The student defended his thesis written by a chatbot. To what extent is AI replacing human labor?

The brainchild of Elon Musk

The chatbot with artificial intelligence ChatGPT was developed by OpenAI (USA), one of the founders of which is Elon Musk. According to open sources, the bot was launched on November 30, 2022, and immediately became super popular with users due to its versatility and wide capabilities. Its developers set themselves the task of creating the most “human” AI that would correctly perceive human speech.

ChatGPT works in interactive mode, can use the context of the dialogue for answers, can write basic program code and look for errors in it, create various kinds of texts, scripts and poems, make translations, conduct financial analysis, compose songs and melodies and give accurate answers. It is impossible to enumerate all its possibilities – their circle is very wide. It is precisely in its ability to perform tasks from a wide variety of areas that ChatGPT compares favorably with other networks with AI – it was trained on huge amounts of data from various sources.

The bot is free to use, and paying subscribers for $20 a month get priority access to the chat even during its peak hours and to its new features. Among others, the bot also supports the Russian language, but it does not officially work in Russia and Belarus, fearing sanctions pressure from the US Treasury. Therefore, it is not so easy for Russian users to access the chat, but, according to experts, it is possible. For example, using a VPN and the phone number of the country where the IP address is located, or using free bots that repeat the skills of ChatGPT – you can easily find a whole list of them on the Web.

How ChatGPT received a diploma from the Russian State University for the Humanities

Alexander Zhdan, a graduate student of the Russian State University for the Humanities, took advantage of this. As a part-time student, he had already worked as a brand manager at the video game developer CM Games, and did not consider that a diploma was so necessary for him to devote a lot of time to working on it.

He rejected the option of buying a thesis for money, and decided to create an algorithm and take advantage of the possibilities of AI. In terms of time, all the work, in total, took him less than a day, but it required certain knowledge and skills from him.

So, the diploma plan, originally drawn up by AI, was rejected several times by the student’s supervisor, and then helped to draw up such a plan. The entire amount of work on writing a diploma – at least 60 pages of text on a given topic – the bot also refused to write. The work had to be divided into separate chapters, and AI easily coped with the task of writing them, backing it up with links to real scientific research. The last stage is the translation of the finished work from English into Russian. The AI ​​coped with it too, but the language of translation seemed “crooked” to the supervisor, so the student had to work on the text.

The uniqueness of the finished work was quite high – 82%, despite the fact that much less was required to defend the diploma – 70%. The defense of the diploma passed without arousing any suspicion among the teaching staff.

“Writing a thesis with ChatGPT is cool. You definitely need to edit it yourself, but most of the machine gives out everything by itself,” the author of the “unconventional”, in his own words, work shared his impressions.

The precedent gained fame by accident. The newly graduated graduate recklessly boasted on social networks, and one of the blog readers took it and complained about what was happening both to the university itself and to the Ministry of Education.

The university responded by immediately initiating an audit. But, given that before this incident, none of the students used this method of creating a thesis, the members of the teaching staff could not come to a consensus on what to do with the graduate and his diploma – to cancel the work? punish a student exclude without the right to retake? Or leave everything as it is? And, most importantly, how to check student work after this incident?

Diploma Alexander Zhdan decided not to cancel. In addition, the Ministry of Science and Higher Education of the Russian Federation stood up for the student. There should be no negative consequences for a graduate who wrote a diploma using the ChatGPT bot, said the head of the ministry, Valery Falkov, in an interview with the media. But after such a “strength test”, universities and universities will have to reorganize, and, at a minimum, think about changes in the approach to testing students’ knowledge.

“The neural network creates texts that pass anti-plagiarism. It is necessary to develop pedagogical and technical methods that will help teachers recognize such a forgery,” Alexander Tsyganov, professor, head of the department of insurance and social economics at the Financial University under the Government of the Russian Federation, agrees with the minister’s arguments. Teachers will be able to resist such schemes, the speaker believes. So, while working on a diploma, the student must show the supervisor that he understands the topic of the work. The second way is to abandon “descriptive” diplomas in favor of project diplomas, works in which students create some kind of practical solution.

