Good article here on what #ChatGPT is and isn’t. It is exceedingly good at mimicry. It isn’t “intelligent” or doing cognition or meaningful semantic processing, and likely never will be without introduction of different models. Right now, it excels at guessing the right next word, so fails miserably at anything requiring reasoning.
https://open.substack.com/pub/theline/p/nestor-maslej-truly-intelligent-ai
Nestor Maslej: Truly intelligent AI isn't upon us quite yet

Should an AI system that scores above 1,500 on an SAT but still makes fundamental errors that defy common sense be considered truly intelligent?

The Line
@davetroy I could not get it to understand that the paper assignment required scholarly research. The only time it used any was when I specifically told it to quote our textbook authors, which it did pretty well.
@Cindy_Conaway yeah I mean that will sorta kinda work for texts it knows about. But that’s going to run into all kinds of practical and ethical issues too.
@davetroy Right. I tend to make my papers pretty complicated (and have a proposal, first draft, and final in my upper level courses) specifically so that even if they are cheating, they have to reverse engineer so much that they might accidentally learn something.
@Cindy_Conaway yeah, I thought it might be a fun assignment to tell students to use #ChatGPT to write a paper, and then correct it and add/check appropriate references. They might end up doing so much work in trying to create something passable that they’d actually have achieved some understanding.
@davetroy Because I teach fully online in a well-established program, it's kind of a big deal to change an assignment, but as I revise or develop new courses, I may incorporate that or at least put something in the assignment where I ask them to contrast scholarly and non-scholarly research.
@Cindy_Conaway yeah I think it would be fun for them because they would think it’ll be easy at first, and they would also learn about the limitations of these tools.