How a 23-year-old founder transformed his student anti-cheating tool into a top startup detecting AI-generated text online

How a 23-year-old founder transformed his student anti-cheating tool into a top startup detecting AI-generated text online

GPTZero: From Senior Thesis to Multi-Million Dollar Startup

Edward Tian, a computer science major at Princeton University, was in search of an exciting project to complete his senior thesis in machine learning. Combining his interest in artificial intelligence (AI) and his minor in journalism, Tian set out to create a tool that could verify whether a block of text was written by a human or generated by an AI tool. Little did he know that this project would lay the foundation for his startup, GPTZero.

Over the course of his winter break, Tian developed a technology using machine learning that could accurately identify AI-generated text. He launched the tool, GPTZero, in January and it quickly gained popularity, attracting 30,000 users within the first week. Educators, in particular, found the tool immensely valuable in detecting student cheating on essays and writing assignments, as traditional plagiarism-detection scanners were not equipped to handle AI-generated content.

Tian’s inspiration for GPTZero came from his previous work at the BBC, where he researched ways to combat disinformation. As interest in GPTZero grew among customers and media, Tian realized he had the potential to turn it into a full-fledged startup. Despite having a journalism job lined up after college, he decided to team up with his friend Alex Cui, a PhD candidate in computer science, to focus on GPTZero full-time.

Securing a $3.5 million seed funding round led by Uncork Capital and Neo, as well as participation from industry leaders such as Lattice CEO Jack Altman, Stability AI CEO Emad Mostaque, ANBLE CEO Tom Glozer, and former CEO of the New York Times Mark Thompson, Tian and Cui expanded their team to 12 employees globally.

GPTZero’s success can be attributed to its ability to detect AI-generated text by analyzing two main qualities: “burstiness” and “perplexity.” Burstiness refers to the variability of word choice and sentence structure in a given block of text, while perplexity gauges how confusing the text is to the AI model scanning it. Although GPTZero is not perfect and can struggle with non-native English speakers’ writing, Tian emphasizes that it should not be the sole determinant of whether a student cheated. While other AI detectors have emerged over time, they often come with errors and inaccuracies.

• Mexican peso at 7-1/2 year high against dollar, more gains expected.
• Laurentian Bank’s shares drop following news that potential t...
• Vietnam’s VinFast, an EV manufacturer, to list in the U.S. in...

To address these concerns, GPTZero recently launched the “Origin” tool, a plugin that integrates into Microsoft Word to track a student’s writing process, providing verification that the work is their own. By focusing on “human” detection, GPTZero aims to overcome the challenges faced by other AI detectors. The company continuously improves its model by building upon new research, setting it apart from competitors who simply rely on existing AI models.

Looking to the future, Tian and Cui have expanded GPTZero’s scope beyond plagiarism detection. The startup now offers a broader AI-detection utility for various industries, including education, finance, journalism, and cybersecurity. Currently, GPTZero is free for texts up to 5,000 characters, while enterprise pricing ranges from $9.99 to $19.99 per month, with an additional $14.99 monthly fee for developer API access.

GPTZero’s journey from a senior thesis project to a multi-million dollar startup showcases the power of innovative thinking and the potential for AI in various fields. With their determination to provide the best technology and make a positive impact, Tian and Cui aim to revolutionize the way AI is detected and utilized across industries.

• ai
• ChatGPT
• startups