GPTZero, a tool created by a Princeton University student and designed to detect AI-generated text, has raised $3.5 million in funding to build out its team to better improve the existing technology and launch a new, automated tool to fact-check the web.
Venture capital firm Uncork Capital and startup accelerator Neo invested in GPTZero, as did Jack Altman, the brother of OpenAI CEO Sam Altman, and former New York Times CEO Mark Thompson.
“They want to see this built toward a journalism-use case, because combating disinformation—and it being a filter for the internet—is seen as important,” says Edward Tian, the 22-year-old founder of GPTZero.
GPTZero, which now employs 10 people, was launched in January 2023. Over 30,000 people tried out the tool within a week of its launch, crashing the servers. Since launch, 1.2 million people have registered with the tool.
The tool analyzes text for its “perplexity” and “burstiness”—two measures of how complex and varied writing is—to determine whether it is likely generated by an AI or not. That approach is not without its flaws: An April 2023 research paper from Stanford University academics found that AI detection tools like GPTZero often unfairly categorized text written by non-native English speakers as AI-produced because of a lack of variance in their writing.
Despite such evidence of bias, interest in detecting AI-generated content remains high—as the investment in GPTZero shows.
A new tool developed by the company, called Origin, launched last week on the Google Chrome store. It allows users to highlight text on the internet, and a browser extension automatically checks if the content is AI-generated. In the future, it will also verify the source of information. “We’re going to be fact-checking every sentence within a website,” says Tian. “If Origin says it’s AI-generated, we’re going to be tracking down the citation to see if it’s false or not.”
Having such a tool is vital for internet users in the AI Age, says Tian. “Detecting AI is just the first step of this long game of what people need to know in the delta between AI and human generations,” he says.
Tian believes that GPTZero is uniquely placed to succeed because of the design issues inherent in the current generation of GPT technology and the importance of identifying where content has been made using those AI tools. “There’s this new problem that the data on the internet, [which] these language models are trained on, is polluted by the outputs of the [same] language model,” says Tian. “So now this whole industry needs a filtering layer to train better language models.”
Many within his team, including cofounder and chief technology officer Alex Cui, share that belief—which is why Cui dropped out of his PhD program in AI research at the University of Toronto to join the team full time.
“This is just the start,” says Tian. “Up until now, I was a student, still in school, with a side project. The moment everyone comes together, it’s full steam ahead.” Tian says that the funding will allow GPTZero staff to devote their labor full time to the project, improving the tool to meet the demands created by the volume of generative AI being posted to the internet. “We are transitioning from this is something we do on the side to something we’re putting our hearts in and going all-in on,” he says.
Beyond new features development, the funding will be used to fund world-leading researchers to come and work with GPTZero, giving them the ability to learn further techniques. Some of the funding will also go toward the further development of Origin, which is currently being coded.
“Having this as a shield for everybody to consume information is important,” Tian says. “Your eyes are not enough to tell the difference between misinformation and truth. Now you need to know the origin of everything.”