a tool already detects AI-generated texts

ChatGPT has removed the technology dashboard to immerse you in a host of experiences coming soon. The curious thing about this news is that the same creators of ChatGPT artificial intelligencecapable of generating exceptional texts, launched a tool to detect them.

[Prueba esta flipante Inteligencia Artificial capaz de clonar tu voz y decir lo que quieras]

Distinguish between text written by a human and an AI

That ChatGPT is so striking in today’s world is due to the amazement it generates among thousands of people always amazed by their ability to answer questions, generate text or simply program among many other things.

Even due to the demand generated by the use of its tool, OpenAI, its creators are already working on a premium service in order to sustain it. And it is that this artificial intelligence is capable of almost everything without knowing its limits yet.

We are developing a new tool to help distinguish between AI-written and human-written text. We are posting an initial version to gather feedback and hope to share improved methods in the future. https://t.co/4dQE3dX6vX

—OpenAI (@OpenAI) January 31, 2023

OpenAI itself has just communicated from its website that it has launched a tool capable of distinguish generated text by a human and one created by artificial intelligence.

They did this using a AI Text Classifier web-based that predicts if the text was created by artificial intelligence. This way, all users who use ChatGPT for a multitude of solutions without warning of their use are stopped.

The fact that Microsoft itself continues to push very hard for this solution by increasing its invested capital and launching a high-caliber message to its competitors clearly shows that a tool like the one launched is vital for identifying texts.

How this tool works

OpenAI dropped some details of how did you train this classifier so that it is able to differentiate between human-generated and AI-generated text. He also clarified that it is not possible for him at the moment to detect all the writings created by artificial intelligence, but it is an important first step.

This makes it possible to identify many of them and thus prevent them from becoming spreading false information, academic dishonesty or AI chatbots that can impersonate humans; even a very important site has already used it for its publications.

A series of premises must be taken into account to understand the complexity and errors that this new tool launched by OpenAI can make:

• The classifier struggles to identify short texts (less than 1000 characters).
• Human-generated text can be mislabeled as written by AI.
• Its use is recommended only for English text, since in others it may not work correctly.
• The classifier cannot distinguish highly predictable text. An example of this would be that you cannot determine if a list of the first thousand prime numbers was generated by a human or an AI.
• AI-generated text can be changed to avoid detection by the classifier and it is unclear whether its detection will be a temporal advantage.
• Classifiers based on neural networks can be erroneously calibrated and make their predictions incorrect for certain inputs that differ from the data used for their training.

We tried it: and it’s not clear with Lorca

As we can see, this is a first step for this classifier which, for the moment, will not always be able to identify whether a text has been generated by artificial intelligence. has its mistakesit will therefore always be necessary to take their qualifications with a good margin of error.

Mistakes that can be found with poetry The Cheating Married by Federico García Lorca since the classifier was unable to identify whether it was created by a person or by the AI. Open AI has already warned that with languages ​​other than English, it really has a hard time. And in fact, when we pass this poetry translated into English, it identifies that it was generated by a person.

Y realmente la llegada de esta herramienta se dirige principalmente al sector de la educación, donde ya miles de jóvenes están usando ChatGPT para completar exámenes o trabajos de estudio.

De hecho, recomiendan a los educadores y a todo profesional en el entorno de este sector que compartan su opinión debido a la importancia que está teniendo ChatGPT. La idea de OpenAI es desplegar actualizaciones al modelo de lenguaje creado por los mismos en una colaboración muy estrecha con las comunidades que más impacto están recibiendo; como la dicha del sector de la educación.

OpenAI ha compartido la herramienta para que cualquiera pueda probarla desde este enlace al igual que se puede usar ChatGPT, para todo aquel que no ha tenido su momento de experimentar su gran capacidad para la conversación natural y otros menesteres.

Te puede interesar

Sigue los temas que te interesan