AI Detector Confirms Article Written In 2014 Was 35% Created By Technology That Wouldn’t Exist For Another Eight Years
INTERNET — A leading AI-content detector announced Tuesday that it had successfully identified traces of artificial intelligence in an article written in 2014, providing researchers with the strongest evidence yet that ChatGPT may possess the ability to travel backward through time.
“This passage is clearly suspicious,” said the detector, examining several paragraphs typed by an actual human nearly a decade before ChatGPT existed. “The sentence structure displays characteristics commonly associated with GPT models, particularly the unsettling use of complete sentences.”
Experts immediately praised the breakthrough.
“For years, physicists believed time travel would require exotic matter, wormholes, or enormous amounts of energy,” said one researcher. “Apparently all you actually need is Grammarly and a paragraph with decent transitions.”
The discovery came after a writer deliberately produced an article without AI assistance, then submitted the exact same text to several AI detectors.
Results showed the article was:
100% AI.
78% AI.
42% AI.
Human.
And, according to two additional services, “$19.99 per month to find out.”
Industry leaders said the wildly contradictory findings demonstrate just how sophisticated AI detection technology has become.
“The important thing isn’t whether the result is correct,” explained one detector company spokesperson. “The important thing is that there is a large percentage displayed next to a circular progress bar.”
Researchers later expanded testing to articles written in 2014, 2019, and 2021.
Several were flagged for possible AI involvement.
“This strongly suggests one of two possibilities,” said analysts. “Either AI detectors are unreliable, or a freelance SEO writer accidentally invented generative artificial intelligence sometime during the Obama administration and simply forgot to mention it.”
Detector companies rejected criticism, stressing that their products should never be treated as definitive proof unless, of course, an employer, professor, editor, client, school administrator, or nervous marketing manager desperately wants them to be.
Meanwhile, genuinely AI-generated articles reportedly passed several detectors after a human changed four sentences and replaced “Furthermore” with “Also.”
At press time, an AI detector examining the Declaration of Independence had rated it 71% GPT-generated and recommended that Thomas Jefferson “rewrite the flagged passages in his own words.”