Shantanu Kumar / Pexels
ChatGPTplagiarists mind , as OpenAI has originate a shaft that is capable of detectingGPT-4‘s writing output with reportedly 99.99 % accuracy . However , the company has spent more than a yearwaffling over whether or not to actually release it to the public .
The company is reportedly bring a “ measured approach ” due to “ the complexness involved and its potential impact on the broad ecosystem beyond OpenAI , ” perTechCrunch . “ The text watermarking method acting we ’re develop is technically promising , but has crucial risks we ’re weighing while we search option , including susceptibleness to circumvention by bad actors and the potential to disproportionately bear on group like non - English speakers , ” an OpenAI voice said .
Shantanu Kumar / Pexels
The text - watermarking system works by incorporating a specific pattern into the model ’s written yield that ’s detectable to the OpenAI tool , but invisible to the conclusion substance abuser . While this tool can reliably spot the writing generate by its own GPT-4 locomotive , it can not observe the outputs of other models like Gemini or Claude . What ’s more , the water line itself can be dispatch by launch the schoolbook outturn through Google Translate , switch it to another oral communication and then back .
This is n’t OpenAI ’s first attempt at building a text - spotting tool . Last twelvemonth , it quietly axed a similar text detectorit had in development due to the tool ’s paltry detection rate and propensity for false positives . Released in January 2023 , that sensor ask a user to manually input sampling text at least 1,000 character in distance before it could make a decision . It managed to correctly assort AI get - content with only 26 % accuracy and mark human - generated subject as AI - derived 9 % of the time . It also led oneTexas A&M prof to wrong run out an total classfor supposedly using ChatGPT on their final assignments .
OpenAI is also reportedly hesitating to release the tool for fear of a drug user backlash . Per the Wall Street Journal , 69 % of ChatGPT substance abuser believe that such a tool would be unreliable and likely leave in false accusations of cheating . Another 30 % reported they would willingly drop the chatbot in party favor of a dissimilar example should OpenAI actually roll out the feature . The party also revere developers would be able to reverse organize the water line and build tools to negate it .
Even as OpenAI debates the merits of let go its watermarking system , other AI startups are rushing to release text detector of their own , including GPTZero , ZeroGPT , Scribbr , and Writer AI Content Detector . However , given their oecumenical lack of accuracy , the human oculus remain our good method of spotting AI - generate content , which is not reassuring .
