{"id":57330,"date":"2025-09-28T22:14:46","date_gmt":"2025-09-28T19:14:46","guid":{"rendered":"https:\/\/geoconversation.org\/artificial-intelligence-at-nornickel-how-technology-is-changing-mining-and-people\/"},"modified":"2026-05-04T18:36:38","modified_gmt":"2026-05-04T15:36:38","slug":"artificial-intelligence-at-nornickel-how-technology-is-changing-mining-and-people","status":"publish","type":"post","link":"https:\/\/geoconversation.org\/en\/artificial-intelligence-at-nornickel-how-technology-is-changing-mining-and-people\/","title":{"rendered":"Artificial Intelligence at Nornickel: How Technology Is Changing Mining and People"},"content":{"rendered":"\n

Artificial intelligence is everywhere\u2014it writes texts, controls machines, recommends movies. But what does this mean for me personally? Will it replace my job, change my daily routine, or is it just a passing trend? At Nornickel, these questions are answered not in theory but in practice: the company is already implementing AI in its plants and offices. <\/p>\n\n

In this article, we learn from Danil Ivashechkin, head<\/a> of AI development and implementation at Nornickel, how the company uses new technologies, and supplement the material with comments from Andrey Doronichev, a Silicon Valley expert and founder of the AI startup Optic<\/a>. <\/p>\n\n

Why AI Is Not Just Another Programming Language<\/h2>\n\n

At first glance, it may seem that artificial intelligence is just another programming language, another IT trend. “The same mathematics, the same algorithms\u2014just a new name.” But this is precisely where the fundamental difference lies. <\/p>\n\n

“Artificial intelligence (AI) is fundamentally different from conventional programming. In classical coding, you write clear instructions for the computer: ‘if A, then do B.’ The algorithm is predictable and executes exactly what is programmed. <\/p>\n\n

\n

Machine learning works differently. Here, algorithms are not written directly\u2014the system learns from large amounts of data and finds patterns that humans cannot always explain. The results of such work can be unpredictable, yet effective.” <\/p>\nexplains Andrey Doronichev, founder of the AI startup Optic.<\/cite><\/blockquote>\n\n

Modern models, such as GPT (General Pretrained Transformer), use hundreds of billions of parameters and process incredible volumes of information. Essentially, they build vast statistical “maps of word proximity.” For example, the word “microwave” on such a map will be near “kettle” or “kitchen”\u2014because in texts these concepts often appear together. Therefore, if you ask the model which appliances are usually in the kitchen, it will suggest exactly such options. This is the principle on which large language models (LLMs) work: they select the most probable continuation based on patterns in the data. <\/p>\n\n

\"Example<\/a>
A simple example of how a neural network works\u2014it analyzes distinguishing characteristics, processes data, and produces an answer. Source: Roskachestvo<\/a> <\/figcaption><\/figure>\n\n

This is why it seems as though the system “thinks.” In reality, this is not thinking but statistical prediction of the next word or action. The illusion of understanding is created because the model can maintain a coherent dialogue, although mathematics remains at its core. <\/p>\n\n

It is important to understand: the idea of machine learning is not new<\/a>. The first research in this field was conducted in the mid-20th century. But for decades it remained theory\u2014humanity simply lacked the computational power. The real breakthrough occurred when modern graphics processors appeared. It was on them that models could be trained on massive datasets, and here NVIDIA played its role, betting on deep learning and transforming from a graphics card manufacturer into a technology giant<\/a>. <\/p>\n\n

AI as a New Arms Race<\/h2>\n\n

The first steps in creating generative artificial intelligence were taken by individual companies. OpenAI invested enormous resources in developing models, assembled the best engineers, and brought the technology to the mass market. Meta and Google followed suit, French company Mistral released its model, and Chinese DeepSeek caused a sensation by claiming it trained a GPT-level system for just $6 million and released it as open source. <\/p>\n\n

But gradually, the conversation stopped being about who would make the next convenient model for users. Artificial intelligence became a matter of strategic advantage<\/strong>, and the market turned into an arena of global competition. Today, this is a true “arms race” between the US and China. <\/p>\n\n

The Americans control key computing power\u2014NVIDIA H100 graphics processors, without which training large models is impossible. Chinese companies are forced to make do with the limited H800 version, but compensate with scale: they create clusters of thousands of processors and receive state support, which invests billions in startups, infrastructure, and technology parks. China has already surpassed the US in the number of patents, while the US still holds the lead in technology. <\/p>\n\n

