in business communication, specifically how women communicate and how to negotiate successfully in the technical community. She writes a frequent column on Substack called “The Pursuit” where she analyzes current events, and is frequently published in a variety of literary and popular journals. She was recently featured in the Yale University Medical journal, “The Perch” where she wrote about misconceptions about addiction and recovery and its impacts on decision making. Susan has a forthcoming book titled “Talk Like a Girl: Fierce and Feminine Language in the Workplace.” She currently lives near Annapolis with her husband and two dogs, Atticus and Mack.
You Can Tune a Piano – You Can Also Tune a Prompt
By now, many of us are already effective at generating prompts to create output in a variety of artificial intelligence tools. Where we are still growing our AI muscles is in prompt engineering, or “prompt tuning” to get the results we want. There are no tricks to this trade, but there are effective editing tools, just like the edits and rules of grammar we learned as early writers. Those are the rules attendees will walk away with in this session.
Anyone who has been to one of my sessions at STC knows that I am a fast-paced and fun speaker who leaves the audience energized with lots to talk about. This session is no different. While the prompt tuning field is changing rapidly, it is also developing a set of guardrails that attendees can learn as quickly as we all learned when to use a modal verb.
This session will introduce learners to a variety of prompt tuning methods and outcomes and will prepare them to add another skillset to their resumes upon completion. See you in Atlanta!
In this session, attendees will learn:
- Clear and concise ways to clean up prompts that are too broad or too specific to get desired results
- Effective ways to help models understand the situation or background relevant to initial prompts
- Sample inputs and resulting outputs from strong, effective prompts in a variety of Gen AI tools
- Zero-shot, single-shot, and multi-shot prompting and tuning for the best results, and what those terms mean
- Tailoring prompts to suit specific domains, then re-tuning the same prompts for cleaner results
- Excellent ways to work with leadership on trust-building, ethical considerations, and building your role as a prompt engineer and tech writing team thought leader
