News for hackers. Stuff that matters.
Re:New AI classifier for indicating AI-written text by swatcoder on Tuesday January 31, 2023 @02:13PM (#34600240)
It's not good by any means, but neither is most assigned, casual, or rote writing.

10,000 students are writing some crappy essay on the Great Depression every day, and ChatGPT has probably trained on a zillion of these. It's optimized to produce those mediocre essays really efficiently, and that's very disruptive to how teachers have been working with students for the last century or so. The internet (and fraternity filing cabinets) were already straining this kind of pedagogy, but ChatGPT breaks it wholesale.

Re:New AI classifier for indicating AI-written text by gunshai on Tuesday January 31, 2023 @04:24PM (#34602381)
What I find interesting about your comment, is that while it can produce that mediocre essay. It can also produce a much better one.

How? Well it's all about how you interact with it. But the majority of use as you said will be taking the first output given the input. What's amazing to me is learning to reject the output in favor of our own vision or conflicting ideas.

If chatGPT helps people get past blank page syndrome and interact with their own ideas better to see the limits of what is returned contrast to what they think. That would be an incredibly useful tool for anyone trying to learn.