Arizona State University is ratcheting up its AI strategy, becoming the first university to form a deal with OpenAI, the creator of ChatGPT.
At the heart of the partnership is unlimited ChatGPT-4 access free of charge for approved university members. At least to start, interested staff, researchers, faculty members, and student workers (such as teaching assistants) will have to submit proposals outlining their ideas for using the tool and evaluating its effectiveness in order to be considered. ASU plans to start accepting submissions on February 1.
Since ChatGPT’s release in November 2022, the tool has largely been left to individual faculty members or departments to adopt and navigate. For Arizona State, a primary goal of starting — and, leaders hope, expanding — an institutional-level partnership is to help “ensure equitable access for students who are going to go into the workplace and have access to these technologies,” and to make “an investment in the future of technology innovation,” Kyle Bowen, the university’s deputy chief information officer, told The Chronicle in an interview on Friday.
Bowen declined to say how much ASU will pay OpenAI through this arrangement. Individual user accounts for GPT-4 generally cost $20 a month.
Momentum for the partnership began last summer, Bowen said, with the release of ChatGPT Enterprise. The new service introduced a solution to privacy concerns officials had: the ability for the university to have its own workspace in ChatGPT — a metaphorical “walled garden” — that pre-approved ASU community members could access with their institutional emails and passwords. Their data wouldn’t be used to train OpenAI’s models.
The University of Michigan system developed a similar “sandbox” setup last year in collaboration with Microsoft.
Bowen said Arizona State doesn’t have a limit on the number of submissions it will approve for Enterprise access. In a news release announcing the partnership, ASU expressed particular interest in the areas of “enhancing student success, forging new avenues for innovative research, and streamlining organizational processes.” Other news outlets reporting on the partnership cited curriculum development and personalized tutoring as potential focuses.
With training data off the table, the answer to “What’s in it for OpenAI?” appears to be better insights into how higher education as a sector is thinking about using generative AI. “We’re keen to learn from ASU and to work toward expanding ChatGPT’s impact in higher education,” Brad Lightcap, OpenAI’s chief operating officer, wrote in a statement.
Excitement and Risk
It’s one thing to invest; it’s another to win over the faculty. Faculty members across institutions have responded to ChatGPT with both excitement and dread, grappling with how to embrace a tool that, while bursting with potential, can also hallucinate and be used to cheat on assignments. Bowen said many ASU faculty members have been eager to work more with ChatGPT; about 20 percent have reportedly participated in a generative AI training program that ASU started offering in the fall 2023 semester.
As Bowen sees it, that’s a key indicator of interest. “People vote with their feet, right?”
Pauline Davies, president of the Tempe campus University Senate, expressed similar optimism. The development “is extremely exciting and forward looking,” she wrote in an email. “My colleagues want to use generative AI to craft the best possible learning experience for our students. We will be redesigning some aspects of teaching to make use of the technology, rather than pretend it doesn’t exist.”
According to reporting by The Arizona Republic, the university plans to form an ethics committee to monitor the partnership — one that will include faculty members. And while not specific to this latest development, ASU provides resources online to help faculty members and students use AI responsibly.
That Arizona State emerged as OpenAI’s first college partner wasn’t surprising to sources The Chronicle spoke with on Friday. Innovation, after all, is a big part of ASU’s brand and raison d’être. The institution has increasingly sought out ways to harness AI; in 2023, for example, it announced AI Acceleration, a new team of technologists assigned to create “the next generation of AI tools.”
Bryan Alexander, a higher-ed futurist, was excited to see ASU take an approach that encourages bottom-up innovation “driven by faculty, rather than imposed by a vendor or by the IT department.”
Still, risks remain. Alexander noted that OpenAI is now battling a growing number of copyright-infringement lawsuits from entities including The New York Times that threaten the foundation of its models (their training data). When asked about how Arizona State is protecting itself against these unknowns, Bowen said the university has an annual agreement with OpenAI and can evaluate terms and conditions on “a regular cycle.”
The arrangement recognizes “that the technology does change and evolve over time,” he said. “It’s designed for flexibility.”
For Ethan Mollick, an associate professor at the University of Pennsylvania’s Wharton School who studies the effects of artificial intelligence on education, ASU’s partnership with OpenAI is “a first step.”
“The partnership is interesting,” Mollick said, but “the tool itself doesn’t do anything, right? It has to be combined with other things,” like curricula, to get at that transformation piece. “That’s the bigger question.”
In other words, meaningfully transforming education will require much more than providing access to GPT-4.
