The growing presence of artificial intelligence (AI) in daily life, both personal and social and professional, is an obvious phenomenon whose relevance is dealt with. 

The irruption of this technology experienced one of its most important and notorious chapters last year when the OpenAI company made available to the public a trial version of its ChatGPT product, a conversational language chatbot that responds to questions raised by the user, and that it has been shown to be capable of generating texts with a very notable level of writing and information that can make them difficult to distinguish from a writing done by a person. For this, we have several AI text detector tools online.

The texts generated by ChatGPT have a level of redaction that can make them difficult to distinguish from those written by a human being

Naturally, this can pose a problem, particularly in academic or even journalistic settings, where a professor or editor-in-chief may have difficulty identifying whether a text has been produced by an AI tool or by a student or journalist.

As might be expected, it is artificial intelligence itself that comes to the aid of those who face these dilemmas which are far from irrelevant, as explained by the American magazine Fast Company in a report in which it presents some of the tools that have been developed for the detection of texts written by AI.

A prediction algorithm

Shortly after OpenAI released its GPT-2, the GLTR (Giant Language Model Test Room) was unveiled. This is an algorithm developed by experts from the MIT-IBM Watson AI Lab and the Natural Language Processing Group at Harvard University that bases its AI text detection system on the idea that similar people recognize each other.

Microsoft plans to integrate ChatGPT into Bing search features

Another AI-generated AI content detector method that OpenAI is working on is to identify all text generated by your ChatGPT with some kind of hidden signal. Apparently, and according to Scott Aronson, a member of his research team, at a recent professional conference, the company already has a prototype of the tool.

"Basically," he said, "we want every time ChatGPT generates a long text that it contains a secret signal in its choice of words that allows us to prove that this text has indeed been created by GPT." The company would be the only one with access to that key.

In addition to those mentioned, other tools that claim to help detect texts created by software endowed with artificial intelligence are Zerogpt, AI Content Detector, from a company called Crossplag, and another called Unfluff.