The launch of the “Sherkala” model underscores the commitment of Inception, a G42 company, in collaboration with the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), to advancing artificial intelligence to serve diverse languages that are underrepresented in the tech world. This collaboration centers on enhancing the capabilities of the Kazakh language and providing robust support to its more than 13 million speakers worldwide. The model is pioneering in its integration of cutting-edge artificial intelligence technologies to meet the linguistic and cultural needs of various communities.
Definition of the “Sherkala” model
The “Sherkala” model is one of the advanced large language models, containing 8 billion parameters and trained on more than 45 billion words primarily covering the Kazakh language, alongside other languages such as English, Russian, and Turkish, to achieve broader linguistic diversity. The development of “Sherkala” relied on advanced techniques inspired by the “Llama 3.1” model, with a 25% expansion of the tokenizer to increase the efficiency of Kazakh language comprehension and high-accuracy text generation. The model was trained using “Condor Galaxy,” one of the most powerful supercomputers dedicated to artificial intelligence applications.
The importance of supporting the Kazakh language
Figures indicate that millions of Kazakh speakers previously lacked technical platforms offering AI-driven linguistic services. Accordingly, “Sherkala” aims to bridge the digital divide in the Kazakh-speaking world by providing unprecedented opportunities for communication, text comprehension, analysis, and generation in a way that reinforces cultural identity. This project complements the efforts of Inception and MBZUAI in empowering other languages, such as Arabic through Jais and Hindi through Nanda, and now extending to the Kazakh language, thereby establishing the principle of inclusivity in the AI world.
Leadership and efficiency in language processing
The “Sherkala” model is distinguished not only by its large size but also by its exceptional ability to conduct conversations, provide accurate answers, and generate high-quality text. It has undergone rigorous testing regarding Kazakh culture, history, and knowledge, which has enhanced its deep understanding of various subjects. Notably, the model managed to outperform other models exceeding 70 billion parameters in certain tests, thanks to its efficiency and speed in generating text.
The role of Inception and MBZUAI
Inception possesses extensive experience in designing and developing cutting-edge AI solutions, and Dr. Andrew Jackson, CEO of the company, confirmed that these efforts aim to serve underrepresented linguistic communities. Its partnership with MBZUAI aims to unite efforts and resources to build a robust model that effectively serves the Kazakh language. For its part, MBZUAI plays a pivotal role in research and development by focusing on natural language processing techniques and providing experts and resources to achieve qualitative breakthroughs in this field.
Inclusivity and responsibility in artificial intelligence
“Sherkala” points to the expansion of artificial intelligence to serve diverse languages without overlooking the importance of responsible innovation. This vision is evident in the pursuit of ensuring model transparency and expanding the scope of benefit to all linguistic and cultural groups, bringing various communities closer to a future driven by innovative technology and smart services. This trend also fosters the spirit of research and academic cooperation between institutions and companies to achieve sustainable and safe AI solutions.
“Sherkala” represents a qualitative leap in the field of large language models dedicated to languages that are underrepresented on smart platforms. This step comes within a broader vision targeting the expansion of cognitive and cultural communication opportunities for millions of people in a world where digitization is growing in importance. “Sherkala” is expected to contribute to enhancing the linguistic identity of multiple communities, especially Kazakh speakers, paving the way toward a fairer and more inclusive digital future in the AI sector.
