27.12.2024 18:31:00
Дата публикации
The training of a large language model KazLLM based on 148 billion tokens in Kazakh, English, Russian and Turkish has been completed. The model was developed by the team of the Institute of Smart Systems and Artificial Intelligence (ISSAI) with the support and coordination of the MCRIAP RK and the MNVO RK.
Such linguistic institutes and research and production organizations as Til Kazyna, JSC NIT, Maqsut Narikbayev University and other institutes contributed to the implementation of this project.
At the initiative of the President of the Republic of Kazakhstan, KazLLM will become the basis for the creation of a larger project - TurkLLM, aimed at developing natural language processing technologies in the Turkic-speaking space.
Almost 4 billion tenge were allocated for the large language model project.
At the first stage, KazLLM will be openly available to developers, startups and companies to stimulate the creation of products and services based on it.
Beeline Kazakhstan and QazCode played a key role in the creation of the technology, providing computing power based on DGX H100 servers. These resources made it possible to reduce the training process to 50 days, ensuring high performance of the model.
KazLLM opens up prospects for the automation of various areas. Companies will be able to develop chatbots, improve customer services, analyze large amounts of data and create educational platforms for learning the Kazakh language.
An AI assistant is currently being tested in eGov, which should simplify the process of obtaining information and interacting with government agencies. The KazLLM language model has been implemented in the project. The service can be used in Kazakh and Russian.
"You can dial the service you need by voice or text, and we will take you to it. It is not necessary to know the exact name of the service. It is enough, for example, to dial: I had a baby. What should I do? Our chatbot will give you a link to the service immediately after you register it,” said Alibi Dzhangildin, chief database architect at JSC NIT.
(The text was translated automatically)
Such linguistic institutes and research and production organizations as Til Kazyna, JSC NIT, Maqsut Narikbayev University and other institutes contributed to the implementation of this project.
At the initiative of the President of the Republic of Kazakhstan, KazLLM will become the basis for the creation of a larger project - TurkLLM, aimed at developing natural language processing technologies in the Turkic-speaking space.
Almost 4 billion tenge were allocated for the large language model project.
At the first stage, KazLLM will be openly available to developers, startups and companies to stimulate the creation of products and services based on it.
Beeline Kazakhstan and QazCode played a key role in the creation of the technology, providing computing power based on DGX H100 servers. These resources made it possible to reduce the training process to 50 days, ensuring high performance of the model.
KazLLM opens up prospects for the automation of various areas. Companies will be able to develop chatbots, improve customer services, analyze large amounts of data and create educational platforms for learning the Kazakh language.
An AI assistant is currently being tested in eGov, which should simplify the process of obtaining information and interacting with government agencies. The KazLLM language model has been implemented in the project. The service can be used in Kazakh and Russian.
"You can dial the service you need by voice or text, and we will take you to it. It is not necessary to know the exact name of the service. It is enough, for example, to dial: I had a baby. What should I do? Our chatbot will give you a link to the service immediately after you register it,” said Alibi Dzhangildin, chief database architect at JSC NIT.
(The text was translated automatically)
