SOUTHFIELD–Seven Lawrence Technological University faculty members and a student engaged in a lively discussion Thursday on the latest advances in artificial intelligence and how they’ll affect the future of education at LTU.
Most of the talk, the last in LTU’s “Humanity + Technology” lecture series for the 2022-23 academic year, concerned the AI writing and research took ChatGPT. But C.J. Chung, professor of computer science, said he’d been using AI for years to generate music. Matthew Cole, associate professor in LTU’s College of Business and Information Technology, predicted that AI won’t take jobs, but will augment them.
The student representative on the panel, Caroline Holowenko, is a junior in LTU’s Technological Humanites major, and spoke about her academic concentrations in cyberpolicy law and ethics and IT security.
There were several gaspworthy moments, if that’s a word. Vivian Kao, associate professor of English, pointed out that the latest version of ChatGPT, GPT4, scored in the 90th percentile of a bar exam, while a version produced only a week earlier GPT3.5, only scored in the 10th percentile. “We have no idea why GPT4 did so much better on the bar exam,” she said. “We have no idea how it learned. There is something frightening how quickly this thing can learn what it was not built to do. What else can it do that we have not discovered?”
Meanwhile, Philip :Plowright, professor of architecture and chair of the LTU Department of Design, said he asked GPT for help in researching a topic he wanted to write an academic paper about. He said several searches produced details on 125 scholarly articles by academics he had heard of, in peer-reviewed journals he knew existed. But when he double-checked the references, and “out of 125 references. ZERO were real. The AI had created a whole series of documents for me based on what I wanted to see, not because those things existed in reality.”
Eric Meyer, associate professor of biomedical engineering, said that there were concerns, but that AI would soon help engineers do better and faster work, especially in 3D modeling of designs.
And moderator Julie Kiernan, assistant professor of communication and director of LTU’s technical communication program, wondered whether anyone had considered that all the energy used in just trying things out with hugely powerful computing was bad for the environment.