The effect of AI in the oil and gas industry may exceed 300 billion rubles per year

The Russian fuel and energy complex can receive an effect of more than 300 billion rubles. per year in the case of active use of generative artificial intelligence (AI), experts from Vygon Consulting believe. Generative AI could take on more than half of design engineering applications in the future. However, for this, the models need to be specially trained, and the cost of creating such world-class systems is approaching $1 billion, which in the Russian Federation is only possible with the consolidation of the efforts of the entire industry. Other experts note that fuel and energy companies have many more pressing tasks to digitalize their processes.

The total effect for Russian oil and gas companies from the use of generative artificial intelligence could amount to 343 billion rubles. in year. This assessment is contained in the review by Vygon Consulting entitled “Possibilities of using generative AI to ensure the technological sovereignty of the Russian fuel and energy sector.”

It is assumed that most of this effect can be achieved by increasing the efficiency of development, field development, geological exploration and production management projects.

Generative AI (GenAI), unlike conventional algorithms that are already widely used in Russian oil companies, has the ability to learn from a much larger volume of data (about three orders of magnitude higher than other types of neural networks). Thanks to this, GenAI will be able to increase automation of processes related to engineering and scientific and technical expertise. We are talking primarily about applied expertise, 59% of which can be automated. The potential for automating the work of professional groups related to engineering using neural networks will almost double, to 57%, according to Vygon Consulting.

A special case of generative AI are large language models (LLMs), which specialize in natural language processing tasks. Their most promising type is multi-agent systems, which in theory are capable of well solving a limited range of rather complex problems, partly replacing, for example, a geologist or a reservoir engineer.

The use of nuclear materials in the oil and gas industry is hampered by several circumstances.

First, training large models requires increased computing power. As an example, Vygon Consulting cites GPT-4, for training which 100 times more computational operations were used than for GPT-3, and the cost of training was close to $1 billion. Secondly, even the best models existing in the world ( Gemini 1, Claude 3 Opus, GPT-4) are poorly suited for solving specific problems in the oil and gas sector due to the lack of complete industry information with country specificities.

“Despite the rapid development, industry knowledge and functionality of nuclear engineering in the fuel and energy sector is still limited,” says Grigory Vygon, head of Vygon Consulting. “But in the near future they will learn to solve multi-level engineering problems, analyze existing technologies and create new ones. It is existentially important for Russian oil and gas industry, living under strict sanctions restrictions, export reorientation and energy transition, to join this race as quickly as possible and provide itself with applied tools to achieve technological sovereignty.”

Sanctions are limiting Russia’s ability to expand computing power, but the key obstacle is the high cost of development. Thus, according to experts, the cost of creating from scratch one megamodel of the Claude 3 level, released at the end of February 2024, exceeds $500 million (approximately 50 billion rubles). For comparison: Russian large IT companies are investing a total of about 48.3 billion rubles in the development of generative AI. in year. “At this level of investment, creating a world-class megamodel is only possible if we consolidate the efforts of all our companies,” the review says. And to create competitive models, experts believe, in the future it is necessary to increase investments to 100 billion rubles. in year.

However, other experts believe that investing in GenAI is not a priority for Russian companies now. Now they are faced with problems even in the field of much less complex systems, such as automated process control systems, notes Timofey Khoroshev, leader of the technology consulting practice at DRT. Companies are dealing with issues of basic automation, so they “have no time for artificial intelligence right now,” agrees Maxim Malkov, head of the practice for providing services to oil and gas companies at Kept. In his opinion, for now oil and gas companies are focused on import substitution of existing software products, most of which Russian vendors will have to develop independently.

Olga Mordyushenko