Seeking Honesty: OpenAI’s Tool for GPT Output Verification

OpenAI’s automated AI-powered chatbot, ChatGPT, has taken the internet by storm, but not without creating a few issues on the way. With writers, coders, marketers, and seemingly everyone else in between using ChatGPT to generate content, companies worldwide are staring down a tsunami of AI-generated content.

MUO VIDEO OF THE DAY

SCROLL TO CONTINUE WITH CONTENT

With issues of plagiarism and stolen content constantly swirling around ChatGPT and its output, OpenAI has now released a new GPT-Classifier, a tool designed to detect and disclose whether the text you’re reading was the work of ChatGPT or one of its other GPT tools.

Disclaimer: This post includes affiliate links

If you click on a link and make a purchase, I may receive a commission at no extra cost to you.

OpenAI Launches GPT-Classifier to Help Check AI-Written Text

OpenAI’s announcement comes as scrutiny on AI-writing and AI-generated text steps up and companies intergrate ChatGPT output.

The new GPT-Classifier attempts to figure out if a given piece of text was human-written or the work of an AI-generator. While ChatGPT and other GPT models are trained extensively on all manner of text input, the GPT-Classifier tool is “fine-tuned on a dataset of pairs of human-written text and AI-written text on the same topic.”

So instead of hoovering up information and recreating its best approximation, the GPT-Classifier attempts to compare similarities between known human text and known AI text to find discrepencies that reveal the source writer.

GPT-Classifier’s Detection Is Currently Limited

While the idea of easily spotting AI-generated text will be music to the ears of editors and educators, OpenAI has cautioned that its “classifier is not fully reliable.”

Currently, GPT-Classifier “correctly identifies 26% of AI-written text” while labeling 9% of human text as AI-written. OpenAI also notes that the tool’s accuracy “typically improves as the length of the input text increases.”

A cursory test of the GPT-Classifier spotted a human-generated example and marked it “very unlikely to be AI-generated,” and also correctly indicated that a ChatGPT-generated snippet on USB issues was “possibly AI generated.”

The Demand for ChatGPT Detection Is High

Even with the GPT-Classifier’s limitations, the demand for reliable ChatGPT detection is likely to see many turn to this tool. OpenAI’s commitment to building and releasing a free GPT detection tool is important because as more students, writers, programmers, and others use AI-text generation tools, understanding and detecting this input will become vital.

For now, although GPT-Classifier is up and running and available for testing, it’s best to take its labeling with a pinch of salt.

Also read: