GPTZero — A bot to counter ChatGPT’s AI Plagiarism?
In this article you will understand what is GPTZero and how it works.
Ever Since OpenAI released ChatGPT, there’s been a rapid increase in plagiarism, recently a new bot GPTZero was released. Now, what is this new bot, and how it work is covered in this article.
“Write an essay on ______” was the only sentence required for any high school student to complete his/her essay since OpenAI’s ChatGPT launched and many teachers were worried sick about it. But not anymore, because Edward Tian, a 22-year-old senior at Princeton University, has built an app to detect whether the text is written by ChatGPT, an AI bot that everyone’s been talking about lately.
Edward is a Computer Science student and he created this bot because of the motivation to stop plagiarism since the release of ChatGPT which he claims can quickly detect the text produced by ChatGPT.
He introduced his innovation on January 2nd through a tweet saying, “I spent New Year building GPTZero — an app that can quickly and efficiently detect whether an essay is ChatGPT or human written”. He also said that he’s been receiving many positive responses from the teacher community after testing it. Within a week users of GPTZero had already reached the count of 30k and the app crashed due to heavy traffic.
Working of GPTZero
In order to determine whether the text is AI-generated or not GPTZero mainly uses two indicators:
- Perplexity (perplexity is a measurement of how well a probability distribution predicts a sample): It returns the complexity of the text. It works on the principle that the more complex the text is, the more likely it is to be written by a human also there’s a unique pattern in how a bot produces a text because it is trained on such data. GPTZero itself is trained on text data, so if it finds the text matter similar to its training data it can immediately return that the text is AI-generated.
- Burstiness ( it is the intermittent increases and decreases in activity or frequency of an event): it measures the variation of sentences. An important feature of a bot is that it tends to generate text with a uniform length of sentences whereas when we write we don’t pay much attention to such things and we write a shorter sentence alongside a long sentence. This idea is used to differentiate AI-generated text from one which is not.
To give you the simplest explanation of how it works, ChatGPT and GPTZero are both trained on nearly similar kinds of text data and that is the reason GPTZero can easily detect plagiarism. (A teacher who checks your paper has more knowledge of the topic)
After testing the bot some users reported bugs in the app to which Tian acknowledged that he is still developing the bot and still needs work to improve its performance.
What makes GPTZero Special?
This app is an eye-opener for all of us which made us aware that we are different from AI and will be no matter how minutely we design it and try to mimic it as humans, at least no time soon.
More about ChatGPT and AI Plagarism
- A research team at OpenAI has been trying to develop a watermark or some trademark on the text generated by ChatGPT.
- The New York City education department wants to block access to ChatGPT on school networks and devices over concerns about its negative impacts on student learning.
Okay, so in this segment, I will generate some text from ChatGPT and then see the results of GPTZero on whether the text was AI-generated or not.
Astronomy is the study of the universe and everything in it. From the smallest subatomic particles to the largest galaxy clusters, astronomers use a wide range of tools and techniques to explore the mysteries of the cosmos.
One of the oldest sciences, astronomy dates back to ancient civilizations, where early humans observed the movements of the stars and planets in the sky. Over time, as technology and knowledge advanced, astronomers were able to delve deeper into the universe and make groundbreaking discoveries.
After generating this text I checked it on GPTZero and here are the results
The model returns the results in terms of perplexity. As of now the model is still new and many more improvements and findings will be made. For now, if you want to try it, you can try it here.
That’s it for this article. See you at the next one
Follow for more!
Hello reader, I write because I have a thing for it and since you are reading this means you are also a curious person just like me. Connect with me on LinkedIn👇