This whole story happened due to the fact that in most universities students are simply not interested in studying, users of the World Wide Web comment. Graduate students are not methodologically helped, the course of the process is not clearly explained, but they cling to formal things. All this demotivates no worse than previous years of poor education. The high bar in some universities is rather an exception, they share their thoughts, hence the negative attitude towards Russian education, in general, and writing a diploma, in particular. After learning about the dissertation written by ChatGPT, former students began to share their practices of “hacking the system”. Most of them came down to copying works posted on the Internet. “And it worked!” they write.

But not everyone thinks so.

“What is the problem with writing a diploma in a whole year, and why such a desire to deceive the system? After all, the system is created, among other things, by academically dishonest students, they are part of the problem. RSUH is far from the worst university. What kind of problem is it to write a human diploma at the Russian State University for the Humanities? Why fight with your own education there? ”, Opponents leave such comments.

“Good work involves the work of a student for a sufficiently long period under scientific guidance, and an unscrupulous student is easy to spot,” says Sergey Pekarsky, Dean of the Faculty of Economic Sciences at the Higher School of Economics.

But now it will be more difficult for other students. Access to the ChatGPT chat bot both at the RSUH itself and at other Russian universities and schools was going to be limited.

The “pioneer” himself, as the media write, was offered a job at a well-known SMM agency. With the help of AI, it is supposed to create advertising campaigns in social networks, write short texts that inform about the draws and encourage users to go to the brand page.

“Rise of the Machines”: Have we over-praised neural networks?

There is an opinion that AI will replace humans in many areas of the economy in the foreseeable future. So, after reports of the case with ChatGPT, the media again began to actively discuss the question in which segments AI ousted a person from his workplace, or is close to it. For example, in the Russian Federation, since the beginning of this month, they were going to make the work of school teachers easier by entrusting neural networks with checking essays. AI will begin deciphering x-rays in public clinics in Moscow. Chatbots have already replaced a significant part of call centers and customer support services. They successfully work in the field of advertising design – such large advertisers as Nike and L’etoile use their capabilities in the development of illustrations and decorations for books and games, work as photobanks, melodic banks, in face recognition systems,

Does this mean that the use of AI will soon lead to mass unemployment, leaving many specialists with the need to retrain and choose other areas of activity? How close are we to the “rise of the machines”?

Despite the impressive capabilities of neural networks, they will not completely replace human labor, many experts believe. At least for the foreseeable future. For example, AI can create a simple melody for a commercial, but cannot write a symphony. He can write an ordinary news article in the media even now, but deep analytical authorial material is not available to him. AI can write code, but this does not mean that it can replace the IT specialist.

A significant part of the time professional developers spend on analyzing business requirements, designing architecture, organizing system monitoring, – comments Artem Bondar, head of the machine learning team at Yandex.Market. Writing code itself is only 20-30% of the value of a strong professional. And although the possibilities of modern neural networks are really impressive, it is too early to talk about the replacement of a person by a machine.

In addition, it turned out that the ChatGPT chat bot itself, which is ahead of Google and Yandex in terms of functionality, is far from ideal. Unlike the constantly updated systems of its direct competitors – Yandex and Google, ChatGPT uses data limited to 2020. Therefore, it may not give real-time answers to modern questions and newsworthy events. This bot has problems understanding some languages, and users have to wait longer for a response. In terms of the quality of data processing, experts write, it loses to standard search engines.

Also, ChatGPT is sensitive to the correctness of the question asked. The answer to it depends fundamentally on the wording. From time to time, experts admit, ChatGPT gives very convincing, plausible, but incorrect or even meaningless answers.

Back in December 2022, the administration of one of the largest forums for IT professionals, Stack Overflow, considered the average rate of correct answers of ChatGPT “too low” and banned the use of content generated by the language model.

But the Internet publisher Buzzfeed, on the contrary, intends to actively use the ChatGPT neural network to personalize content, increasing the number of publications. On this news, the company’s shares on the Nasdaq stock exchange rose by 119% in one trading session. The publication was inspired by a story told online by journalist Henry Williams. When he was assigned to write an article by The Guardian, he decided to get ChatGPT to work, and got the material ready in 30 seconds. Grammar, punctuation, fact-checking and syntax were flawless, the journalist admitted. But the text itself was “dry” and unemotional, and therefore required the author’s improvements. True, the editor did not complain about this – he received $ 615 for writing the article, or 42.5 thousand rubles at the current exchange rate. But after this incident, he cast aside all doubts that AI would soon deprive him of his job.