To Innovation - Breaking Politics, Economics, Crypto & IT News
|
||||||||||
TOP STORIES
At least in the distant past the Red planet had a liquid bodies of water
ChatGPT designed its own AI detector
 Unlike ChatGPT itself is far from real use 06.Feb.23 8:32 AM By Shawn Highstraw Photo Photobank |
|
|
The creators of chatbot ChatGPT themselves came up with a solution to recognize texts generated by artificial intelligence. The AI Text Classifier program tries to estimate whether a text was created by a human or a machine. Texts created by ChatGPT are usually indistinguishable from texts written by people. The chatbot causes problems in education, among other things. For example, the program can give answers to test questions. ChatGPT developer OpenAI warns that AI Text Classifier is not reliable. In English-language tests, the program extracted 26 percent of texts written by artificial intelligence. The program incorrectly marked texts written by humans as "probably written by AI"in 9 percent of the tests. The longer the text entered, the more likely it is that AI Text Classifier can recognize whether the text was written by a human. For texts with less than a thousand characters, the program is very unreliable, the creators warn. The developers of ChatGPT made the program available for free. Although the software is not yet perfect, according to the creators, they want to hear from users whether it can still offer a solution. In addition, they are still working on improvements. OpenAI recommends that the assessment of a text is not based entirely on the AI Text Classifier. However, according to the creators, it can complement other methods of determining the source of a text. | |
Related Information:
