The Ukrainian AI model has been named ‘Syaivo’: how the vote went at ‘Dii’
This has been reported by the Ministry of Digital Transformation.
The Ministry of Digital Transformation stated that, following the open vote, the name “Syaivo” for the national large language model was supported by 22,601 users. A total of 136,090 people took part in the vote.
The statement notes that “Syaivo” won by a margin of three thousand votes. Second place went to “Slovo”, and third to “Gomin”.
The final ten also included “Kavun”, “Pytay”, “Shipit”, “Yadro”, “Hoverla”, “Dzvinka” and “Shukay”. According to the Ministry of Digital Transformation, the results were as follows: “Syayvo” – 22,601 votes, “Slovo” – 19,546, “Gomin” – 18,301, “Kavun” – 16,978, “Pytay” – 15,214, “Shypit” – 11,228, “Yadro” – 8,635, “Hoverla” – 8,492, “Dzvinka” – 8,379, “Shukay” – 6,716.
The article also notes that ten options made it to the final selection, chosen from over 3,000 user suggestions.
What is known about the Ukrainian large language model
In June, the Ministry of Digital Transformation, together with Kyivstar, began joint development of a national large language model. Its beta testing is scheduled for late spring this year.
Large language models – LLMs – are artificial intelligence systems that use neural networks to process, analyse and generate text. They are trained on large volumes of data, enabling them to understand language, context and style, as well as to create coherent texts.
The article explains that LLMs are built on transformer architecture and predict word sequences to generate responses, code or articles. Such models are used in chatbots, translation tools, text analysis and programming. Notable examples include GPT-4, Gemini and Llama.
It is expected that the Ukrainian language model will enhance security, as foreign LLMs may ignore the Ukrainian context and be vulnerable to propaganda. Local data processing, as noted, is critically important for the application of artificial intelligence in defence, medicine and finance.
What else has been written about artificial intelligence
Last year, UNIAN reported that the most advanced artificial intelligence models failed the Ukrainian External Independent Testing (ZNO). They found visual-text tasks the most challenging.
The article notes that some tasks required logical thinking, whilst others demanded an accurate interpretation of instructions in the Ukrainian language. According to the test results, the artificial intelligence failed to reach the threshold of 70% correct answers.
It was also reported that so-called ‘memes’ can impair cognitive abilities and critical thinking not only in humans but also in artificial intelligence. According to experts, language models trained on large amounts of popular but low-quality content may, over time, lose effectiveness and become ‘dumber’.
As a reminder, the Diya portal has opened registration for beta testing of a new service for active-duty personnel of the Armed Forces of Ukraine – the submission of applications for a one-off cash grant from the Ministry of Defence of Ukraine. The service is being developed as part of the expansion of digital services for defenders and their families.
