The press service of the anti-plagiarism system, which detects borrowings in term papers, theses, academic papers and other textual materials, announced the implementation of a new feature that allows you to determine whether the analyzed text was generated using ChatGPT or other major language models.
Image source: “Antiplagiarism”
Based on many years of research in the field of natural language processing (NLP), Anti-Plagiarism has developed an algorithm for identifying and marking text fragments generated by AI systems.
The new feature is already available. It can be used by users who are paid subscribers to Anti-Plagiarism. The ability to search for “artificial” fragments does not provide full protection against auto-generated content. The materials of the service itself emphasize that these are fragments “presumably” created using the GPT-2, GPT-3 and ChatGPT language models.
If Antiplagiarism detects fragments of machine origin, the system flags the relevant materials and the document is given a status of “suspicious”.
According to Antiplagiarism executive director Yuri Chekhovich, the neural network “always leaves a trace”. In addition, all typed texts are in his opinion “Still different from human writing”. This feature allows you to identify documents created by artificial intelligence.
Add Comment