OpenAI, the company behind ChatGPT, has reportedly developed a watermarking tool that can instantly detect chatbot-generated text, but the company is reportedly hesitant to make the tool available due to the risks it poses.
Generative AI models have now reached incredible levels. Right now ChatGPT You can ask any question you want to chatbots and get an answer. Whatever you want, it prepares it for you in seconds. Of course, this situation is also abused. For example, students can have artificial intelligence write their homework. Or we often see content written by artificial intelligence on the internet.
There are many tools available to detect text written by artificial intelligence, but the vast majority of them do not provide reliable answers. A report published today by The Wall Street Journal found that OpenAI It will allow you to detect things written in ChatGPT it turns out that something has developed.
OpenAI is hesitant to release tool to detect text created with ChatGPT
According to the WSJ report, the company has written in ChatGPT can describe any text tool has been developed. An OpenAI spokesperson confirmed to TechCrunch that the tool is being developed. It could be very useful for understanding situations like having AI do homework.
But OpenAI is a watermarking tool without introducing this technology defined as indecisiveThe company said the vehicle was promising but carries with it significant risks It says it has. These risks include the possibility that it could be used by malicious people or that it might not produce disproportionate results in non-English languages. So it’s not clear whether it will be available or not. Time will tell.
So how does this tool work?
Let’s explain the tool a little bit. According to the information received, it will be very different from the current detection methods and will be able to give better results. The tool will only work on texts created with ChatGPT. In other words, it will not work on other artificial intelligences. It will make some minor changes to the words selected by the model and eventually tell them “an invisible watermark” will be added. Thanks to this, texts created with ChatGPT, by different tools can be detected.
We don’t know if the company will release this tool. However, considering that artificial intelligence is becoming widespread and such uses are increasing, it would be very useful to come up with a method that really gives accurate results.
RELATED NEWS