\"NVIDIA<\/a>
Since 1993, NVIDIA has been developing graphics processors that quickly became popular among gamers. In recent years, the company has expanded its presence in other industries (cryptocurrency mining, AI), thanks to which NVIDIA’s stock value has soared to new heights, making it one of the most valuable technology companies in the world. Source: Nasdaq<\/a>, Medium<\/a> <\/figcaption><\/figure>\n\n

What About Russia?<\/h2>\n\n

Russia has not stayed on the sidelines in this race. We have our own developments\u2014for example, models from Yandex and Sber, which were created from scratch rather than based on open source. They show good results in Russian and are already used in products. <\/p>\n\n

But alongside this, there are serious limitations. The main one is the absence of a developed venture capital market. In the US and China, venture capital fuels innovation: hundreds of startups test hypotheses, dozens of them succeed and receive new investments. In Russia, there is almost no such mechanism: startups have nowhere to go for funding<\/a> except rare grant programs or individual investors. <\/p>\n\n

There is also a second problem\u2014the ideas themselves. Even at hackathons and pitch days, where innovation should be thriving, solutions that have long been implemented are often repeated. Danil Ivashechkin admits: “I come to a hackathon, and they propose visual recognition<\/a> for a conveyor or detection of people by hard hats. Such systems have already been implemented both in Russia and abroad. I want to see fresh ideas, not the tenth variation on computer vision.” <\/p>\n\n

Ultimately, Russia is formally present in AI development, but there is no systemic growth driver. Nevertheless, a niche for development remains: the country could focus on creating domain-specific models<\/strong>\u2014specialized solutions for industry, energy, or finance. Such systems do not become global products, but help to understand one’s own processes more deeply and increase efficiency, which is especially important for companies like Nornickel. <\/p>\n\n

How AI Already Works at Nornickel<\/h2>\n\n

At Nornickel, they decided not to limit themselves to theory and discussions\u2014artificial intelligence was immediately put to the test. And this is not about one-off experiments, but about projects that are already operational or being prepared for implementation. <\/p>\n\n

The first case is document search<\/strong> for lawyers and tax specialists. If previously one had to manually flip through hundreds of pages of contracts, the new model can find exactly the text fragment needed and immediately provide a link to the document. This saves hours of routine work and reduces the risk of errors. <\/p>\n\n

The second case is the “chief metallurgist’s assistant.”<\/strong> Specialists have hundreds of technological instructions, and previously, to answer a question, one had to go through them manually. Now the model reviews the array of documents itself and formulates an answer in a short and clear form. <\/p>\n\n

Another direction is operator work in production<\/strong>. Here, AI acts as a “co-pilot”: the system reports in advance what ore will arrive, tracks changes in production parameters, and even warns by voice if something goes wrong. The operator remains in charge of decision-making, but the algorithm takes on routine tasks, increasing speed and accuracy. <\/p>\n\n

A similar principle can be compared to Tesla’s autopilot: the car drives itself, but if it encounters a situation it cannot handle, the voice assistant immediately warns the driver\u2014take control. This increases the sense of control and safety: the person remains responsible, but the system makes the process more convenient and efficient. <\/p>\n\n

The next step, which the team is already working on, is a personal employee assistant<\/strong>. Unlike specialized solutions for individual departments, this is a tool for everyone. The assistant will be able to read a long document and summarize it in seconds, prepare bullet points for a presentation, translate text, compose a letter, or draft meeting minutes. Essentially, this is a universal assistant that handles all bureaucratic and routine workload. <\/p>\n\n

The main principle is that AI does not replace people but frees up their time. Where hours used to be spent on searching or rewriting, employees can focus on more complex and important work: analysis, decision-making, and creative tasks. <\/p>\n\n

\"Nornickel<\/a>
Nornickel developed Axioma\u2014a domestic solution that predicts and forecasts air pollution levels in real time. The product was created using AI technologies and digital twin technology. Source: Nornickel<\/a><\/figcaption><\/figure>\n\n

Replace or Assist: What to Actually Expect from AI<\/h2>\n\n

The question “will AI replace people?” always arises. Ivashechkin has a personal example from retail, where his career began. At that time, he was involved in demand forecasting: how many bananas, milk, or pasta would be purchased the following week. Initially, all work was based on Excel and simple algorithms, then the company decided to implement machine learning. The algorithm learned to “run through” thousands of products, forecast demand, and automatically order the required quantity. <\/p>\n\n

“It turned out that the department, where everyone was responsible for their category\u2014fruits, groceries, and so on\u2014seemed to become unnecessary. But we were not fired. We simply stopped sitting in Excel and SQL and began managing the development of these algorithms. Some went into data science, some became product managers. The team remained, the work just changed,” explains Ivashechkin.<\/p>\n\n

A similar story repeated at Nornickel, only now\u2014with industrial professions. When algorithms began to be implemented, operators initially thought they were going to be replaced. <\/p>\n\n

