Several media outlets reported this month that the Vancouver School Board (VSB) and other districts in the Lower Mainland are planning to introduce AI chatbots into classrooms as soon as this academic year. VSB associate superintendent Pedro Da Silva stated that the board anticipates AI will be “incredibly important” in supporting learning and pathways to work and post-secondary education.
A Teacher Educator's Perspective
Joel Heng Hartse, a teacher educator in British Columbia for 15 years, offers a direct message to the VSB and other districts considering partnerships with Microsoft and others to incorporate generative AI into high school classrooms: “You don’t have to do this.”
Hartse reports that as a post-secondary educator, the importance of generative AI is greatly exaggerated. The primary effect he has observed is helping students outsource the very activities that university education is meant to foster: reading, writing, and thinking.
Problems with AI Outputs
Hartse has encountered numerous issues with generative AI, including producing false quotes from literature, introducing fake sources into bibliographies, fabricating nonexistent people in reports, and editing student work until it resembles advertising copy. He acknowledges that he may sound like an anti-technology crank but emphasizes a critical point: what these chatbots produce is not information.
It is challenging to explain this to students and administrators accustomed to searching for answers on Google. Large language models (LLMs) generate plausible sequences of words—or tokens, parts of words converted into numbers—that mimic patterns in their training data. While they are adept at producing text resembling human communication intended to convey verifiable facts, the connection between LLM output and essential educational values like truth, facts, information, and knowledge is incidental.
Confident Errors
This is why generative AI consistently and confidently produces incorrect text. Hartse provides examples: asking for former members of his favorite band yields a guitarist who never existed; requesting relevant academic sources results in five imaginary articles with fake URLs; and generating a biography of a scholar he knows attributes expertise in topics they have never written about.
Some LLMs are better at avoiding these errors, particularly those that include footnotes from online sources. However, all LLMs produce text that does not correspond to reality, and only a person with deep subject-matter knowledge can discern the difference. Hartse cites Jon Christian, editor of the tech media company Futurism, who described an LLM as “a virtual dumbass who is constantly wrong.”
The Simple Alternative
The great thing about this technology, Hartse concludes, is that you can simply choose not to use it. No one is forcing adoption. If a job requires it, one can learn as they go, much like learning to create Pivot Tables in Excel at a first office job after college.
