A: This consultation service aids researchers who are trying to utilize AI Tools (U-M GPT GenAI Tools or others) for research purposes.

• For Administrative workflows - contact Generative AI Center of Excellence
• For AI Tools in Teaching workflows - contact Teaching and Learning

A: Not all ChatGPT or Claude versions have been sanctioned by U-M Information and Technology Services (ITS) due to enterprise security, governance, or compliance controls and the sensitivity of the data. Public versions of these tools should only be used if the classification of the data is considered LOW.  If you are unsure of the classification of your data, you can use the following guides to help you:

• Data Classification Levels
• Examples of Sensitive Data by Classification Level
• Sensitive Data Guide

A: Complete the Specialty Software Purchases & Support Form and submit it in TDx. You will be contacted through TDx if any additional information is required. Here are some additional resources, if needed:

• Software, Licensing, and Cloud Services Procurement Compliance SPG 601.03-3
• Software Needs and Assistance

A: U-M GPT tools are free for the researcher to use however, they do come with limits. If you require more tokens to complete your project, contact LSA-TS-AI-Research (Group) to help you decide next steps, or if you are interested in purchasing a paid subscription, you can submit a software purchase request through the TDx ticketing system.

A: Yes, there are several safety concerns and potential risks associated with using Generative AI (GenAI):

• Data Privacy and Risks
• Accuracy and Hallucinations: AI Tools do not understand information in the same way that humans do. This can lead to “hallucinations” or presenting incorrect information resulting from inaccurate statistics or events, non-existent citations or out-dated answers based on the version of the tool. U-M does not recommend accepting AI output without review.
• Bias and Ethical Concerns
• Compliance and Regulatory Risks
• Human Responsibility

A: As artificial intelligence (AI) continues to gain popularity, it is essential for members of our campus community to develop skills in prompt literacy. Prompt Literacy for Academics outlines many best practices to consider when learning how to interact with GenAI. Other resources for learning about prompt engineering:

• Training: Generative AI Prompt Literacy Canvas Course
• Visual Aid: One-page format document developed by UM Flint
• U-M GenAI Prompt Library for a comprehensive resource designed to support the diverse needs of the University of Michigan community

A: At its simplest, meta prompting is "prompting about prompting." It shifts the focus from content (what to say) to structure (how to reason). Meta prompting is an advanced prompt engineering technique for people to interact with large language models (LLMs). Instead of giving direct instructions to the AI models, we are influencing their thinking process by teaching them to generate their own effective prompts.

A: Human-in-the-Loop (HITL). In a human-in-the-loop system, humans are actively and directly involved in the decision-making process or the continuous improvement of an AI system. Their input is essential for the process to function to its full potential as well as evaluating the performance.