Kim Kardashian was asked about her AI use in a new video from Vanity Fair published this week. And the reality TV star blamed OpenAI\u2019s ChatGPT for giving her the wrong answers while studying for tests.<\/p>\n
Kardashian has been pursuing a law career through non-traditional means since 2019 and took what\u2019s called the \u201cbaby bar\u201d in 2021. She earned her law degree in May and took the bar exam in July, though she\u2019s still awaiting the results, according to Entertainment Weekly<\/a>.<\/p>\n The actress talked about her use of ChatGPT during a Vanity Fair YouTube<\/a> video\u2014part of a series where celebrities answer questions while hooked up to a lie detector. Teyana Taylor, star of the recent film One Battle After Another, asked Kardashian questions like whether she considers AI a friend.<\/p>\n \u201cNo. I use it for legal advice,\u201d Kardashian said. \u201cSo when I am needing to know the answer to a question, I\u2019ll take a picture and snap it and like put it in there.\u201d<\/p>\n Taylor jokingly asked whether she was cheating and Kardashian clarified that it was just to study for her tests, but it often gave the wrong answer. \u201cThey\u2019re always wrong,\u201d Kardashian said, stone-faced. \u201cIt has made me fail tests all the time. And then I\u2019ll get mad and I\u2019ll like yell at it and be like, \u2018You made me fail, why did you do this?’\u201d<\/p>\n Kardashian then said that \u201cshe,\u201d referring to ChatGPT, will \u201ctalk back to me.\u201d Kardashian said that she tells the robot that it will make her fail, asking it how that makes \u201cher\u201d feel.<\/p>\n \u201cAnd then it\u2019ll say back to me, \u2018This is just teaching you to trust your own instincts. You knew the answer all along,’\u201d Kardashian explained.<\/p>\n Generative AI is notorious for giving bullshit answers because the technology fundamentally doesn\u2019t understand what it\u2019s saying. The responses are perhaps best compared to a magic trick that sounds like it\u2019s capable of reasoning and logic, but it\u2019s really just guessing at the most statistically likely word to put in front of the other. It\u2019s a neat magic trick, but a magic trick just the same. And it\u2019s the reason that AI chatbots struggle<\/a> with seemingly simple questions like how many R\u2019s are in the word strawberry.<\/p>\n But Kardashian\u2019s response about trying to guilt ChatGPT about giving the wrong answer speak to what\u2019s going on under the hood with all of these AI chatbot. That glimpse at a reassuring voice trying to encourage the user to trust their own instincts is yet another part of the magic trick. ChatGPT\u2019s o4 was notorious for being overly supportive of the user in an almost disturbing way. The upgrade to version 5 was troubling for many users who found o4 to be just what they needed in their lives.<\/p>\n