FARK.com: (12737916) ChatGPT's company OpenAI has a tool to rate a document's likelihood of AI origin, from a scale of Meatbag to If I Did It. Fark: only gets it 1 in 4 correct

		ChatGPT's company OpenAI has a tool to rate a document's likelihood of AI origin, from a scale of Meatbag to If I Did It. Fark: only gets it 1 in 4 correct (buzzfeednews.com)		8
		More: Fail, High school, Blog, University of Pennsylvania, Tool, Education, Writing, BuzzFeed, Human

• • •

282 clicks; posted to STEM » on 01 Feb 2023 at 11:53 AM (7 weeks ago) | Favorite | share:

8 Comments (+0 »)

View Voting Results: Smartest and Funniest

cretinbob

2023-02-01 11:28:30 AM

all they have to do is put an invisible header in.
Fun Fact. When you use your printer, there is a code that you can't see that identifies where the printout came from.
It all comes down to language skills but DON'T TELL ME HOW TO WORDS! amirite?

Mr_Vimes

2023-02-01 11:59:22 AM

Like any real living human writes for BuzzFeed.

akallen404

2023-02-01 12:05:39 PM

So media buzz notwithstanding, it's about as reliable and as useful as ChatGPT is.

devine

2023-02-01 12:28:49 PM

Just think of how powerful this ChatGPT tool will be for intelligence agencies. I'm already hearing about it being used for some business proposals and other internal reports at the company I work at as well as others.

Glorious Golden Ass

2023-02-01 12:36:47 PM

cretinbob: all they have to do is put an invisible header in.
Fun Fact. When you use your printer, there is a code that you can't see that identifies where the printout came from.
It all comes down to language skills but DON'T TELL ME HOW TO WORDS! amirite?

I've heard that. It's the only thing that keeps me from sending death threats to politicians I don't like and celebrities that won't date me. But if it's true, wouldn't a lot me people who didn't know that be on jail?

padraig

2023-02-01 1:33:54 PM

I worry more about false negatives than false positives. Some student could get royally farked if one of his reports got wrongly flagged as AI-written.

Fark_Guy_Rob

2023-02-01 1:54:33 PM

padraig: I worry more about false negatives than false positives. Some student could get royally farked if one of his reports got wrongly flagged as AI-written.

They have had plagiarism detectors for years, if not decades. They get it wrong all the time too.

paulleah

2023-02-01 9:51:53 PM

They say it only works reliably 1 in 4 times. If it's reliable 1 - 4 times, swap the yes / no output and you are right 3 - 4 times.

Consistency is consistency.

Displayed 8 of 8 comments

View Voting Results: Smartest and Funniest

Redisplay/refresh comments

This thread is closed to new comments.