The company was tutoring English Literature as one of its subjects.
They were generating English Literature exam problems - for their users - using the ChatGPT web UI.
They would upload the marking spec, and say: "Give me an excerpt from something that might be on this syllabus, and an appropriate question about it".
Naturally, their users - the high school students - were getting, often, hallucinated excerpts from hallucinated works by existing authors.
I think the kids will be fine - it'll be their world, at some point, and that world will look a lot different to now. Maybe that's too optimistic!
I would hope, in that world, LLM literacy amongst adults has increased.
Because I feel really, really bad for all the kids who are beating themselves up about getting badly marked by ChatGPT (I assume) on an imaginary excerpt of an imaginary Wordsworth poem by their functionally imaginary tutor.
It makes me laugh, and reminds me of one of my favourite jokes, about the inflatable boy who - being of a rebellious nature - takes a safety pin to the inflatable school. Chaos ensues. Afterwards, the inflatable boy's inflatable teacher says:
"You've let me down; you've let the school down, but worst of all, you've let yourself down."
I guess I'm suspicious of the linked article. Call me full of hot air, but is it actually a safety pin? Or is it just designed to look really good on an application for an inflatable college?