OpenAI's AI Detector Turns Out To Be A Dud, Can't Tell Robots From Humans

In a major setback for ChatGPT-parent, OpenAI, the much talked about classifier, proved to be a disappointment, as it struggled to distinguish between human and AI-generated content, leading to its closure.

What Happened: OpenAI’s AI detector, initially hailed as a potential solution to identify AI-generated text, failed to meet expectations. The company discontinued the tool after acknowledging its remarkably low accuracy rate.

In an updated blog post, OpenAI said that the classifier was never effective at differentiating between human and AI-written text and even tended to produce false positives, incorrectly labeling human-written content as AI-generated.

OpenAI plans to develop new methods to detect AI-generated audio or visual content. Though specifics are yet to be revealed, the company stated that it remains committed to finding effective solutions.

The Sam Altman-led company released the classifier to detect AI-generated text earlier this year. At the time, OpenAI admitted that it is “impossible to reliably detect all AI-written text.”

Why It’s Important: In November last year, OpenAI released chatGPT for free public testing, receiving praise from numerous leading experts. However, as its popularity grew, so did concerns over its potential misuse for creating automated misinformation campaigns and fostering academic dishonesty.

In January, the New York City Department of Education took action and prohibited teachers and students from accessing OpenAI’s chatGPT on their devices and internet networks.

This decision was prompted by concerns raised by schools and educators who feared the technology could foster dishonesty and plagiarism, ultimately rendering traditional writing assignments obsolete.

Image Courtesy: Shutterstock.com

Check out more of Benzinga’s Consumer Tech coverage by following this link.

Market News and Data brought to you by Benzinga APIs