ChatGPT's Achilles Heel

I just posted a discussion of ChatGPT's major weakness for the Sentient Syllabus Project: https://sentientsyllabus.substack.com/p/chatgpts-achilles-heel

It falls prey to its "Schrödinger Facts".

In a nutshell: ask the AI for sources - and then check whether they actually exist. They usually don't. Requests for specific, provable sources turn out to be ChatGPT’s achilles heel.

That's a big deal: (a) we can (and should) always require students to attribute ideas anyway, and chasing down the source of a ChatGPT claim may be more work than working from an actual source; (b) ChatGPTs information is not useless though - and can be very valuable to point to the right direction; (c) while we become masters in attribution, the collaboration benefits in other ways, and the end result is improved.

Find our resources on academia and the new AI at https://sentientsyllabus.substack.com and on http://sentientsyllabus.org

... and share this information to whoever might find it useful.

#SentientSyllabus #ChatGPT #AI #Education #HigherEd #University #Academia #syllabus