OpenAI releases a tool to identify text written by artificial intelligence

OpenAI, which has made a name for itself with ChatGPT recently , has now released a tool to identify texts written by artificial intelligence . In fact, the company has developed a product to help reveal what’s been done with ChatGPT and similar tools. The most important reason for the development of this is undoubtedly that almost everyone starts to produce content with ChatGPT and these contents seem to be created by human hands; hence to prevent plagiarism. 

The company said it’s impossible to reliably detect all AI-written text, but it believes good classifiers can reduce false claims that AI-generated text was written by a human. Therefore, the classifier, i.e. the detection tool, by OpenAI is actually not completely reliable. The company itself specifically states this. 

The classifier tries to figure out whether a particular piece of text was written by humans or an AI generator. While ChatGPT and other GPT models are extensively trained on all types of text input, the GPT-Classifier tool is tuned on a dataset of human-written text and AI-written text pairs on the same topic.

According to the information shared by OpenAI, in its assessments of an English text difficulty set, the classifier correctly identified 26 percent of the AI-written text as “probably AI-written”. The tool tagged only 9 percent of human-written text as artificially-written. In other words, although the result is not perfect, at least 15 percent difference proves that the product is open to improvement. 

Recently, many people have been wondering if artificial intelligence will take our jobs in the future. To what extent this turns into reality, it is difficult to comment yet. However, it is admirable that a company like OpenAI, which is already doing artificial intelligence-focused work, also produces the antidote. 

Let’s remind you that you can try OpenAI’s new tool here. Because the company will continue to improve the product with feedback from users.