“The neural network that creates the text always leaves its traces”
The team of the Russian service “Anti-plagiarism” announced the launch of a new functionality. In the text borrowing detection system, automatic detection of machine-generated texts has become available.
“Anti-plagiarism” for texts written by various AI systems, including GPT-2, GPT-3 and ChatGPT, is available to all users of paid versions. As noted in the press service, the machine-generated text detector has been trained and tested on a large data set, which minimizes the risk of false positives. AI generated texts are marked as “Suspicious” and generated fragments are highlighted.
“Antiplagiarism” has learned to detect texts written using ChatGPT
Yury Chekhovich, Executive Director of Antiplagiat, explained:
The neural network that creates the text always leaves its traces. When working on the definition of the generated text, the developers of the Antiplagiarism company also use artificial intelligence algorithms, naturally trained for detection tasks. Despite all the success of ChatGPT or Jasper, their texts are still different from those written by a person.