Kazakhstan has launched ISSAI KAZ-LLM, its first artificial intelligence (AI) Large Language Model (LLM), based on advanced neural network technology and specially designed for the Kazakh language, local press reported.
This model, which serves as the basis for the Kazakh version of ChatGPT, was designed by researchers from the Institute of Intelligent Systems and Artificial Intelligence (ISSAI) of Nazarbayev University in Astana and was presented to the President Kazakh Kassym-Jomart Tokayev.
ISSAI KAZ-LLM is tailored to the multilingual and multicultural environment of Kazakhstan and supports Kazakh, Russian and English, with additional functionality for Turkish.
The model addresses language barriers and serves to improve the field of generative artificial intelligence for resource-limited languages.
« Our model understands Kazakh, Russian, English and Turkish and can perform tasks such as translation and text summarization, which are particularly useful for analytical work“, Madina Abdrakhmanova, deputy director of ISSAI for external relations and head of the data science team, told the press.
Training data comes exclusively from publicly available materials, such as Kazakh websites, news articles and online libraries, supplemented by contributions from various organizations.
Founded in 2019, ISSAI started with a small team tasked with advancing AI research and building datasets for the Kazakh language.
The ISSAI team has already developed the Soyle application, Kazakhstan's first multifunctional voice application. Unlike KazLLM, a research project, the Soyle app is a product developed entirely based on previous research and was launched last November 20.
Kazakhstan is also preparing to launch the alem.ai International Center for Artificial Intelligence, set to become a hub for transforming the country into an AI-driven economy, encouraging innovation, attracting investments and supporting startups. By 2029, exports of Kazakh AI solutions are expected to reach $5 billion.
Kazakhstan joins countries such as Korea, China, the United Arab Emirates, the United Kingdom and France in developing linguistic models adapted to their cultural and historical contexts.
With MAP