Shinobi Skills: Mastering the Art of Prompt Engineering
Target Audience: Librarians, archivists, developers, and information professionals interested in exploring the practical applications of large language models (LLMs) in library and information science contexts. No prior coding experience is required.
Workshop Description: Step into the world of the shinobi (aka, ninja), the legendary agents of stealth and strategy in feudal Japan, and discover how their skills translate to the modern art of prompt engineering. This interactive, hands-on workshop will empower you to become a "Prompt Shinobi," mastering the techniques to effectively communicate with and control powerful LLMs.
Why Shinobi? The shinobi were renowned for their adaptability, resourcefulness, and precision—qualities essential for successful prompt engineering. This workshop draws inspiration from their tools and techniques to create a fun, engaging, and memorable learning experience.
Workshop Objectives: By the end of this workshop, participants will be able to: - Understand the fundamental principles of prompt engineering and their relevance to library and information science. - Craft effective prompts for various tasks, including information retrieval, text summarization, code generation, and creative writing. Apply defensive and offensive strategies to mitigate risks and maximize the potential of LLMs. Explore real-world applications of prompt engineering in library settings. Collaborate with peers and share knowledge in a supportive and engaging environment.
Workshop Activities: Introduction to Prompt Engineering: A brief overview of prompt engineering, its core principles, and its potential impact on researchers and libraries. Historical context: Exploring the connection between the skills of the shinobi and the art of prompt engineering.
Kata Challenges: A series of hands-on exercises focused on mastering fundamental prompt construction techniques. Participants will practice crafting prompts for different tasks and LLMs, receiving immediate feedback and guidance.
Kemuridama Challenge: The Art of Misdirection: A challenge focused on using "misdirection" techniques in prompts to achieve specific goals, such as bypassing safety filters (for red-teaming efforts) or generating creative outputs indirectly.
Tessen Challenge: The Art of Defense and Offense: Participants will learn how to craft prompts that balance defensive strategies (mitigating risks) with offensive strategies (achieving specific outcomes).
Ninja Missions: Collaborative, team-based challenges that simulate real-world library scenarios. Participants will work together to solve problems and achieve objectives using prompt engineering techniques.
The Dojo: Dedicated time for open discussion, Q&A, and sharing of ideas and resources.
Workshop Format: This workshop will be highly interactive, with a mix of: Short presentations and demonstrations Hands-on exercises and challenges Small group activities and discussions Individual experimentation and reflection
Materials: Participants must bring their own laptops (tablets and phones will create a suboptimal experience.)
Why this Workshop is Important: Prompt engineering is rapidly becoming an essential skill for librarians and information professionals. This workshop provides a unique and engaging opportunity to build these skills in a fun and supportive environment, drawing inspiration from the rich history and culture of the shinobi.
Expected Outcomes: Participants will leave this workshop with a solid foundation in prompt engineering and the confidence to apply these skills in their own work, improving their ability to leverage the power of LLMs for the benefit of their libraries and communities.
Attendee Requirements: You need a robust laptop that will handle multiple browser tabs, must be comfortable with data privacy issues that may arise in the creation and use of accounts for one of the premiere LLM/multimodal generative AI platforms, such as ChatGPT, Claude, Gemini, or Copilot. Free versions are fine, particularly if you pair up with another participant in case you hit limits. No coding skills required.