OpenAI releases its own AI-written text detector

After a 22-year-old from Toronto made a third-party ChatGPT detector instrument to find out whether or not a chunk of textual content was written by a human or an AI, OpenAI is itself now introducing a brand new “AI Classifier” for “indicating AI-written textual content.”

“We’re launching a classifier educated to differentiate between AI-written and human-written textual content,” reads the instrument’s web page on OpenAI’s web site. The corporate says that whereas it’s unattainable to “reliably” detect AI-written textual content, it believes that good clarifiers can inform mitigations for false claims relating to a chunk of textual content that was written by AI however claimed to be human-written.

The corporate makes it very clear that its classifier is not absolutely dependable. In its exams, the classifier appropriately recognized 26 % of AI-written textual content, whereas it incorrectly labeled human-written textual content as AI-written 9 % of the time. “Our classifier’s reliability sometimes improves because the size of the enter textual content will increase,” mentioned OpenAI. Texts with lower than 1,000 phrases are liable to be labelled incorrectly.

OpenAI additionally says that the instrument shouldn’t be used as a “main resolution making instrument,” and as a substitute, ought to complement different methods of figuring out whether or not the piece of textual content is human-written or not. Additional, OpenAI recommends utilizing the classifier just for English textual content. “It performs considerably worse in different languages and it’s unreliable on code.”

The corporate is making the classifier publicly out there to gather suggestions from customers, whereas it continues to enhance the instrument. Check out the classifier here.

Picture credit score: Shutterstock

Supply: OpenAI