ChatGPT may have met its match.
GPTZero, an app made by Princeton student Edward Tian, has recently gone viral on Twitter for its alleged capability of detecting whether an essay is written by a human or by ChatGPT.
ChatGPT, as we recently featured, is a free-to-use conversational model designed “to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”
Developed by artificial intelligence research and deployment company OpenAI, the AI chatbot has sparked concerns of misuse, particularly with regard to students potentially using ChatGPT to write essays or other types of written work.
In the same Twitter thread, Tian provides a demo of how GPTZero works:
As seen on the demo, GPTZero uses “perplexity” and “burstiness” to find out whether text was written by AI or by a human. According to Neowin, perplexity is a term used in machine learning that describes how surprised a language model is when it encounters new data. Meanwhile, burstiness is how often a term is used in a given text.
Neowin says that human writing has more burstiness while language models (like ChatGPT) are able to predict text better with lower perplexity.
Tian says that his main motivation in developing the app was “increasing AI plagiarism.”
“[T]here’s so much [ChatGPT] hype going around. is this and that written by AI? we as humans deserve to know!,” says Tian.
Currently, GPTZero’s beta version is free to try and is continually being developed and improved by Tian.
