18 {{ upvoteCount | shortNum }}
18 {{ upvoteCount | shortNum }}

Detecting AI in Universities

by Kay King
16 replies
A professor in Canada has found a foolproof way to 'detect' papers from students that are AI written.


I had posted here before about 'reference books' listed in an AI written paper that were not real books. They were references CREATED by AI to provide 'proof' and 'further information' on an AI written article.

The Canadian professor found footnote references used in AI 'academic' papers turned in by students had the same 'tell'. The references listed as 'sources' were totally fake.

He said he now simply looks up a few of the footnote references and can quickly determine which students used AI rather than writing their own papers.

I think it's funny - a system designed to be so 'real' it tries to create an illusion of reality. And a human smart enough to figure it out.
#detecting #universities
Avatar of Unregistered
  • Profile picture of the author RMRC
    Lol very ironic.

    AI will only ever be a tool, it will never be able to surpass the human mind. How embarrassing for the students who got caught!
    {{ DiscussionBoard.errors[11753967].message }}
    • Profile picture of the author DWolfe
      Originally Posted by RMRC View Post

      AI will only ever be a tool, it will never be able to surpass the human mind.
      Yeah, however, Ai can complement the mind currently. Are you aware that Ai can pass both the Bar Exam and Medical Exams? The teacher who caught the students is smart knowing students are lazy.

      Just something to consider - "Study Finds ChatGPT Outperforms Physicians in High-Quality, Empathetic Answers to Patient Questions
      While AI won't replace your doctor, the JAMA Internal Medicine paper suggests physicians working together with technologies like ChatGPT may revolutionize medicine. "

      "The panel of healthcare professional evaluators preferred ChatGPT responses to physician responses 79% of the time. "

      "Additionally, ChatGPT responses were rated significantly higher in quality than physician responses: good or very good quality responses were 3.6 times higher for ChatGPT than physicians (physicians 22.1% versus ChatGPT 78.5%). The responses were also more empathic: empathetic or very empathetic responses were 9.8 times higher for ChatGPT than for physicians (physicians 4.6% versus ChatGPT 45.1%). "
      Source - https://today.ucsd.edu/story/study-f...ient-questions
      {{ DiscussionBoard.errors[11753989].message }}
  • Profile picture of the author max5ty
    This post is completely off topic...and you can delete it after you read it.

    I visit the forum sometimes without logging in...as it seems the majority of people do by the statistics at the bottom.

    I've noticed spammers will come by and do a spam post with a link, and then the post is deleted and they're sometimes banned.

    However, when viewing the forum offline, the post with the link will still show up for hours and hours and usually days. I've seen them stay up for a couple of weeks or longer.

    I think some people have learned to drop by and do a spam post with a link knowing it will be deleted but also knowing it will show up on the forum when someone views the forum without logging in as it's how most people view the forum.

    Would seem like if a post were deleted for spam, it would also be deleted for those visiting the forum without logging in...but it's not.

    Didn't know where else to mention this, but I think it's a secret some spammers have figured out.
    {{ DiscussionBoard.errors[11754003].message }}
    • Profile picture of the author WF- Enzo
      Administrator
      Please try appending /?cache-bust at the end of a Warrior URL.

      Example:

      https://www.warriorforum.com/artific...ce/?cache-bust

      This could be caused by a slow cache refresh. Thanks.

      Originally Posted by max5ty View Post

      This post is completely off topic...and you can delete it after you read it.

      I visit the forum sometimes without logging in...as it seems the majority of people do by the statistics at the bottom.

      I've noticed spammers will come by and do a spam post with a link, and then the post is deleted and they're sometimes banned.

      However, when viewing the forum offline, the post with the link will still show up for hours and hours and usually days. I've seen them stay up for a couple of weeks or longer.

      I think some people have learned to drop by and do a spam post with a link knowing it will be deleted but also knowing it will show up on the forum when someone views the forum without logging in as it's how most people view the forum.

      Would seem like if a post were deleted for spam, it would also be deleted for those visiting the forum without logging in...but it's not.

      Didn't know where else to mention this, but I think it's a secret some spammers have figured out.
      Signature
      {{ DiscussionBoard.errors[11754208].message }}
      • Profile picture of the author Princess Balestra
        Originally Posted by WF- Enzo View Post

        slow cache refresh

        On certain niche datin' sites, this pickup is a guaranteed WIN.
        Signature

        Lightin' fuses is for blowin' stuff togethah.

        {{ DiscussionBoard.errors[11754320].message }}
  • Profile picture of the author Kay King
    Interesting - mods have pointed out some synch problems and this may be part of it. How desperate must you be to post spam and then hope it 'lasts' for a while?
    Signature
    Saving one dog will not change the world - but the world changes forever for that one dog
    ***
    I firmly believe that animals make people more human.
    {{ DiscussionBoard.errors[11754007].message }}
    • Profile picture of the author max5ty
      Originally Posted by Kay King View Post

      Interesting - mods have pointed out some synch problems and this may be part of it. How desperate must you be to post spam and then hope it 'lasts' for a while?
      If you visit the forum right now without logging in and go to this section and thread, there is a post above my last one that was deleted 7 hours ago and is still showing.

