By now, it is likely that everyone working in a STEM field has heard of recent developments in artificial intelligence generally, or of ChatGPT in particular. For over a year, the media has been saturated with coverage ranging from fawning over the revolutionary capabilities purportedly now available to everyone, to hand-wringing over the potential obsolescence of all clerical and knowledge workers.

In the post-secondary environment, you may have heard about Generative AI from colleagues concerned about the implications for academic integrity or excited about the pedagogical opportunities presented (or both), or even from enthusiastic IT administrators struggling to balance excitement with their training and knowledge of technology hype cycles.

This OER does not aim to instill either excitement nor anxiety in educators. Rather, we endeavour to provide a balanced and sober view of Large Language Model (LLM) technology and the chatbots it powers. We offer:

• a basic historical background,
• an understanding of the operation, functionalities, and limitations of LLM-based tools,
• some useful analogies for thinking about them, and ultimately,
• an appreciation of the specific issues and opportunities introduced for higher education STEM teaching.

How Widespread Could This Really Get, Anyway?

It’s reasonable to ask whether the amount of ink spilled over the potential impacts of LLMs on education—positive or negative—is warranted. It is, after all, in the interests of the companies selling these tools to generate as much interest and hype as possible about the “AI revolution” they are shepherding. This serves to maximize both sales and buy-in from public investors and venture capitalists. We have seen the leaders of these companies publicly call for regulation of the technology and warn of its potential dangers, which a sympathetic observer might interpret as good corporate citizenship. A less charitable observer might point out that the loudest calls for a “pause” come from the early market leaders, and that they conspicuously refer to future developments, rather than to their products already on the market. Keeping the potential of the technology in the news cycle— while imposing greater regulatory burden on competitors playing catch-up—would certainly be a shrewd business strategy.

So beyond what’s in the popular press, what do we really know about adoption? How much use are students making of this technology, how is it impacting higher education today, and what might we expect in the near future?

The Data

According to the 2023 Global Student Survey, 40% of university students world-wide are using GenAI in their studies, but this number is higher in Canada (where 54% of students use it) and substantially lower in the United States, with only 20% of university students using GenAI. Worldwide, 50% of the students using GenAI in their studies input a question once or more per day. In Canada, 65% of students are using it from 2 to 10 times a day and 80% of students who use GenAI are using it more than once per day (only Turkish students are using GenAI more than Canadians, at 81%.)  Forty-four percent of students worldwide (but only 31% in Canada and 39% in the US) say that their understanding of complex concepts or subjects has improved since using GenAI for their studies. When asked if they would like their curriculum to include training in AI tools relevant to their future career, 83% of Kenyan and Indian students agreed. The world average was 65%; 63% of Canadian students (but only 47% of American students) would like training in AI tools (Chegg.org, 2023).

The Anecdotes

This Vox video, produced by journalist Joss Fong, provides some additional data and anecdotal colour on ChatGPT use, along with a concise and engaging introduction to some of the issues, from the point of view of both instructors and students:

These statistics are primarily focused on ChatGPT, which up until recently, has been the most widely available and easily accessible tool. However, Microsoft is now rolling out the same GPT4 technology powering ChatGPT across its entire suite: in the Edge browser and Bing search engine, built into all Microsoft Office 365 tools via Copilot, and even embedded within new versions of Windows itself. These products are widely used (often, the only institutionally supported option) across post-secondary institutions in Canada, and if past experience is any guide, it’s likely that many IT departments will simply “turn on” these features as soon as they are released, to staff, faculty, and students. It’s safe to say that by the time you are reading this, a full-featured ChatGPT equivalent will likely be just a click or sidebar expand away in Word, PowerPoint, Windows, and other tools that you —and your students —use every day.