Users of GPTZero's new authorship platform will be able to compile and share data about their writing process such as their copy/paste history, the number of editors they had, and how long editing took. — Reuters
GPTZero, a software startup aimed at identifying text written by artificial intelligence, announced that it's diving deeper into the world of AI detection with a new tool that tracks whether AI contributed to a particular document.
The goal is to move away from an all-or-nothing paradigm around AI writing towards a more nuanced one, says co-founder and chief executive Edward Tian in an exclusive interview with Inc. GPTZero aims to "bridge the gap" between "writers and readers," he added, whether that's between an employee and their boss or a teacher and their student.
As text-writing AI programs such as ChatGPT, Claude, Llama, and Gemini flood offices, schools, and the internet with machine-generated text, the task of differentiating AI words from authentically human ones has taken on a new urgency.
There are a number of different ways to go about that process. One recent study sought to hone in on AI's favourite words, which can include terms like "delves" and "crucial," while AI developers have also created (but not yet widely released) tools that can "watermark" AI text by subtly tweaking its diction.
GPTZero already offers an AI classification model, trained on a corpus of pre-existing human and AI text, that seeks to detect AI text in the wild. That's what helped the startup secure a US$10mil (RM42mil) Series A round over the summer. But the company's latest offering features a different type of AI detection software: tools that seek to identify the use of chatbots by monitoring the writing process itself.
Users of GPTZero's new authorship platform will be able to compile and share data about their writing process such as their copy/paste history, the number of editors they had, and how long editing took. The software will also help clarify how much AI someone is allowed to be using (if any) and record a real-time video of their "typing patterns and behaviour" that third parties can later review. The more traditional AI detection that GPTZero already offers is also included.
Tian, the co-founder, said the new tool was beta tested over the summer with the American Federation of Teachers and the University of Virginia School of Education. But aside from academic applications, AI authorship tracking has relevance in professional settings, too.
"AI detection has been shifting a lot from education to workplaces," Tian said, especially as computer-savvy students become computer-savvy workers. "We've received a lot of adoption with hiring managers, with recruiting (and) cover letter analysis."
Last month, the writing software juggernaut Grammarly announced plans to launch its own tool, called Grammarly Authorship, to track the provenance of written text. A Grammarly spokesperson told Inc on that the tool will begin rolling out in beta on Google Docs soon, with Microsoft Word and Pages integrations coming later. It will track which parts of a document were typed, pasted or AI-generated, including through a replay feature, and offers both academic and workplace applications. – Inc./Tribune News Service