      Just a heads up.
      {{ DiscussionBoard.errors[11754009].message }}
  • Profile picture of the author spartan14
    Well dont subestimate AI
    Soon it will be able to think like a real human
    The movie terminator its a reality of our next future
    Signature
    How I Make 150$ Daily With Free Traffic
    {{ DiscussionBoard.errors[11754271].message }}
  • Profile picture of the author socialentry
    Originally Posted by Kay King View Post

    He said he now simply looks up a few of the footnote references and can quickly determine which students used AI rather than writing their own papers.

    I think it's funny - a system designed to be so 'real' it tries to create an illusion of reality. And a human smart enough to figure it out.

    You mean he didn't bother checking the references prior?



    hahahahha what a retarded professor.
    {{ DiscussionBoard.errors[11755641].message }}
  • Profile picture of the author WF- Enzo
    Administrator
    Eventually someone will "create" a software that can detect AI in academic use in the same way there are platforms used to detect plagiarism
    Signature
    {{ DiscussionBoard.errors[11755670].message }}
  • Profile picture of the author Hadi Kieleweke
    Interesting. I think students should try hard to use the gift of intellect bestowed on them to answer their homework. The AI tools should be used to just give directions on how to tackle the questions asked
    {{ DiscussionBoard.errors[11755916].message }}
  • Profile picture of the author Timofeii Dvornik
    What an interesting catch! This underlines how even the most advanced AI, though remarkable, can't truly mimic human nuances, including academic integrity. It also serves as a cautionary tale for students who might rely on AI to cut corners. Just goes to show that there's no substitute for genuine learning and effort. Plus, it's a gentle reminder about how tech, while incredibly helpful, can sometimes trip on its own complexities!
    {{ DiscussionBoard.errors[11760655].message }}
  • Profile picture of the author Princess Balestra
    How may we most easily & practically discern effuse of vomit from reliably advantageous means of succor?

    I would wish for pernickety discernment forevah to maintain its place upon our most hoomanly pinnacles.

    This is how garbage gets collected an' recycled steada consoomed till'n it infects all bloodstreams.

    Thing is, all leaps through hoops demand courage an' smarts.

    & why so?

    Because when you dun the leap, you proved sumthin' -- to yusself, an' for the benefit of evrywan else.

    rn we have a potentially corruptin' short cut to success.

    You gotta hope the professorials checkin' alla this stuff can distinguish the garbage from the IRL sweat & brain contribyootion.

    The moment the gatekeepahs let anythin' through, you don't gaht no gates no more.

    Worth thinkin' about next time your Mom is sick, you want a noo roof, or emergencies beyond your range of experience mean you gotta get sum 'spearienced schmuck in to fix your NIGHTMARE.
    Signature

    Lightin' fuses is for blowin' stuff togethah.

    {{ DiscussionBoard.errors[11760667].message }}
  • Profile picture of the author bobbyswinson
    Yeah that's a good way to catch it tho gpt4 has less hallucinations. But if you're so lazy that you don't even double check the references then you probably deserve to fail.

    Tho i will say the ai boom is still very early so i think being indistinguishable from humans isn't too far out.

    Most detectors have very bad methods of detecting AI text which is just looking at how 'predictable' the text is as a whole. But the AI text is trained on human text so it implies a homogenized way of people speaking on the training set = AI written else human written.

    But if someone just actually writes like AI then they'll get flagged as false pos.

    And if you're clever about it you can just use LLMs to one-shot evade detectors pretty easily.
    {{ DiscussionBoard.errors[11761057].message }}
  • Profile picture of the author Princess Balestra
    Question is: what is universality of import, direction, or delivery?

    Less'n you gaht a yooniverslly agreed drop point of wisdom an' intel, likely you gaht nuthin' but self-intrested maraudahs or uninformed bomb-schwanglahz.

    Anythin' crooshl gowin' on demands the critical.

    Nitzy spitzy joosy flowsy.

    For evrywan.

    Prolly we always standin' on a plateau of knowledge an' wisdom them people gotta carve teeth outta wood mighta discovahed STUNNIN'.

    (An' I am talkin' VIKINGS here bcs they bit them Yookay folks to death with their miracyool teeth, not thuh axes an' cudgels!)

    What can I say?

    Likely the emergin' gap between no facts and no observably actionabyool facts means I can now take 10 d**** up my a** without even bendin' ovah!

    Physiological beauty an' promise scripts such eternally achievable horizons.

    That is my start point always for glimmahs of fyootyoore brilliance gonna set nowan free.

    PRINCESS! YOU'VE SPOKEN MANY TIMES ABOUT YOUR IMPOSSIBLY PERKY BOOBIES!
    SO WHY NOT MAGNIFY THEIR VALUE BY WHOOSHEEING THEM OUT INTO 100% SQUIDOS?


    Yeah, see, bcs I am naht a moron, K?
    Signature

    Lightin' fuses is for blowin' stuff togethah.

    {{ DiscussionBoard.errors[11761176].message }}
  • Profile picture of the author Greg Morris
    Well, if the students do a little bit of smart work, they can totally fake it. I have used ChatGPT for my university assignments and never got caught despite the fact that my teachers used AI detection tools.
    Remember that AI is only to give you information. You can rewrite all of the stuff in your words and nobody can even guess if you have used AI.
    The fact is that we can use AI but we can not be fully dependent on it.
    {{ DiscussionBoard.errors[11761285].message }}
Avatar of Unregistered

Trending Topics