ChatGPT owner launches ‘imperfect’ tool to detect AI-generated text

ChatGPT is a free app that generates text in response to a prompt, including articles, essays, jokes and even poetry. It has become spectacularly popular in the short time since its release in November, while simultaneously raising inevitable concerns about copyright and plagiarism.

The AI classifier, OpenAI’s language model trained on the dataset of pairs of human-written and AI-written text on the same topic, aims to distinguish the text which was written by the latter. It uses a variety of providers to address issues such as automated misinformation campaigns and academic dishonesty, the company said.

In its public beta mode, announced in a blog post today (Wednesday 1 February), OpenAI acknowledges the detection tool is currently “very unreliable” on texts under 1,000 characters and that AI-written text can also be edited to trick the classifier.

“We’ve trained a classifier to distinguish between text written by a human and text written by AIs from a variety of providers,” OpenAI wrote.

“While it is impossible to reliably detect all AI-written text, we believe good classifiers can inform mitigations for false claims that AI-generated text was written by a human: for example, running automated misinformation campaigns, using AI tools for academic dishonesty, and positioning an AI chatbot as a human.

“We recognise that identifying AI-written text has been an important point of discussion among educators and equally important is recognising the limits and impacts of AI-generated text classifiers in the classroom. We’re making this classifier publicly available to get feedback on whether imperfect tools like this one are useful”.

Since ChatGPT’s launch late last year – almost instantly garnering millions of users – some of the largest US school districts, including New York City, have banned the AI chatbot over concerns that students will use the text generator to cheat or plagiarise on their classwork.

Other companies have created third-party detection tools, such as GPTZeroX, to help educators detect AI-generated text.

OpenAI said it is engaging with educators to discuss ChatGPT’s capabilities and limitations and will continue to work on the detection of AI-generated text.

Last week, it was revealed that ChatGPT had successfully generated essays to an acceptable academic standard on topics ranging from constitutional law to taxation, enabling the chatbot, in theory, to pass law exams at Minnesota University Law School.