The state's initiative to create its own national language model (LLM) has sparked lively discussion among the Ukrainian IT community. IT professionals share different opinions, weighing the potential benefits and possible risks of such a project.
The state's initiative to create its own national language model (LLM) has sparked lively discussion among the Ukrainian IT community. IT professionals share different opinions, weighing the potential benefits and possible risks of such a project.
The day before, it became known that the Kyivstar company will directly deal with the creation of the Ukrainian national LLM, plans for which the Ministry of Digital Affairs announced in March. The national LLM will be trained on Ukrainian data, will understand the national context, dialects and terms, and on the basis of the national LLM, AI products for the state and the defense sector will be launched .
Some developers see significant potential in this initiative. Volodymyr Kyrylenko, a fullstack developer at DevIT, hopes that, provided it is independent of political influence and developed as a purely technological product, Ukraine can get “a useful and cheap tool for government services without government money.”
Dmytro Voytovych, Frontend developer, supports the principle of "it's better to do something than nothing." In his opinion, even in case of failure, the project will allow you to gain valuable experience. "Maybe a normal European AI will finally appear," he adds, pointing to broader prospects.
However, there are those who express doubts about the feasibility of such a step. Maksym Merkulov, Co-founder of ReadEd and Product Manager, believes that the state should focus on its core functions, and not on complex technological solutions, especially when ready-made analogues already exist. “I want to be wrong, but this will be a finely tuned GPT model raised on its own server,” he suggests, hinting at the possible secondary nature of the development.
In contrast to the skeptical views, Tim Roman, UX/UI Designer at Extrawest LP, Electriqua, praised the Ministry of Digital Affairs’ decision to create a Ukrainian LLM. He urges critics not to ridicule the country’s attempts to do something of its own, citing arguments in favor of the project:
A similar opinion was expressed by Maksym Korzhenevsky, founder and CTO of MK-Consulting, who in an interview with dev.ua emphasized the importance of launching a national LLM.
“In the conditions of Russian aggression, this is extremely important. Yes, many models already know how to speak Ukrainian very well, but they are trained on open data, which is distorted, including by Russia Today. When a new model comes out and you ask her about Crimea or what happened to the Malaysian Boeing MH 17, she will answer that the flight “crashed”. That’s right, it just took off and crashed. Or what is the UPA? And the model answers: “This is an organization that tried to make Ukraine ethnically pure,” said Maksym Korzhenevsky.
He also added that from an economic point of view, investing in your own LLM is quite advisable, since most LLMs in the world spend significantly less resources on processing requests in English than in Ukrainian.
"Roughly speaking, the same sentence in Ukrainian and English costs different money. Processing in English can be twice, and sometimes even three times cheaper than in Ukrainian, because there are more tokens. If we create our own model with our own tokenizer, we will be on the same level as in English and it will be cheaper to simply twist and use these models," added the CTO of MK-Consulting.
EPAM reported yesterday that access to the national LLM via API will significantly reduce the cost of using AI technologies. The company emphasized that according to MarketsandMarkets, the global LLM market was estimated at $6.4 billion in 2024 and will grow to $36.1 billion by 2030, which will further stimulate the development of the national ecosystem of AI solutions.
Read more about the agreement between the Ministry of Digital Affairs and Kyivstar in the dev.ua material following yesterday's presentation.
Let us remind you that today
