Chris Selley: We aren't slaves to AI, unless we want to be

National Post

8 hours ago

Article content
Immediate disbarment seems much more appropriate to me for any lawyer caught faking precedent — AI-generated or otherwise — but ChatGPT tells me that's unreasonable. Someday, though, presumably, this is going to lead to a truly disastrous outcome. That goes for every profession, not just the law.
Article content
I have a suspicion — or maybe it's more of a hope — that in 25 years we'll look back on the AI craze and chuckle at our own gullibility. It's not that it's useless; Google's AI-generated search results are a hell of a lot more useful than its actual search engine, if only because Google seems intent on making its actual search engine non-functional. AI can provide interesting information and answers — but then, and this is critical, you have to actually check that they're accurate.
Article content
That's assuming, of course, that we still know how to think for ourselves in 25 years. The stories you hear out of the education world about students relying on AI for their work — and struggling in vain to perform without it — are downright dystopian.
Article content
Article content
Tales abound of students using AI to generate entire essays, and I often detect a tone of hopelessness among teachers and professors lamenting it. At the high-school level, I've heard stories of parents going to the wall to protect their kids who've been caught plagiarizing — and of teachers basically giving up the fight, because unlike traditional plagiarism, it can be difficult to prove conclusively that something is AI-generated. ChatGPT doesn't always give the same answers to the same questions, which sounds to me more like a bug than a feature … unless you're looking to plagiarize an essay, of course.
Article content
(I'm amused to see that Grammarly, the website that allowed an entire generation to opt out of learning how to write an English sentence, now provides a service that will flag AI-generated text. Pick a lane, fellas!)
Article content
The frustrating thing is, it's not remotely difficult to detect AI-generated text qualitatively once you're familiar with it. It is absolute trash, always, and immediately recognizable to anyone paying the slightest bit of attention: unrelentingly bland, often repetitive, punctuated by weird verbs and jarring turns of phrase. Teachers in particular, who will (hopefully!) know how a student actually writes, will take notice when that student submits something completely out of character.
Article content
'Beyond a reasonable doubt' shouldn't be the standard for implementing academic sanctions. But if flagging AI plagiarism is too much too ask, there is a very easy solution: In-person, handwritten exams. Schools and individual teachers and professors are increasingly resorting to that — and finding that many of their students have no idea how to write, both structurally and in terms of penmanship.
Article content