The AI classifier for indicating AI-written text being developed by OpenAI, the creator of the hugely popular ChatGPT, seems to be performing poorly – a development which may have an impact on journalism as it tries to lean heavily on artificial intelligence.

“…the AI classifier is no longer available due to its low rate of accuracy. We are working to incorporate feedback and are currently researching more effective provenance techniques for text, and have made a commitment to develop and deploy mechanisms that enable users to understand if audio or visual content is AI-generated.,” states an update on a January 31 OpenAI blog.

This AI update indicates that, among other things, journalism too may be impacted since it would not be possible to effectively differentiate between human written text and artificial intelligence generated text.

Just imagine that a copy lands up at the desk of a sub-editor, who would find it difficult to know whether a reporter has actually written the story, or it has been generated through AI.

A case in point is of an ‘interview’ of Formula One world champion Michael Schumacher, who suffered a serious head injury. The so called ‘interview’ was run by Die Aktuelle. It later transpired that the answers to the questions had been generated by Character.ai

The AI classifier project had an ambitious beginning when OpenAI had said that it has “trained a classifier to distinguish between text written by a human and text written by AIs from a variety of providers”.

However, things seem to have gone south now.

“Our classifier is not fully reliable. In our evaluations on a “challenge set” of English texts, our classifier correctly identifies 26% of AI-written text (true positives) as “likely AI-written,” while incorrectly labeling human-written text as AI-written 9% of the time (false positives). Our classifier’s reliability typically improves as the length of the input text increases. Compared to our previously released classifier, this new classifier is significantly more reliable on text from more recent AI systems.”

On the limitations, OpenAI said: “The classifier is very unreliable on short texts (below 1,000 characters). Even longer texts are sometimes incorrectly labeled by the classifier. Sometimes human-written text will be incorrectly but confidently labeled as AI-written by our classifier.”

It went on to recommend using the classifier only for English text. “It performs significantly worse in other languages and it is unreliable on code.”
It is quite clear that guard rails need to be established as artificial intelligence is rapidly scaled up. Journalism needs those checks and balances and it is needed right now.