Dear Dominik,

Yes, the fact that most students (and the general public) do not understand what language modeling is or how it works indeed leads to entertaining productions. I have received papers on Plato's Phaedrus that include quotations "from the text" that are wholly invented by chatGPT. "Plausible" is perhaps the best word for the text thus produced.

Best,

Tyler

On Fri, Mar 24, 2023 at 1:09 PM Dominik Wujastyk via INDOLOGY <indology@list.indology.info> wrote:

I expect many of us are beginning to see student essays that are partly or wholly generated by chatGPT and friends. Reading these is a strange experience. The texts are like a fever-dream of a real essay. Almost correct, often plausible, strangely vague, sometimes insanely wrong and sometimes quite fantastical. They are textual versions of the AI pictures "woman laughing alone with salad," with two rows of teeth and indeterminate numbers of fingers.

One of my favourite features is imagined bibliographies. We all enjoyed the hilarious fake indological bibliography in Lee Siegel's Love in a Dead Language. Now, chatGPT is producing its own almost-real bibliographical entities. E.g., from one of my students this term,

Bryant, E. F., Chakravarty, K. K., & Pal, J. N. (2001). The excavations at Adamgarh: A Protohistoric site in Central India. Oxford & IBH Publishing Co.
Are you being faced with fake bibliography entries like this?

Best,
Dominik

_______________________________________________
INDOLOGY mailing list
INDOLOGY@list.indology.info
https://list.indology.info/mailman/listinfo/indology