Here's a first... I just got an email because ChatGPT suggested an article I wrote to somebody. Could I send them a copy? Except, I never wrote the article, it doesn't exist. PLEASE realize right now that this tool isn't pulling out cool references for you. It's making plausible titles and matching them to authors names.
@L_howes yeah, we are testing #ChatGPT now as part of a project led by a brilliant 4th yr student, Devin, and she is finding that refs to articles are all made up, with titles worded to make them more "relevant" to the query, while author names are picked out of a hat to sound plausible. Can't pull DOIs either
On an interesting note, Devin found that refs to books are generally correct, at least from her sampling so far 🤔 we are planning to write up a bioRxiv about it
Before you write your paper and submit it to arXiv, you should be aware that this has been reported over and over again. I myself have written on that some six weeks ago (you are welcome to cite that).
Such "Schrödinger Facts" are exactly expected from the way the generative models work, there is nothing nefarious or malicious involved, and once one understands why this happens, one can still make good use of the results.
https://sentientsyllabus.substack.com/p/chatgpts-achilles-heel
This invites us to pay more attention to fact checking. That's not a bad thing.
And no, books are just as readily confabulated in my experience.
🙂
#SentientSyllabus #ChatGPT #HigherEd #Education #SchroedingerFacts
Nice to meet you Elisa. It's always a pleasure to encounter colleagues who share my interests.
🙂