“There was a constant wave: if we succeeded at something, people got scared\u2014that’s it, we’re no longer needed; if we didn’t succeed, they said\u2014we told you people are better,” Ivashechkin recalls.<\/p>\n\n

Nornickel’s experience shows: artificial intelligence does not displace specialists but changes the nature of their work. Algorithms take on routine processes, while people remain at the center of the system\u2014they make key decisions, are responsible for results, and use technology as a tool. Ultimately, AI becomes not a replacement but a partner<\/a> that helps employees be more effective and frees up space for creativity and new tasks. <\/p>\n\n

\"Bill<\/a>
Bill Gates stated that AI can compensate for the shortage of doctors and teachers. Writers, fitness trainers, illustrators, transcribers, programmers, and even pop singers may end up on the “scrap heap of professions.” Source: Business Insider<\/a>, Dea<\/a>t<\/a>h by AI<\/a> <\/figcaption><\/figure>\n\n

How to Tame Technology: Control Without Slowing Down<\/h2>\n\n

If at the company level we already see practical benefits from AI\u2014from document search to operator support\u2014then at the societal level another question arises: what risks does it carry and who will be responsible for their consequences? Any powerful technology is like medicine: in the right hands it helps, in the wrong hands it becomes poison. Therefore, the conversation about artificial intelligence inevitably turns to the topic of regulation. <\/p>\n\n

When the conversation turns to regulating artificial intelligence, Ivashechkin immediately recalls the most obvious threats: deepfakes and voice forgeries<\/strong>. <\/p>\n\n

\n

“Today, algorithms can not only change a face in a video but also completely copy a voice. Imagine: supposedly your child or grandchild calls and urgently asks you to transfer money. Or as in the story with ‘Brad Pitt’\u2014when a woman was deceived by a voice substitution. It’s funny until you realize this is a real threat.” <\/p>\nsays Ivashechkin.<\/cite><\/blockquote>\n\n

Such scenarios, in his opinion, need to be regulated now. But at the same time, Danil Ivashechkin warns against excessive government pressure<\/strong>. He cites the example of Europe, where at some point companies were required to disclose in advance what their models were trained on and prepare dozens of pages of documentation. <\/p>\n\n

“It seems harmless, but in fact it kills startups at the start. The model already costs millions, and then there’s paperwork on top. As a result, many projects in Europe simply didn’t take off. This is not about safety, this is about slowing down progress,” notes Ivashechkin. <\/p>\n\n

In Danil Ivashechkin’s opinion, regulation should be targeted and risk-based, not based on the principle of “regulate everything that moves.” Otherwise, innovations simply won’t take off. But there is another extreme\u2014letting the process run completely unchecked, which is fraught with increased fraud and the emergence of dangerous technologies. <\/p>\n\n

Andrey Doronichev adds: to develop effective rules, a deep understanding of the technology itself is needed, and that is not yet available. Artificial intelligence is too powerful a tool, comparable in impact to nuclear physics. Therefore, government involvement is inevitable, but it is important that professionals are included in the process. Responsibility also lies with the companies creating foundational models: they will have to cooperate with the government and establish responsible practices. <\/p>\n\n

In Doronichev’s opinion, it is impossible to effectively regulate the industry right now. But in the long term, it is inevitable: like nuclear technology or energy, AI will sooner or later be framed by a system of rules. Today, it is important to maintain balance: industry and government must move together. <\/p>\n\n

\"Politico<\/a>
Progress cannot be stopped\u2014otherwise your country risks being left behind, as has already happened with Europe, where excessive regulation has slowed down technology development. Source: Politico<\/a> <\/figcaption><\/figure>\n\n

How Your Life Will Change with Artificial Intelligence<\/h2>\n\n

The future of artificial intelligence is envisioned on several levels. For companies, this is primarily domain-specific models\u2014assistants that know everything about the internal life of the business. They take on routine tasks: processing documents or finding the necessary data. This approach saves time and optimizes processes, leaving people more space for creativity. <\/p>\n\n

For individuals, AI in the future will become a true assistant in everyday life, entertainment, and medicine. Imagine: to go on vacation, you don’t need to spend hours searching for tickets and hotels. It’s enough to write to the assistant “flying to Thailand\u2014buy everything I need,” and the system will select flights, accommodation, and services on its own. The same goes for entertainment: if you want tickets to a match\u2014just specify the price and row, and AI will do the rest. <\/p>\n\n

Games will also change beyond recognition. Characters (NPCs) that previously mechanically repeated scripted lines will now react like living interlocutors\u2014you can talk to them, and each dialogue will be unique. The storyline will become variable and adapt to the player: at the beginning of the game, you can specify interests and preferences, and the story will unfold accordingly. <\/p>\n\n

