The linguist talked about ways to determine the AI of a text using template phrases.

Texts that have been created using artificial intelligence (AI) can be recognized by their pattern, frequent semantic repetitions, and clear structure. This was announced on March 12 by Andrey Grigoriev, Acting Head of the I.G. Dobrodomov Department of General Linguistics at Moscow Pedagogical State University.

"The text created by artificial intelligence is devoid of individuality and emotionality. It contains only common, often formulaic phrases, well-known comparisons, semantic repetitions, and has a clear structure. In this sense, it is boring to read it, the eyes slip over the lines, the eye does not cling to speech inaccuracies," he said in an interview with RIA Novosti.

According to the expert, the generated texts may look too smooth and grammatically correct, but they often contain inaccurate or fictitious facts. AI can also refer to non-existent books and avoid specifying sources, which raises doubts about the reliability of the information.

In addition, people who regularly read high-quality texts can intuitively feel that the text was written by a neural network, since, despite its external correctness and the presence of some facts, it leaves a "feeling of emptiness."

On the same day, Inna Filippova, Chief Operating Officer of the SOK smart office network, said that AI is increasingly being introduced into office work, from document processing to data analysis. However, even the most advanced algorithms are not able to completely replace employees. At the same time, it is able to automate a significant part of routine processes — data processing and structuring, preparation of standard documents, information retrieval, primary analysis of reports and large amounts of data.

All important news is on the Izvestia channel in the MAX messenger.