AI has taken the internet by storm, like a new app everyone’s suddenly obsessed with. In response to this surge, OpenAI has developed a tool that can detect the texts written by ChatGPT. In this article, we will talk in detail about this new AI detection tool from OpenAI and its effectiveness.
OpenAI’s new AI detector
Watermarking method for ChatGPT’s writing detection
As per the company, OpenAI has developed a text watermarking method that adds a hidden mark to their text. However, as reported by The Wall Street Journal, the company is still deciding whether to make this tool available as they are also considering the risks. These include the possibility of people finding ways to remove the mark and also the chance that it might affect certain groups, like people who don’t speak English.
OpenAI says it’s taking a ‘deliberate approach’ to releasing tools that can detect writing from ChatGPT https://t.co/kjAggCkZGn
— TechCrunch (@TechCrunch) August 4, 2024
This is not the first time OpenAI has come up with an AI detection tool. Last year they shut down their previous text detector – “AI classifier” due to its low rate of accuracy. However, it is said that this time the approach would be different.
The watermarking method would only target writing generated by ChatGPT, not by other companies’ models. This method involves subtly altering how ChatGPT chooses words, creating an invisible watermark in the text that a specialized tool can later identify.
The company says they have found that text watermarking is effective for catching changes made to the text locally, like rephrasing. However, it struggles with more global changes, such as using translation tools, changing the text with another AI model, or inserting and then removing special characters between words.
- Also, read
OpenAI AI Detector: Effectiveness
Someday using AI to write will be akin to using a calculator. While ChatGPT already has strong competitors like Copilot, Gemine, and Claudie, the idea of developing a tool to detect AI-generated text from OpenAI and exploring watermarking seems like digging your own grave and falling in.
On the one hand, this idea could be proven good in maintaining academic integrity and creating responsible users. On the other, this new method will only detect the content generated by ChatGPT, so using a tool like Quilt Bot and simply paraphrasing the idea is like a piece of cake to many.
- Meanwhile, watch our Lenovo Legion Pro 5i (2024) Review video
