Craig Froehle
#ChatGPT is the real deal, folks. This answer, which it generated from scratch, is better than 99% of the answers I would've gotten from the undergrads I teach, and probably better than 50% of the grad students I teach. If you're an academic and still relying on essay or short-answer questions on your quizzes and assignments, you might want to reconsider it...or at least ban access to the Internet. Otherwise, you may be getting evidence of ChatGPT's mastery instead of the student's.
Dec 12, 2022 at 5:06amWeb
Show threadCraig FroehleDec 12, 2022
Wow...it even noticed the essence of the Trolley Problem from a vague ethical question. This is damned impressive. #ChatGPT
Show threadCraig FroehleDec 12, 2022
I asked #ChatGPT about the risk to humanity posed by AI and part of its answer made me realize we should be expressing the exact same fears about ultra-rich sociopaths. I mean, if you replace "AI" with "corporation," the sentence is still 100% true.
Show threadCraig FroehleDec 12, 2022
Finally found a fairly basic technical question where #ChatGPT's answer isn't very good. It overly relies on the words rather than the actual meanings of the words. That said, I suspect this would pretty easily pass the Turing Test as a plausibly human response.
Show threadCL158Dec 12, 2022
@cra1g 😂
Show threadTristanDec 12, 2022
@cra1g Do you really think it’s that impressive? The answer is incredibly vague and shows no insight or depth. It basically just re-phrases the question. From what I’ve seen, I’m not convinced there is any advantage to be gained from this programme to cheat … except for very basic definitions. Which they could just as easily find on Wikipedia…
Show threadCraig FroehleDec 12, 2022
@Tris123 Sure, the information is "out there," but students still need to find it, read it, parse it, then rewrite it into a form that answers the question so as to avoid the plagiarism finders. All that takes work. This tool generates a unique response (well, in the sense that it can't be found by search engines) specifically formatted to address the question being asked, all with no more than copying and pasting the original question into the chat bar. 😕
Show threadCraig FroehleDec 12, 2022
@Tris123 This could easily pass for a reasonable response to a short-answer question on a freshman or sophomore literature course. I'm just saying that there hasn't before been a tool before that generates novel answers on the spot. And honestly, I expect the quality (sophistication and scope) of its responses to increase fairly rapidly now that the basic model is in place. #ChatGPT
Show thread@pineywoozle ‘s #3WordNoteDec 12, 2022
@cra1g ElNo tweeted re deactivating accounts on #Twitter - My real fear is that he or some1 he’s aligned w/ cld use #ChatGPT to spoof us & drive division in 24. ChatGPT is hard for me to spot & in 16 I told every1 I could buttonhole that there was an obvious syntax issue w/ tons of #farL accounts, dividing Dems. In college I had a prof who’s test language was off. I asked if he wrote it in a diff language & translated it. He said I was the first to notice in 20 yrs. #UsPol #Musk
Show threadChris O’keefeDec 12, 2022
@Pineywoozle @cra1g the use of ChatGPT for disinformation is pretty scary. Imagine an army of troll bots on steroids with the aim of dividing us. It’s probably already starting.
Show threadCraig FroehleDec 12, 2022
@nockee @Pineywoozle Yep! #ChatGPT
Show threadCraig FroehleDec 12, 2022
@nockee @Pineywoozle This one made me laugh. #ChatGPT
Show threadChris O’keefeDec 12, 2022
@cra1g @Pineywoozle haha. The thought of conspiracy types consuming and creating crap with this is insane.
Show thread@pineywoozle ‘s #3WordNoteDec 12, 2022
@cra1g @nockee Thats not a theory that’s just the facts man😉 Vegans are some scary pale mofos.🤣🤣🤣👍
Show thread@pineywoozle ‘s #3WordNoteDec 12, 2022
@nockee @cra1g Yep thats the fear. We need to start now connecting IRL so that we have more ways of uncovering fakes.