Quick access to main page (top) Direct access to main contents Quick access to main page (bottom)

OpenAI Can Spot Its Own AI Text with Crazy Accuracy – The Debate Over Public Release Heats Up

wikitree Views  

OpenAI is debating on releasing their new development of technology that detects text written by ChatGPT to the public.

OpenAI image / Vitor Miranda-shutterstock.com
ChatGPT image / Marina Ny-shutterstock.com

The Wall Street Journal reported that OpenAI can detect text generated by ChatGPT with 99.9% accuracy, but debates on releasing this technology to the public have been ongoing.

The technology works by inserting an invisible watermark into the text, allowing it to decipher ChatGPT-generated text and providing a score indicating the likelihood that AI was used.

OpenAI reportedly was developing this technology before the launch of ChatGPT in November 2022, and internally, the technology is ready for employment.

Scott Aaronson, a theoretical computer scientist, blogger, and director of the University of Texas Austin’s Quantum Information Center, reportedly took a leave of absence to work on AI safety at OpenAI for two years to develop this technology.

OpenAI is grappling with the decision to release the technology, weighing user acquisition against transparency in its usage.

While some employees support the release by emphasizing its positive impacts, others raise concerns that the watermark can be easily removed. Employees pointed out that users could alter AI-generated text by translating it or adding and removing emojis, changing the original structure. Surveys conducted internally at OpenAI are complicating the decision to release the technology.

Last April, a survey revealed that 69% of ChatGPT users expressed that AI detection technology could lead to suspicion or blame directed at AI users. Schools are calling for the release of this technology to prevent plagiarism through the use of AI. The Democracy Technology Center reported that 59% of middle and high school teachers in the U.S. believe that students use AI to complete assignments. This percentage increased by 17% compared to a year ago.

According to The Wall Street Journal, OpenAI plans to build public support for AI transparency by this fall and prepare to respond to potential legislation.

wikitree
content@viewusglobal.com

Comments0

300

Comments0

[LATEST] Latest Stories

  • NCT Dream to return in November: report
  • Seoul shares open sharply lower on profit-taking after record rally
  • Asking Husband for Financial Transparency 남편에게 경제적인 투명성을 요구하기
  • Art Basel Paris opens amid Louvre heist, bringing global art scene back under Grand Palais dome
  • Two decades after revival, Seoul's Cheonggyecheon flows with 8 times more fish
  • COVID-19 vaccines may help some cancer patients fight tumors

Share it on...