Skip to main content
Advertisement
Live broadcast

The expert warned students about the risks of using AI

Bogatyreva expert: it is simply useless to fight the use of AI in education
0
Select important
On
Off

Research shows that 82% of young people aged 18 to 24 are aware of neural networks with text functions, and many actively use them to complete educational assignments and write academic papers. On Friday, April 5, Lyudmila Bogatyreva, the head of the IT department of the Polylog agency, the author of the Telegram channel of the State IT Bogatyreva, told about the risks of using AI in the educational process in an interview withГазета.Ru ".

As the expert pointed out, universities are increasingly expressing concern that AI is becoming a standard learning tool. Many universities are already implementing AI text detection systems and setting rules for their use.

However, according to Bogatyreva, these measures often turn out to be ineffective — students find ways around them. A popular method is the use of special services ("humanizers") that paraphrase AI text so that detectors cannot distinguish it from human text. The specialist noted that it is pointless to ban AI.

"Instead, it's important to teach students to think critically, check facts, and be aware of the limitations of neural networks," she explained.

Bogatyreva argues that the key limitation of AI is the so—called "hallucinations", when the system generates false or fictitious information. This is due to insufficient training data, incorrect interpretation of queries and other factors, the expert said.

She said that even the most advanced ChatGPT-4.5 model today demonstrates "hallucinations" in 37% of cases, which is the best indicator among all versions. Despite this, according to a survey by the National Research University of Higher School of Economics, about 31% of students trust AI to find sources for their work.

The phenomenon of AI "hallucinations" poses a serious threat to the academic environment, Bogatyreva added.

"But it is even more dangerous if the error goes unnoticed and the work is published on the Internet — then other AI models can use it in training, spreading incorrect information further," the expert concluded.

Earlier, on April 1, Izvestia got acquainted with the results of a study by the APUM marketing agency, impulse.guru and social networking networks for IT and digital specialists "Grid" with the support of SpyWords, according to which more than 30% of Russians surveyed actively use neural networks to search for professional information, and about 50% of respondents periodically turn to neural search. Frequent users include marketers and lawyers.

Переведено сервисом «Яндекс Переводчик»

Live broadcast
Следующая новость
На нашем сайте используются cookie-файлы. Продолжая пользоваться данным сайтом, вы подтверждаете свое согласие на использование файлов cookie в соответствии с настоящим уведомлением и Пользовательским соглашением