But the greatest hopes are associated with medicine. Doronichev describes how his team uses AI for drug selection: algorithms analyze pharmaceutical data accumulated over decades and test billions of molecular combinations in minutes. Previously, specialists could test only a few thousand variants per day, now\u2014billions. This is how a candidate molecule against breast cancer was found. <\/p>\n\n

“Every experiment is like launching a rocket: at the start, you did everything possible, but you only find out whether it will fly at the moment of launch,” explains Doronichev.<\/p>\n\n

\"Transformer<\/a>
In the first stage (a), the neural network is pre-trained; in the second (b), it identifies the target that the selected molecule can reach. Source: ACS Publications<\/a> <\/figcaption><\/figure>\n\n

In terms of scale of impact, artificial intelligence is increasingly compared to electricity: as soon as it appeared, literally everything around changed. But, like any technology, AI rests on three pillars\u2014people, data, and computing power. Data and power multiply every year, but people remain the main link. It is up to us how to use this tool: accelerate the search for medicines, create new formats of education and entertainment, or use it for spam and fraud. <\/p>\n\n

What kind of future do you want to see with artificial intelligence? Let’s dream: what could our world become\u2014from production and science to everyday life? What are the boldest predictions that come to mind? <\/p>\n\n

The material was prepared with the support of the Russian Ministry of Education and Science within the framework of the Decade of Science and Technology.<\/p>\n","protected":false},"excerpt":{"rendered":"

Artificial intelligence is everywhere\u2014it writes texts, controls machines, recommends movies. But what does this mean for me personally? Will it replace my job, change my daily routine, or is it just a passing trend? At Nornickel, these questions are answered not in theory but in practice: the company is already implementing AI in its plants […]<\/p>\n","protected":false},"author":12,"featured_media":57337,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_seopress_robots_primary_cat":"none","_seopress_titles_title":"Nornickel and Artificial Intelligence: How AI Is Transforming Industry","_seopress_titles_desc":"How Nornickel Implements Artificial Intelligence: Examples, Trends, and Risks. What Awaits Industry and People in the AI Era ","_seopress_robots_index":"","footnotes":""},"categories":[572,549],"tags":[574,573,606],"tag-cat":[599,594],"class_list":{"0":"post-57330","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-it","8":"category-mining","9":"tag-artificial-intelligence-in-geology","10":"tag-automation-and-robotization","11":"tag-digital-technologies-in-geology","12":"tag-cat-it-and-artificial-intelligence","13":"tag-cat-mining"},"acf":[],"pbg_featured_image_src":{"full":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1.webp",1707,1139,false],"thumbnail":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-150x150.webp",150,150,true],"medium":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-300x200.webp",300,200,true],"medium_large":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-768x512.webp",768,512,true],"large":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-1024x683.webp",1024,683,true],"1536x1536":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-1536x1025.webp",1536,1025,true],"2048x2048":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1.webp",1707,1139,false],"bricks_large_16x9":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-1200x675.webp",1200,675,true],"bricks_large":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-1200x801.webp",1200,801,true],"bricks_large_square":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-1200x1139.webp",1200,1139,true],"bricks_medium":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-600x400.webp",600,400,true],"bricks_medium_square":["https:\/\/geoconversation.org\/wp-content\/uploads\/2025\/09\/oblozhka-1-1-600x600.webp",600,600,true]},"pbg_author_info":{"display_name":"\u042e\u043b\u0438\u044f \u0424\u0440\u043e\u043b\u043e\u0432\u0430","author_link":"https:\/\/geoconversation.org\/en\/author\/giulia-nikolaevna\/","author_img":false},"pbg_comment_info":" No Comments","pbg_excerpt":"Artificial intelligence is everywhere\u2014it writes texts, controls machines, recommends movies. But what does this mean for me personally? Will it replace my job, change my daily routine, or is it just a passing trend? At Nornickel, these questions are answered not in theory but in practice: the company is already implementing AI in its plants…","_links":{"self":[{"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/posts\/57330","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/users\/12"}],"replies":[{"embeddable":true,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/comments?post=57330"}],"version-history":[{"count":1,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/posts\/57330\/revisions"}],"predecessor-version":[{"id":57338,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/posts\/57330\/revisions\/57338"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/media\/57337"}],"wp:attachment":[{"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/media?parent=57330"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/categories?post=57330"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/tags?post=57330"},{"taxonomy":"tag-cat","embeddable":true,"href":"https:\/\/geoconversation.org\/en\/wp-json\/wp\/v2\/tag-cat?post=57330"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}