Post AVb7df1HEtU0SoHkDA by acm_redfox@jawns.club
(DIR) More posts by acm_redfox@jawns.club
(DIR) Post #AVYjl1XB8uOT6PGVSi by grammargirl@zirk.us
2023-05-11T22:30:43Z
0 likes, 0 repeats
This week, I've seen two different professors talk about catching students cheating with ChatGPT because the AI incorrectly cited information from the professor's own textbook. Wild!#ChatGPT #AI
(DIR) Post #AVYkRWsD0xt2SgN0ZU by EricaFriedman@mastodon.social
2023-05-11T22:38:24Z
0 likes, 0 repeats
I ran a test with Google Bard today, it fabricated a quote by me. It's not useful.
(DIR) Post #AVYpSMY1OEp8Uv90jI by genewitch@fediverse.projectftm.com
2023-05-11T23:23:23.104Z
0 likes, 0 repeats
@grammargirl@zirk.us it's poor at programming help too. I tried to explain it today as the first response from the chatbots is like the question asked on stackexchange, and the code never works correctly, if at all. Repeated requests to fix it will garner responses similar to the replies by people to the original question on stack exchange.It took about 1.5 hours to have a chatbot write pygame code that displayed something resembling raindrops falling, full screen. the unity code never worked, and never changed outputs, even after a similar amount of time.
(DIR) Post #AVYsrtzoT0l3L00tMm by aulia@octodon.social
2023-05-12T00:12:48Z
0 likes, 0 repeats
@grammargirl maybe I’ve forgotten what it was like being a college student but it’s insane that they don’t even bother fact checking the responses. Then again fact checking requires them to do the work they’re trying to avoid by using ChatGPT, so … 🤷🏻♂️
(DIR) Post #AVYw2PVcSO7ppdaoro by pa28@social.vivaldi.net
2023-05-12T00:48:19Z
0 likes, 0 repeats
@grammargirl the output of language models depends on the strength of the model for the text being generated. Every word produced is simply the highest scoring word for that place in the text. Wherever there is weakness associating, for example, an author with a citation the odds of getting it wrong go up. Increased model size hides but does not correct these mistakes.
(DIR) Post #AVYyeQVxMnxTrY0QUK by grammargirl@zirk.us
2023-05-12T01:17:30Z
0 likes, 0 repeats
@skotchygut I actually don't know, but it sounded like at the very least the students got in some kind of trouble.
(DIR) Post #AVZG2eeCAfeHrjzwR6 by edross@mas.to
2023-05-12T04:32:29Z
0 likes, 0 repeats
@grammargirlAnother similar situation https://mastodon.social/@kashhill/110351791881068388People need to realise that these things aren't replacements for encyclopedias
(DIR) Post #AVZPMICsDJqVe8ep4y by coffeenlanguage@mastodon.social
2023-05-12T06:16:51Z
0 likes, 0 repeats
@grammargirl I'm finishing my last two papers for college at the moment. All I can think is that, suddenly, my English-inspired bad habits in Spanish are an advantage because they'll make it clear I didn't cheat.
(DIR) Post #AVZoWBnLPBBvViX3YW by design_law@mastodon.social
2023-05-12T10:58:45Z
0 likes, 0 repeats
Okay, fellow profs, what would your ChatGPT "tell" be? I.e., what kinds of statements being attributed to you would immediately alert you to shenanigans?E.g., "As Burstein, Rajec & Sawicki tell us in the casebook, design patents protect general design concepts...."
(DIR) Post #AVZobEAA1opMIpbgR6 by design_law@mastodon.social
2023-05-12T10:59:40Z
0 likes, 0 repeats
Okay, fellow profs, what would your ChatGPT "tell" be? I.e., what kinds of statements being attributed to you would immediately alert you to LLM interventions?E.g., "As Burstein, Rajec & Sawicki tell us in the casebook, design patents protect general design concepts...."
(DIR) Post #AVZq5fEa2gceu8kdUG by toast@sfba.social
2023-05-12T11:16:23Z
0 likes, 0 repeats
@grammargirl what AI is essentially doing is poisoning the well. Minus any kind of check on veracity, bad information from AI-written term papers is going to find its way into the intellectual ecosystem. Also keeping me awake — if a despot like Putin decides to publish 10,000 websites stating Putin is a cool dude, eventually that information will make its way into an AI’s understanding of “facts”.
(DIR) Post #AVZvZqvNdrlHcTz7rc by gallaugher@mastodon.world
2023-05-12T12:17:51Z
0 likes, 0 repeats
@grammargirl @MakeAppPie now if this was an article citation of said work, it’d improvise the prof’s Google Scholar ranking, so the gaffe would be embraced and encouraged. All hail ChatGPT article authors.
(DIR) Post #AVa5UVjK0wohyoR6QK by grammargirl@zirk.us
2023-05-12T14:08:57Z
0 likes, 0 repeats
@coffeenlanguage True!
(DIR) Post #AVb7df1HEtU0SoHkDA by acm_redfox@jawns.club
2023-05-13T02:07:44Z
0 likes, 0 repeats
@grammargirl makes you wonder how many other crap citations are under the radar...