Want to See what ChatGPT Has to Say About You? How ChatGPT Gets It Wrong

       Have you wondered what ChatGPT might say about you if you want to write up a bio or cv?  Can you same time by letting AI write up your bio? 

       One author of multiple books and films, Gini Graham Scott tried to do just that by asking ChatGPT to write a bio of her and after generating her bio five times, discovered that ChatGPT is definitely not ready for Prime Time.  Each time, it made multiple mistakes and FAILED the accuracy test.

       The only thing that ChatGPT got right was the opening statement in each bio which featured some variations. For example, in Version 1 it described Scott as "an accomplished American author, filmmaker, and sociologist who has published over 50 books in various genres, including self-help, business, and crime fiction. In Version 2 it described her as "a prolific author, filmmaker, and consultant who has written and published over 200 books, both fiction and non-fiction, covering a wide range of topics, including business, self-help, personal development, and social issues. She has also produced and directed over short films and videos."

        But after that the errors in the five bios piled up.  For example, here are some of many of the errors, as Scott described them in her ChatGPT-created bio:

  • "ChatGPT variously claimed I was born on March 16, 1944; April 13, 1944, June 22, 1945, July 17, 1948, and 1944 in San Francisco, California, but none of these dates are correct, and I was born in New York, New York.  (I'm leaving out the actual date, since this might be used by scammers to steal personal identity information).
  • ChatGPT got my family background wrong, such as in one bio claiming my father was a chemist and my mother was an artist, when my father was actually a tax lawyer and my mother was a teacher…   
  • Another incorrect statement was that I attended Lowell High School in San Francisco, where I grew up – no, no, it was Great Neck High School in Great Neck, New York…
  • Also wrong. I did not teach sociology at the University of New Orleans and the New College of California, though I did teach sociology for a few semesters at West Georgia College in Carrollton, Georgia and at San Francisco State.
  • ChatGPT got the names of my books and films wrong, too.  It falsely stated that I wrote several self-help titles, such as "The Very Next New Thing: Commentaries on the Latest Developments that Will Be Changing Your Life;" but this is a social commentary book, not a self-help book. Also, I did not write a crime fiction novel called "Murder, Treason, and Seduction," though that might be a nice idea for a novel…
  • Still other books wrongly attributed to me were "Finding the Right Job," and The Complete Idiot's Guide to Self-Publishing," though it did correctly credit me with "A Survival Guide to Working with Humans" and "A Survival Guide to Working with Bad Bosses." 
  • ChatGPT also got my film career all wrong.  For example, it claimed I produced a documentary, "The Suicide Club" which "explores the phenomenon of underground social clubs in San Francisco" in one bio or "explores the world of suicide support groups" in another bio. In still another bio, ChatGPT lists this documentary as "The Suicide Prevention Handbook," a documentary on social issues.  But in fact, the one film I wrote and produced having anything to do with suicide is actually a dark comedy: "The Suicide Party #SaveDave," about a guy losing everything who decides to throw a suicide party. If he can raise enough money to save his house and get back on his feet, he'll live, and if not…well, that's when he'll commit suicide or maybe not."

       And this is just a partial listing of all the things that ChatGPT got wrong.As Scott concludes the article:

"Just about the only thing that ChatGPT got right was some general comments about my working in a number of fields besides being a writer and filmmaker, including being a consultant, trainer, coach, speaker, and conducting workshops on seminars on a wide range of topics.  And they got it right that I do seminars and workshops and speak on writing and publishing, personal and professional development, business communication, leadership skills, management, and social issues.  Then, too, ChatGPT correctly credited me with creating a publishing and consulting company, Changemakers Publishing and Writing, dealing with a wide range of topics including business, psychology, relationships, and social issues.

      But otherwise, almost everything else in this ChatGPT bio is completely WRONG!  Talk about misinformation and inaccuracy.  If I was a professor or teacher grading such a bio, I would give it a failing grade.  I'm not sure how ChatGPT put together all of this incorrect information, since I have correct bios on my websites and on the websites of multiple organizations, including Amazon and several writing groups and business associations.  There's even a bio about me on Wikipedia.  But for whatever reason, ChatGPT completely failed the test of generating a correct bio despite all the correct information about me, which is readily available online.

      So all these errors raise the question: given this disaster of creating an accurate bio – despite five attempts to generate one, how can we trust ChatGPT or other AI text-generating programs to accurately write anything?  This failed bio-creation task is certainly an example of why we still need humans to research and write accurate articles, books, and other materials.

      In the meantime, if you are curious, ask ChatGPT to create your own biography, and see what the program has to say about you."

        It's a pretty damning description of how ChatGPT can create perfectly reasonable sounding documents – and yet many if not most of the facts can be totally wrong.  So now that ChatGPT is being used by millions to create all types of written materials, this experience of writing a bio, even where biographical material on the author is available on hundreds or more websites, is a warning sign. 

        You can read the full article on Medium at https://ginigrahamscott.medium.com/what-if-someone-wants-to-know-about-you-using-chatgpt-how-chatgpt-gets-it-all-wrong-4670f1c21255 or at https://tinyurl.com/2p856vbc

Karen Andrews, Executive Assistant

Changemakers Productions

www.changemakersproductionsfilms.com

San Ramon, California

changemakersprod@att.net   

(925) 804-6333

Adblock test (Why?)



source http://www.expertclick.com/NewsRelease/Want-to-See-what-ChatGPT-Has-to-Say-About-You-How-ChatGPT-Gets-It-Wrong,2023291940.aspx

Comments

Popular posts from this blog

Consumer Safety Concerns For Hemp-derived Delta-9 THC

#141 Bawdy Jokes & Patter Songs

Digitalatto Ltd Launches IDO Phase 2, Ushering in a New Era of Innovative DApps and an AI Chatbot Platform for Online Businesses