This 'GPTZero' Is An Anti-ChatGPT, Designed To Fight AI Plagiarism And Deepfake Text

Robots typing

To counter a smart AI, another smart AI is needed.

That is at least what Edward Tian, a computer science student at Princeton University, had in his mind. When he came to understand what OpenAI's ChatGPT is capable of, he spent the holiday period to build what he calls the 'GPTZero'.

GPTZero is an AI made to counter ChatGPT, in which it's capable of detecting whether an essay was written by ChatGPT, or other generative AI engines.

Tian, who is a former open source investigator, created GPTZero to identify deepfake text, a subject that has attracted a growing amount of criticism amongst researchers and educators.

Realizing that ChatGPT can create queries in a human-like fashion never ever before encountered, they have sounded the alarm on plagiarism.

They are concerned that the viral popularity of ChatGPT will lead to waves of generic-sounding, mostly AI-written essays, that mean nothing more academic dishonesty.

Since OpenAI introduced GPT-3, and later, the ChatGPT, the academic and business world have been in a debate into how to respond to the potential misuse of this kind of technology.

And Tian's project here, is one of the attempts to counter generative AI engines.

He shared two proof-of-concept videos comparing the AI's analysis of a New Yorker article and a letter written by ChatGPT. In the video, Tian showed how the AI managed to correctly identified that they were respectively written by a human and AI.

According to Tian, his AI works by processing submitted text to find indicators of AI origins, which can include randomness and complexity in how things are written.

This is technically referred to as “perplexity and burstiness.”

Tian's motivation for creating GPTZero was academic in nature, over what he termed "AI plagiarism."

Tian tweeted that he thought it was unlikely "that high school teachers would want students using ChatGPT to write their history essays."

When Tian uploaded his app for people to download, GPTZero was so popular that it crashed the server of gptzero.me website was hosted "due to unexpectedly high web traffic."

Tian had to step up its website hosting capacity to better serve visitors who wish to download GPTZero.

"I was expecting, at most, a few dozen people trying out the app," Tian said in an interview, but "suddenly, it was crazy in usership with over 2000 people signing up for the beta in a few hours."

"So many teachers have reached out to me," said Tian. "From Switzerland, France, all over the world."

"I’m awestruck that it blew up and went so viral," he added.

When using ChatGPT, people could simply query the AI things like: "How Did the World War II Influenced Modern Policy," and that the AI will generate a human-like results within seconds.

For freshman-level classes, this is a game-changer.

ChatGPT has put some educators on high alert when they found some of their students cheating using ChatGPT in order to create essays.

But what makes it even worrying is not actually plagiarism. What makes it concerning the most is ChatGPT's ability to make things up, writing pure nonsense, but make it seem and sound that it's real.

As for OpenAI, the company certainly creates various AI and launches many machine-learning products to advance humanity. But regardless, the company's own researchers are worried too.

For example, one of the researchers said in a talk at a Texas university that they "want it to be much harder to take a GPT output and pass it off as if it came from a human."

According to reports, OpenAI is also working on a feature for "statistically watermarking" ChatGPT outputs so that machine readers can spot buried patterns in the AI's text selections.

"I want people to use ChatGPT," Tian said, because as a computer scientist, he cannot say that he's against using AI tools for writing.

"AI is here to stay," explained Tian. "AI-generated writing is going to just get better and better. I’m excited about this future, but we have to do it responsibly."

“And it's only going to be normalized, but it has to have safeguards.”

While some people began scrambling, Elon Musk, the tech billionaire, who was an early investor of OpenAI, finds this amusing.

"It’s a new world. Goodbye homework!" he said in a tweet.

Read: A Reddit User Starts 'Deepfake'

Published: 
06/01/2023