The emergence of ChatGPT has unleashed a new wave of innovation and disruption in various domains. Students are using ChatGPT to cheat on their assignments, writers are publishing generative content without attribution, and researchers are producing papers based on prompts.
How to Detect ChatGPT Content with GPTZero
GPTZero is a tool created by Edward Tian, a Princeton senior, who wanted to stop students from using ChatGPT to cheat on their assignments. He developed a web-based app that can detect if a text was written by ChatGPT or not. Image from GPTZero
The app is very easy to use. You can either paste your text or upload a pdf, docx, or txt file. Then, you just need to click on the “GET RESULTS” button and you will see a summary, an average Perplexity score, and a Burstiness score. The app will also highlight the parts of the text that are likely to be generated by AI.
How Accurate is GPTZero?
OpenAI, the organization behind GPTZero, claims that the tool has an accuracy rate of over 98%. They also admit that there is still room for improvement and they are working hard to reduce the error rate to less than 1%. This shows their dedication to enhancing the accuracy of GPTZero and making it a trustworthy AI detection tool.
Reddit User’s Experience with GPTZero
However, a Reddit user has reported that GPTZero is not accurate at all. They said that GPTZero flagged their own paragraph as AI-generated, even though they wrote most of it themselves. This personal experience raises doubts about the accuracy of GPTZero and highlights the difficulties of achieving perfect accuracy in AI detection.
Futurism’s Testing of GPTZero
In a testing done by Futurism, it was found that GPTZero correctly identifies whether a text was written by ChatGPT or a human. However, it also falsely accused nearly 20% of students of academic dishonesty. This false positive rate raises concerns about the reliability of GPTZero in academic contexts and the potential implications of incorrect identifications.
Nerdschalk’s Testing of GPTZero
On the other hand, a testing done by Nerdschalk found that GPTZero consistently determines texts generated by ChatGPT with high accuracy. However, it faces challenges when checking texts written by humans. This suggests that GPTZero may have trouble distinguishing between AI-generated content and human-authored content, which could affect its overall accuracy.
Follow my posts on LinkedIN
