Kazakhstan will soon have its first national AI model
Author - Mansur Ismagulov
The Tech AI
Get the key points in seconds - a short summary from The Tech AI.
ISSAI NU is developing a large language model, KAZ-LLM, for the Kazakh, Russian, and English languages. The partner is Beeline Kazakhstan and QazCode, which provided 8 DGX H100, open data, and specialists. Previously, the operator had already launched the Kaz-RoBERTA-conversational model, which has been downloaded more than 3,000 times on Hugging Face.
Artificial intelligence opens up new opportunities for business, but not all users can fully leverage its benefits. Most AI systems are developed for high-resource languages such as English, Spanish, or Russian, creating a significant language gap and depriving many people, including Kazakhstani citizens, of access to advanced technologies in their native language, which are essential for development and improving lives in countries with low-resource languages.
In Kazakhstan, the issue of the AI language gap is being addressed by a scientific working group led by the Institute for Smart Systems and Artificial Intelligence at Nazarbayev University — ISSAI NU. They are working on creating a large language model called KAZ-LLM, which will cover the key languages for Kazakhstani citizens: Kazakh, Russian, and English, so that everyone can use digital technologies in their native language. This foundational model will serve as the basis for developing local services and products, as it will be published in open access.
Beeline Kazakhstan and its subsidiary IT company QazCode are partners in the creation of the national large language model KAZ-LLM. This is not the operator's first experience in developing AI. Last year, Beeline launched and made publicly available for all developers the Kaz-RoBERTA-conversational model, which is actively used for servicing subscribers on digital platforms. On the Hugging Face platform, it has been downloaded more than 3,000 times.
Thanks to Beeline Kazakhstan and QazCode, the project to create KAZ-LLM, overseen by ISSAI, is equipped with the necessary infrastructure. The company provided cloud computing resources totaling 8 DGX H100 units, significantly increasing the volume of training data and the model's training capabilities. The company also provided collected open data, and QazCode data scientists joined the joint working group for model training.
This contribution to the development of the national large language model KAZ-LLM reflects Beeline's "digital operator" strategy, as well as its commitment to overcoming the language gap in AI technologies. To this end, the operator signed a Memorandum with Barcelona's supercomputer center, which specializes in the development of AI for various language groups, and also announced the creation of the first GPU cloud in Central Asia for developing AI products based on NVIDIA technology.
The Tech on Telegram Follow The Tech on Telegram
Keep up with the latest technology and startup news.