It’s one of 2023’s hottest topics: artificial intelligence, or “AI.” Read any news outlet or spend some time on social media and someone, somewhere, is showing off something that Chat GPT “wrote” or that an image generator “drew.”
Science fiction becoming science reality? Not quite yet. But it’s fair to say that the power of these machine-learning tools, and the speed at which they have advanced into something approaching the fantastical, has created a mix of hype and hysteria can be hard to parse through.
Read also: Ask the experts: Where will artificial intelligence go next? (Dal News, June 5)
Within higher education, initial conversations around tools like Chat GPT have largely focused on academic integrity. Back in April, Dal’s Faculty of Open Learning and Career Development hosted a webinar on the topic, featuring a range of perspectives from across the university. One of the participants in that event was Christian Blouin, professor and associate dean academic in the Faculty of Computer Science, who is working to support faculty though multiple disruptions in recent past and recently appointed as institutional lead (AI strategy) for Dalhousie.
And while Dr. Blouin sees the academic integrity conversation as an important one, he’s also keen to broaden the AI conversation at Dalhousie into something much more holistic.
“If we assume pace of disruption is increasing — even if it stays constant — we don’t want to find ourselves, in a where we’ll constantly be criminalizing everything new,” he says. “Instead of defining ourselves by what’s not allowed, we need to be clear on what we’re trying to achieve as a university.”
Leslie Phillmore, associate vice-president academic, says the pace by which AI is impacting and will continue to affect academic work makes this a critical conversation to have now.
“Having Christian help facilitate that conversation at Dalhousie not only will give this important work a focal point but will allow us to better connect with other universities across Canada to share information and strategies,” she says.
Developing systems and supports
As for what Dal is trying to achieve with AI, that’s a conversation with Dr. Blouin right at its centre. For the next couple of years, a portion of his time will be spent consulting with staff and faculty, answering their questions and helping the university develop policies and guidelines with respect to the use of AI and machine-learning systems in the classroom, in research and in administrative work.
“AI is not really a technology question — it’s more a people question,” explains Dr. Blouin. “Where is it appropriate or ethical to delegate automation or decision-making to algorithms and software systems — and where is it not? Especially within a university, a place where we disseminate knowledge, it’s important that we empower everyone to be part of that conversation.”
Dr. Blouin already hosted meetings and delivered presentations with many Faculties and faculty councils on the subject, with more to come. His initial focus is on putting together a guidance document for fall courses on how these AI tools (such as large-language models) should be considered.
“People want to know the boundaries of what they can and can’t do, and September is coming soon for faculty who may be looking to adjust their course plans or their syllabus,” he says. “The idea is a mix of pedagogical support and guidance-level advice — a working document that gets folks talking about it and feeling like they can start to get engaged in the subject.”
Review: Working draft: Guiding principles for large-language models - 2023-24 academic year [PDF]
Longer term, it’s about helping Dal prepare itself for an AI-informed digital future in which the pace of change is accelerating. Dr. Blouin wants to ensure the university isn’t caught off-guard by new developments but, instead, has the processes and people in place to carefully consider opportunities and challenges as they emerge. Most importantly, that we get better at coming together and make nuanced decisions in a multi-disciplinary and collaborative manner.
The human element
While the term “AI” is still perhaps best known for its sci-fi context in popular fiction like Terminator or The Matrix, its current application isn’t about artificial consciousness akin to actual human thinking. It’s about computer processes that consider massive amounts of data, whether words or numbers, to perform certain tasks very quickly.
What makes it seem “intelligent,” though, is that the tasks being performed have, traditionally, been distinctly in the human domain such as writing complex text in particular styles or creating realistic-looking images. Through AI tools, computer software can now perform these functions — and can do so at a much higher quality level than ever before.
Scary stuff? It can seem that way. “The first time someone uses a tool like Chat GPT it can be pretty overwhelming,” says Dr. Blouin, referring to the text-generating software developed by OpenAI that, since its launch just seven months ago, has become the standard-bearer for what modern AI can do. “These systems designed to generate language exercise quite a bit of analytical skills, and that’s disturbing, because we thought we [as humans] had a monopoly on that.”
But these sorts of big-data systems can also be incredibly helpful. They work so fast, and on such a huge scale, that they can accomplish easily automatable tasks or processes that take up significant amounts of time — particularly ones that don’t require or benefit from creativity and analysis. Dr. Blouin cites an example of being asked to summarize a 90-page proposal: a large language tool can review that and provide back bullet points in seconds, versus taking hours to read through it and take notes.
“It gives me the ability to scan so much more information more quickly,” he explains. “But we should never make critical decisions based on that work alone.”
Empowering people
If we’re looking to maintain the essential human role in our work with AI, we need to make sure the humans know what to do with it all. And there is a lot to consider here — not just issues of authorship, but bias, privacy, copyright and (given the carbon footprint of the servers that run these tools) environmental implications as well.
“Just asking faculty to figure this out on their own — it’s not realistic or fair,” says Dr. Blouin. “That’s why we want to figure out how we provide guidance to faculty who are designing courses and programs on how to bring this effectively and ethically into their work. And this applies to staff as well. How do we provide the Dal community with hype-free information and guidance on what’s appropriate to do, and to help them adapt to a rapidly changing situation and make the best of it?”
In a way, Dr. Blouin sees his appointment as working towards his own redundancy as institutional lead of AI — to help Dalhousie reach a point where Faculties, departments, instructors, students all feel like they can engage with the bigger AI discussion in their own work or study.
“The problems and opportunities that AI represents in various fields and disciplines are unique to those disciplines. A computer scientist wouldn’t necessarily understand them. So over the next decade, everyone has to ‘own’ AI, not just computer scientists. But we can only do this if there’s a baseline understanding, and people feel they have the authority to make good, informed decisions.”
“If you empower people with knowledge to form their own opinion, and to have the confidence to do so — that’s how we navigate the ethical nuance of AI.”