SPINLAB @ MSU

Kyle Godbey, Pablo Giuliani, Josh Belieu, and I facilitated a discussion about the use of generative AI in higher education.

Our main take away messages are summarized below:

1. Tools, in general, are only useful to the skilled user. No tool, enabled by no technology, exists without some caveat or limitation that the user should bear in mind.
2. “Let it do your chores, but not your thinking.” – Maxwell (Yuchen) Cao (MSU PhD 2020)
3. Research groups should develop group-specific generative AI policies that are consistent with journal and funder policies and consider sharing their generative AI group policies as it relates to research on the ASCSN forum.

Josh presented his study of how well he could summarize a research paper written by Pablo and Kyle in comparison to a variety of generative AI models. The full slide deck is here:

I presented my observation that graduate students in my E&M course disclosed the use of generative AI on their homeworks about 10 times out of about 5000 opportunities to do so:

The key insight for me about why this disclosure rate was seemingly so low was the word “onerous” in the word cloud reaction from the audience. I never once considered that use of generative AI was so widespread and pervasive in my course that the disclosure of every use would be severely time consuming. If this is the correct conclusion, then it implies that I need to face the world as it is and seriously consider changing my course policies as well as my research group’s policies as it relates to the use of generative AI.

Having said all that, I claim that generative AI for an unskilled and untrained user is the equivalent of a loaded self-firing gun handed to a baby.