That’s the thing. If a kid approaches learning with a language model responsibly, they stand to learn a lot very quickly and solve difficult problems that would otherwise be next to impossible for them.
The thing is, we need to teach them that today rather than tell them it’s cheating and try to catch them using it on essays and deal some kind of consequence.
I now use it professionally fairly regularly and it’s an easily justified expense. I’ve already delivered things to clients faster because of it. Most recently I reasoned through prototyping a sort of minimal CMS experience using a self hosted CMS API connected to Next.JS, and had a viable plan and prototype at the proposal stage in as much time as I’d normally just do the research on something like this.
If it’s feasible to accelerate learning and research for real world work, I think we should seriously consider how it integrates with education rather than encourage kids to avoid it entirely. Of course, we don’t have that awareness in our education workforce in Canada, but I wonder if it’s harmful to discourage the use entirely rather than accepting it and ensuring kids are still producing the work that’s expected. If it’s clearly GPT regurgitation with hallucinations and no bibliography, the kid has still failed to deliver. If they manage to do their work faster with technology (the main difference here is that they haven’t googled bunch of stuff, frankly) then great, they’re still learning something.
And of course, the more you tell kids not to use it, the more they’ll want to (which I’ve come to love, honestly).