ChatGPT’s AI-generated dialogue has become quite 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 single person. or ChatGPT.
I’ve spent the new year building GPTZero — an app that can quickly and efficiently detect whether an essay is ChatGPT or man-written
– Edward Tian (@edward_the6) January 3, 2023
In a recent series of tweets, Tian provided examples of GPTZero in progress; The app identified John McPhee’s “Frame of Reference” essay as written by one person and one LinkedIn post generated by bot. On Twitter, he said he created the app over the holidays and was motivated by the growing possibility of AI plagiarism.
On January 3, Tian tweeted that GPTZero was down, possibly due to larger-than-expected web traffic. in one Sub-stack message Tian announced today, he said that more than 10,000 people have tested the publicly available version of GPTZero on Steamlit. (At the time of writing, both gptzero.me and reasonable version is showing an error, possibly due to traffic.) In the newsletter, Tian said he has updated the GPTZero model to “significantly reduce the false message rate and improve the output”.
GPTZero uses “confusion” and “outburst” to determine if a passage was written by a bot. Confusion is how random the text in a sentence is and whether the sentence construction is unusual or surprising to the application. Burstiness compares these sentences with each other, determining their similarity. Human writing has more boom – that is, we tend to write with more sentence variation.
Plagiarism concerns have increased since the launch of OpenAI ChatGPT on November 30, 2022. More than a million people have used it within five days of launch. AI-powered software can write basic essays and mimic the styles of established writers. For example, you could direct ChatGPT to copy Shakespeare’s voice or write in the style of a New York essayist. There were obstacles to the execution, but the results were recognizable in style. It’s not hard to get AI to write a high school English essay and get good results indistinguishable from an assignment written by a student. That said, there are still limitations to what it can do. so easy Confused by the puzzleand sometimes just make up the truth. StackOverflow also forbids any encrypted responses generated by ChatGPT due to the frequency of errors.
In December, OpenAI said it would “watermark” ChatGPT output to prevent plagiarism.
in him newsTian said he’s working on making more updates to GPTZero, including “improving the model’s capabilities and extending the entire application.”
