ÌÇÐÄvlog¹ÙÍø

Generative AI is a type of artificial intelligence that uses machine learning models to create new content — like text, imagery, and videos — based on a prompt. Once generative AI tools are fed and trained on datasets they learn how to form new content by identifying patterns and creating new variations based on those patterns.

This technology has been around for many years, but recently it's been newly simplified and made accessible to the mainstream.

There are many exciting uses for generative AI, including automation and productivity enhancement, content and image generation, code debugging, and much more.

• Unsure where to start? ITS holds a series of events inviting the campus community to explore various AI tools. Keep your eye out on the events calendar.
• ITS has also reviewed and compiled a list of popular generative AI tools that may be helpful for our work. 

Avoid entering confidential data, per the University’s Data Categorization Policy, into AI tools. This includes but is not limited to non-public finance, human resources and medical information, and student records.

Generative AI tools using default settings may not be private and could expose sensitive data to unauthorized parties.

Carefully examine any generated content for accuracy, bias, or copyrighted content before publishing. Don’t assume all generative AI services are safe, unbiased in their results, or completely accurate.

To ensure the safety and protection of ÌÇÐÄvlog¹ÙÍø’s data, consult with ITS before procuring internal or external AI tools for use on the ÌÇÐÄvlog¹ÙÍø network or ÌÇÐÄvlog¹ÙÍø devices. ITS will ensure the tool meets appropriate privacy and security protections.

This pilot project brought together 25 participants from the two units to experiment with their choice of genAI tools. Many librarians experimented with generating metadata for cataloging records, while members from ITS explored generative AI from the perspective of generating code and Universal Design principles.

The trial period uncovered useful insight into the use of genAI workflows: specific prompt guidance leads to better results, it’s helpful to trial multiple similar tools and compare results, and there are potential biases and limitations in relying heavily on AI for information retrieval. Several participants highlighted concerns about the human, environmental, and ethical costs that may outweigh the potential benefits of generative AI.

Quick facts and outcomes:

• AI tools tested: ChatGPT, Perplexity, Elicit, Microsoft CoPilot, GitHub CoPilot, Woodpecker, Synthesia Creator, Otter.AI, and Ollama.
• Sarah Cross, Instructional Designer for Learning Platforms and Universal Design for Learning (UDL), found AI useful in customizing prompts for educational purposes and in generating alternative assignment ideas aligned with UDL principles.
• Peter Tagtmeyer, Outreach and Engagement Librarian for the Natural Sciences and Mathematics, experimented with GPT4and integrated it with Zotero to analyze a large set of articles on Cognitive Load Theory and cognitive offloading. He shared examples of prompts and AI responses related to the "Google Effect," its impact on memory, and its relation to concepts like the Dunning-Kruger effect. Peter noted the potential biases and limitations in relying heavily on AI for information retrieval and emphasized the importance of balancing AI use with traditional study methods.