[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

more ChatGPT experiments, on its sources of data



Hi all,

It occurred to me that a way I could get around ChatGPT’s not saying where its paraphrases come from (so you can check its sources) might be to ask for an essay and then say “  and follow it with a list of all the websites from which you took information for the essay.”


So, I went back to my Austen essay topic and it did tell me in a “sources” list where it was getting its information and its quotations: gradesaver, shmoop, cliffnotes, and sparknotes! That explains why all the English lit essays I’ve generated have been C-range. There are in fact plenty of more academic websites not behind paywalls that it could have used (the journal Persuasions, for example), but it went only for dreck by exploited student writers. I wondered why.


I next asked it to find three articles on Northanger Abbey from that online journal, knowing that a google search will in fact bring them up and link you to them. This is what ChatGPT said: 

I'm sorry, but as a language model AI, I do not have access to the internet and I am not able to find articles from specific journals. However, I can tell you that Persuasions is the journal of the Jane Austen Society of North America (JASNA), which is a scholarly publication that features articles and essays about Jane Austen and her works. It is likely that the journal has published several articles about Northanger Abbey, as it is one of Austen's most popular novels. If you want to find more information about this, you can visit the JASNA website and look for the Persuasions Journal.”

It doesn’t have access to the internet? Then how did it find shmoop and gradesaver? Really, anyone who understands that,I’d welcome an explanation.


I next tried an essay about the uses of solenoid circuits with sources included, and the AI gathered its info from some websites I’m guessing to be ok but not great: engineersedge, allaboutcircuits, machinedesign. They all have ads on their pages but they don’t look as shady as shmoop.

And I tried an essay on the Canadian constitution with a list of the websites it used, and got the Canadian encyclopedia, the govt of Canada, a couple of parliamentary papers, and the CBC. Pretty good.

This might explain why ChatGPT essays can get better marks in other disciplines than mine?


Nicky



Nicky Didicher, English, 

Simon Fraser University,

778-782-4337, didicher@sfu.ca

pronouns she/elle A button with "Hear my name" text for name playback in email signature 

Reduce, Organize, Communicate, be Kind!


At Simon Fraser University, we live and work on unceded traditional territories of the Coast Salish peoples of the Səl̓ilw̓ətaʔɬ (Tsleil-Waututh), Sḵwx̱wú7mesh (Squamish), and xʷməθkʷəy̓əm (Musqueam) Nations.

I am from British and German settler peoples. I was born in North Bay, Ontario, on the territory of the Nipissing First Nation Anishinaabe, covered by the Robinson Huron and Upper Canada treaties. I want to acknowledge and make visible the harm that settler peoples have done and continue to do to the First Nations, and I hope that there can be a reconciliation that actually recognizes their land rights. For a start, we SO need to change the name of this university!