College student made app that exposes AI-written essays

ChatGPT’s artificial intelligence generated dialogue has gotten pretty sophisticated — to the point where it can write convincing sounding essays. So Edward Tian, a computer science student at Princeton, built an app called GPTZero that can “quickly and efficiently” label whether an essay was written by a person or ChatGPT.

In a series of recent tweets, Tian provided examples of GPTZero in progress; the app determined John McPhee’s New Yorker essay “Frame of Reference” to be written by a person, and a LinkedIn post to be created by a bot. He stated that he developed the app during the holidays and was inspired by AI plagiarism.

On Jan. 3, Tian tweeted that GPTZero wasn’t working, likely due to a larger than anticipated amount of web traffic. Tian today published a Substack Newsletter stating that more than 10,000 people have tried out the Steamlit public version of GPTZero. Tian stated that both gptzero.me (and the Streamlit version) are still showing errors due to traffic volume. In the newsletter, Tian said he updated the GPTZero model to “significantly reduce the rate of false positives and improve output results.”

GPTZero uses “perplexity” and “burstiness” to determine whether a passage was written by a bot. The app measures perplexity by how random the sentence’s text is and whether it is unusually or unexpectedly constructed. These sentences are compared to determine their similarity. Human writing has more burstiness — which is to say, we tend to write with more sentence variation.

OpenAI’s launch of ChatGPT, Nov. 30, 2022 has prompted many concerns regarding plagiarism. Within five days, it was downloaded by over a million users. AI-powered ChatGPT can create basic essays or mimic the writing style of famous writers. You can direct ChatGPT to copy Shakespeare’s voice, for example, or write in the style of a New Yorker essayist. Although there are some pitfalls in execution, the results can be recognized as being in the correct style. It’s not hard to get the AI to write a high school English-style essay, and to find the result pretty indistinguishable from an assignment written by a student. There are limitations, however. It’s easily baffled by riddles, and sometimes just makes up facts. Due to its frequent errors, StackOverflow has also removed any ChatGPT-generated coding feedback.

In December, OpenAI said it would “watermark” ChatGPT output, in order to combat plagiarism.

In his newsletter, Tian said he’s working on more updates to GPTZero, including “improving the model capabilities, and scaling the app out fully.”