I wrote this about #ChatGPT in early January. I was a bit worried - I was commenting on new technology and I wasn't convinced I understood it.
After 2 more months of folks experimenting and commenting: I think I nailed this one. For once! https://scientistseessquirrel.wordpress.com/2023/01/10/chatgpt-did-not-write-this-post/
@StephenBHeard on my brief foray into #ChatGPT I asked it to give citations to some papers it claimed to be referring to. It produced titles of two real articles attached to completely wrong authors, journals, and even doi's. Which, of course, is just what you’d predict from the algorithms.
@fatamorgana Yes! I'm surprised by how many people are being surprised that it makes up citations. Of course it does - that's exactly what it's designed to do!
@fatamorgana @StephenBHeard
I wrote on those "Schrödinger Facts" a few weeks ago. Why they happen, why they're actually useful. In case you're interested ...
https://sentientsyllabus.substack.com/p/chatgpts-achilles-heel
BTW my new #Academia mantra is: Have the algorithm think with you, not for you.
Imagine what would happen to thinking, once the algorithm starts getting its Schrödinger Facts right ...
🙂