The DeepSeek neural network has reached the top of app stores in a few weeks, including in Russia. Applications for brand registration in our country have been submitted to Rospatent in two versions – verbal and visual. DeepSeek will be added to some Chinese smartphone models for free. In short, the Chinese IT giant has already made a lot of noise.
"DeepSeek has become the first platform that openly demonstrates the reasoning chain of artificial intelligence when solving a user's query. Instead of the standard phrases "I think, I think, I analyze, I check," which even advanced ChatGPT models used until recently, DeepSeek reveals a step-by-step decision-making process. This approach is not only striking in its frankness, but also has educational value. This innovation turned out to be so significant that it actually launched a real "arms race" in the artificial intelligence market, - Dmitry Shevtsov, head of the e-commerce department at Postnoff & Co., explains the DeepSeek phenomenon.
Photo: ru.freepik.com
The indisputable advantage of the new Chinese neural network is that you can use it for free. Those who use artificial intelligence in their work have already tested DeepSeek in action.
Kamilla Nigmatullina, Head of the Department of Digital Media Communications at St. Petersburg State University, Professor, Doctor of Political Sciences, told MEDIA TALK about her impressions in an interview: "The phenomenon of DeepSeek consists of only two things – it perfectly replaces paid analogues, being free, and due to deep thinking technology often shows better results. I immediately compared the four most popular models for my research purposes and made sure that DeepSeek's answers satisfied me the most. Whether this will be a temporary wow effect and whether something even more convincing from China, the United States or Russia will appear in the near future, the year 2025 will show, because the speed in the race of large language models is amazing."
Kamilla Nigmatullina is also a leading analyst at the St. Petersburg State University Center for Artificial Intelligence. A new master's degree program "Artificial Intelligence in Journalism and Media Communications" has been launched at the Higher School of Journalism this academic year. This science is currently being mastered by 14 people, most of whom have journalism education or work experience in the media. In two years, students will learn how to use neural networks in their work, formulate technical specifications for IT specialists and verify execution, manage media and IT projects that use artificial intelligence, as well as explore neural network practices and evaluate their effectiveness. Among the teachers are employees of Yandex.
They also teach at Moscow State University, where there is also a course on neural networks. Universities of the two capitals are partners in the development of this area, and joint events are planned for 2025. Even earlier than in St. Petersburg, a similar course was opened at the UNN. Lobachevsky.
In the wake of the popularity of neural networks, private courses have also been opened, designed for an average of two months. Kamilla Nigmatullina is sure that it is impossible to comprehend all the subtleties in such a short time: "Two months is a very shallow immersion in such a complex topic. Personally, I studied for six months and continue to improve my skills. I believe that most of the courses offered cannot yet be compared with what the university offers. We train mainly managers who will be able to competently implement AI in media production, so their training focuses on the basics of management, project management, the psychology of teamwork, certain aspects of media production, and, of course, research on both neural networks in media and research prepared using automation of big data processing. The academic approach also involves an analysis of professional standards of work, ethical issues, and global changes in the media system."
The topics of the students' theses will be, in particular, SMM automation, advertising strategies using AI, trust issues and ethical challenges. The interest of higher education institutions in neural networks is natural and this area is developing very actively now. A leading analyst at the St. Petersburg State University Center for Artificial Intelligence is confident that neural networks will inevitably become the main tool of media production in the future. New developments in the artificial intelligence market will contribute to this.
The majority of current journalists are still looking at artificial intelligence with caution. Those who started their careers 10 or more years ago say that they use the neural network only as a technical tool: for example, to decrypt large interviews. They pay special attention to the facts that the neural network can produce, without fail checking the original source.
At the moment, the so-called "fact checking" is indeed the weak side of neural networks. "The peculiarity of AI's work today is that it is poorly able to do fact checking, and this important part of the work should remain with the employee who is responsible for publishing," Igor Krasnoshchek, a member of the Council of the Marketing Guild, said in an interview with MEDIA TALK. At the same time, the expert has no doubt that very soon every media employee will need the skills to use AI in their work: "The prospects are quite big. AI can already do quite a lot of things that can be used in the work of media and mass media. There will be competition, but rather indirect — it's like comparing the work of the media with the ranking or aggregation of the news feed by search engines. But it is unlikely to do without interaction: those editorial offices whose employees do not learn how to skillfully use AI or AI agents in their work are likely to lose out to their faster and more skilled colleagues in the long run."
It is still difficult to predict which countries will take the leading positions in this "arms race". Experts' opinions on the prospects of Russian neural network developers vary.
"The current lag of Russian developments like GigaChat (BEAC) and YandexGPT (Yandex) from global giants (ChatGPT, DeepSeek) is understandable: their architecture was originally created for a narrow market. However, in the niche of Russian-language queries, they demonstrate competitive advantages due to their in-depth understanding of linguistic nuances, laws and culture. Similarly, DeepSeek dominates in China not only due to its technical strength, but also due to fine-tuning to the local mentality. Russian-Chinese cooperation is particularly promising here — combining the resources of the two countries can create a new type of neural network that combines the scale of Asian data with the linguistic expertise of Russia," says Dmitry Shevtsov, head of the e-commerce department at Postnoff & Co.
Andrey Smirnov, head of Serverspace, ITG Corporation, spoke more cautiously: "At the moment, there is no full-fledged Russian alternative to high-quality language models. The main reason for this lies in the scale and amount of resources required to create them. To train ChatGPT and DeepSeek-level models, you need huge GPU clusters or specialized chips (TPUs) that work for weeks and months without interruption. There are no comparable data centers in Russia, such as Google or Microsoft, and renting resources from foreign providers is complicated by sanctions and a high price tag. The creation of a full-fledged language model depends primarily on the level of investment, the availability of suitable computing resources and the ability to assemble a large team of highly qualified specialists. Large cloud platforms or our own computing infrastructure, competent organization of the learning process and access to extensive datasets are all critically important for the success of the project. With sufficient political and financial support, as well as coordinated work by the scientific community, business and the state, language models can reach the level of ChatGPT or DeepSeek."
It is obvious that such a powerful and rapidly developing tool as a neural network will take its place in many areas, including in the media. The higher school here is ahead of the curve and trains a new generation of journalists in accordance with the challenges of the time.
Earlier, we reported that Alibaba added $87 billion in a month thanks to breakthroughs in AI.
