We learned how students started using ChatGPT to do classwork, but OpenAI has launched a text detector generated by artificial intelligence (AI). This is a tool that, although far from perfect, could become a valuable resource for teachers.
This move by the company led by Sam Altman comes amid growing concern about the conversational Chabot. Some schools and universities have already banned it and have become the epicenter of a debate about current education models.
The ChatGPT Detector From The Creators of ChatGPT
At least for now, one of the challenges looming over the educational system is finding a way to detect text generated by AI. It is not enough to prohibit it. The solution from the creators of ChatGPT joins others that have emerged recently to address this critical challenge.
OpenAI warns that it correctly identifies 26% of the text written by AI. Likewise, it incorrectly labels human-written text as AI-written by 9%. That is, it is a tool that is not infallible.
A positive aspect of all this is that the detector is open to the public and can be used for free.
To test it, go to:
- Paste the desired text and click ‘Submit’.
After analyzing it, the classifier will return a result.
OpenAI ChatGPT AI-Generated Text Detector
The results will be ranked based on how likely an AI will generate the text. This way, users will have several options: improbable, unlikely, unclear or possible.
Additionally, it is essential to remember that the detector needs a minimum of 1,000 words to do its job and still needs to be trained to detect text co-written by humans. The analysis considers text entirely created by humans or an AI.
AIs: new Google?
This artificial intelligence has almost as many uses as you can imagine. It gives conversations on various topics, like the Mixx.io podcast, but you can also ask for movie scripts or song lyrics. Even make you a version of an existing work, like “A Day in the Life by The Beatles”. Of course, he can also write a coherent and well-written essay in a few seconds, which may be a challenge for education in the near future and journalism itself.
It may not be fear that causes it, but a slight unease. The amount of text and content that humans can generate using this tool can flood the Internet with seemingly realistic content: “students presenting work to their teachers using this type of tools, or web pages that generate massive content with monetary incentives.”
Instead, there is a reason to fear it, as “the same systems that generate this type of content are the ones that are capable of detecting it instantly.” The challenge is that in the future, there will be an infinite number of models, each with its weights and parameters, and then it will be more difficult.
Beyond fraudulent uses, the list of utilities is almost endless. There are code reviews like those mentioned, but you can also ask me to collect all kinds of information, such as a tutorial or all sorts of comparisons. This makes many think that it can be a substitute for tutorials and even Google’s search, which until recently viewed TikTok as its main threat.
“It is the closest approximation to how it would work in a fully developed system, and it should scare Google,” they said in TechCrunch, where they did a test in which they asked about the different types of Pokémon and their characteristics.
The search engine par excellence could improve, and a look through social networks returns many similar comparisons.