The University of New South Wales (UNSW) is treading carefully around generative AI applications as it maps out places to incorporate the technology into its student services.
![UNSW treads carefully on generative AI experiments](https://i.nextmedia.com.au/Utils/ImageResizer.ashx?n=https%3a%2f%2fi.nextmedia.com.au%2fNews%2funsw_2.jpg&h=420&w=748&c=0&s=0)
The Sydney-based university has started conversations to tap into the disruptive technology but said it won’t rush implementation.
Associate professor in regulation and governance at the UNSW Business School Rob Nicholls said one of the applications envisioned for generative AI is integrating the technology into chatbots that help prospective students find courses.
Speaking with iTnews at a recent Tech Leaders event, Nicholls said the university is “nowhere near [having a] minimum viable product” ready for testing, but continues to explore generative AI's potential.
As the government seeks input into the use of generative AI in the Australian education system, Nicholls said having a “fine-tuned” generative AI capability, trained on course handbooks, could reduce complexity for future students when seeking to “work out what courses to do”.
Such a capability could “match up the students to the courses they want to do.”
Nicholls characerised progress towards this capability as an "in-house drafts", noting there is no internal consensus on “what large language model to use as the base for it and how you ensure that reasonable privacy expectations are met”.
“That almost certainly means not an online service ,or if it's an online service, then an isolated instance of that online service," he said.
He said the “finetuning of the large language model” is “perhaps one of the most important parts” so course relevance and access are up-to-date.
UNSW’s approach to practical applications of the technology may be considered "risk averse”, he said.
The university didn't want to get it wrong, either in the design or implementation process.
Nicholls said universities generally were likely to be slower adopters of the technology due to being highly risk averse.