OIT has been closely following the development of various generative AI products and evaluating their potential implications for teaching, learning, research, and university operations. Here, we can define generative AI as a technology that can generate new content similar to what they have learned from training data. Generative AI is a subset of AI that focuses on creating previously unseen outputs and can include text, images, music, and more.
You may want to take some time to experiment with some of these technologies because they represent the first step in a fundamental shift in the landscape of technology. For example, I recently experimented with ChatGPT’s ability to examine a complex dataset to demonstrate some of the capabilities of these models. To accomplish my analysis, I anonymized a very simple retention dataset and used the following prompt:
I'm uploading a file of several variables. I'm most interested in the Retained column, where 1 is retained and 0 is not retained for college students. I'm interested in finding any unique patterns or relationships. This data is anonymized.
I then conducted a conversational data analysis, including descriptive statistics, correlations, and logistic regression. The model had several insights that were to be expected but still very insightful and accurate. I used the conversation output to demonstrate how AI could be used and to highlight the importance of careful data usage. Why?
Many models use the prompts and uploads for continued training. For example, ChatGPT saves conversations by default and will use chats for future models. Unchecking “Chat history & training” prevents chat use in future models.
Several Data Governance Steering Committee members will work on policy updates, though it is worth providing some initial guidance while the updates make their way through the process.
Guidance
Below are some general guidelines that may help you experiment safely with generative AI.
Accuracy Concerns – Remember that these models produce incorrect results. They are notoriously bad at basic math.
Understand the Platform’s Privacy Settings – This is an excellent chance to read the privacy policy. We know we don’t always read those long documents, but this one may be worth checking.
Respect Copyrighted Material – Avoid uploading copyrighted material.
Datasets – Be wary of uploading datasets into the various AI models. Care must be taken to avoid uploading identifiable datasets – mainly student or confidential data.
Proprietary Information – proprietary procedures or documentation should not be uploaded.
Avoid Oversharing – Avoid sharing too much of your personally identifiable information.
Keep Learning – Don’t be scared – use caution.
Generative AI is Here
Generative AI is here to stay and will only continue to grow. Spend some time experimenting and learning about the capabilities. If you want something free, you can use the Copilot license included with Microsoft 365. Want to talk more about this topic? Reach out to me, and let’s chat!
