During the launch of Ukraine’s first center of excellence for integrating AI solutions, the Ministry of Digital Transformation announced the creation of a national language model (LLM) for government services. Deputy Minister of Digital Transformation Oleksandr Bornyakov shared details of the development.
During the launch of Ukraine’s first center of excellence for integrating AI solutions, the Ministry of Digital Transformation announced the creation of a national language model (LLM) for government services. Deputy Minister of Digital Transformation Oleksandr Bornyakov shared details of the development.
«The Ukrainian LLM can be a superstructure over an existing system, for example, the same ChatGPT. We are currently considering how to try to build a domestic model based on Llama from Meta», — Bornyakov said in an interview with Mind.
According to him, Ukrainian LLM can use the core of another language model. But unlike ChatGPT and other AIs, this model should be focused on Ukrainian data.
«In general, Llama is a free open-source model. Use it as much as you want. However, Meta is interested in cooperation, so they provided us with resources — consultants, developers — to help develop our projects based on Llama», — said Oleksandr Bornyakov.
He said that specialists from the AI Center of the Ministry of Digital Affairs will coordinate the work on creating a Ukrainian LLM, but they will also be able to attract external resources. «We will negotiate with other companies, such as Meta. There are donors, governments that are interested in joining,» said the Deputy Minister of Digital Transformation.
We will remind you that in early February, the first AI Center Excellence in Ukraine was launched in Kyiv.
Oleksiy Minakov, an expert in PR, communications, and the application of AI technologies, told dev.ua that the idea of creating a national language model is quite logical.
«AI sovereignty is part of digital sovereignty. But here it is a question of details — there are doubts that at this historical stage and the level of costs for creating AI models, we are able to do it „from scratch“. Still, it is easier and more realistic to fine-tune ready-made models and, as far as I understand, this is exactly what will happen, because plans to use solutions from Meta, in particular the open source LIama model, have already been publicly announced», — the expert noted.
