Since the launch of Boston University’s first generative artificial intelligence platform in May, the University has been assessing the platform’s use and performance.
Created by BU’s AI Development Accelerator team, TerrierGPT acts as a gateway for students, faculty and staff to access different generative AI models, including ones by OpenAI, Meta and Google, said John Byers, AIDA’s co-director of academics.
Statistics from September suggest TerrierGPT has gained around 8,000 users since its launch, AIDA’s Co-director Yannis Paschalidis said. About 2,000 of those users are faculty and staff, and the rest are students, he added.
Paschalidis said “quite a few people” are using TerrierGPT “on a daily basis,” with the platform averaging 8,000 prompts a day.
“Student pickup … has steadily increased throughout the semester, as people become aware of the existence of TerrierGPT and how they can use it,” he said.
To measure the effectiveness and impact of TerrierGPT and other generative AI models on the BU community, AIDA organized panel-style discussions for faculty members to share how they’re implementing AI in the classroom, Paschalidis said.
Byers said the BU community has expressed a desire for TerrierGPT to offer the most updated versions of generative AI models.
“It’s not always up to date with the absolute highest end, expensive versions,” Byers said. “Keeping it current is really important.”
Byers said the team has found people are finding TerrierGPT to be useful, and students appreciate the platform is free through the university.
Although TerrierGPT is free for users, Byers said he is “pleasantly surprised” it hasn’t been costly to maintain.
With the launch of TerrierGPT, faculty members are outlining AI policies in their syllabi and supporting ethical use — like Margaret Wallace, associate professor of the practice in media innovation at the College of Communication.
“[My] students are, in certain contexts, certainly encouraged, [but] not forced, to use generative AI tools,” Wallace said.“If it helps them with some of their workflow or some of their creative projects or research, it’s totally fine for them to use it.”
Wallace said while her students can use any AI platform they chose, TerrierGPT is included as an option.
“Every semester, I have my graduate assistant keep this giant list of generative AI tools, and certainly, TerrierGPT is one of the tools that we list,” she said.
Lisa Liberty Becker, master lecturer in COM and director of the COM Writing Center, said during her copyediting course, she created an exercise where students used TerrierGPT to see if it would catch all writing errors.
“We tested it out trying to break it, basically,” Becker said. “It was a larger, big-picture discussion, asking ‘What are the tools not catching?’ [and] ‘What do we need to be aware of?’”
Becker said since TerrierGPT is not perfect, editing skills are still essential.
“We still need to know these fundamentals in the context of these tools,” she said. “So you almost have to be a better copyeditor now, because the tools are not infallible.”
She added that discussions around generative AI tools do not pertain to whether they are necessarily “good” tools, but rather whether students are losing the “thinking process” behind writing.
Zoey Schlueter, a sophomore in COM, said her professors encourage generative AI use in class through different activities — but not necessarily in independent creative work.
“[TerrierGPT] is a necessary step for Boston University to take if they want to remain relevant in this growing technological age,” Schlueter said. “I’m glad that they’re taking the initiative to incorporate that into our learning environment, because I think it’s gonna be a large part of hirability in the future.”
Sophomore Madeline Wines said that although she has used TerrierGPT to help formulate emails, she is concerned about its environmental impact.
“While it is a really great platform with a lot of different models combined into one, I am super hesitant about the potential environmental impacts that it may have,” she said. “Encouraging student use, I think we should also encourage student awareness about these impacts.”
Among its goals for improvement, Paschalidis said AIDA plans to integrate “enhanced” privacy protections “at the HIPAA level” to support research conducted on BU’s Medical Campus.
“The goal is to also have models that have HIPAA protections, so that people in the medical school could more easily provide the model clinical data without concerns about violating privacy rules,” he said.
Moving forward, the TerrierGPT developers plan to update the platform, ensuring the BU community has access to the latest technologies.
“This will be a continuing process,” Paschalidis said. “As models evolve, the plan is to be updating the set of models that are offered through TerrierGPT.”
