Indosat Ooredoo Hutchison (Indosat or IOH) and Tech Mahindra have recently announced a memorandum of understanding (MoU) at Mobile World Congress (MWC) Barcelona 2024 to collaborate on the development of Garuda. Garuda is a large language model (LLM) designed to preserve the Indonesian language and its various dialects, emphasizing linguistic diversity and cultural heritage. The joint initiative aims to support digital transformation in Indonesia, aligning with the Golden Indonesia 2045 vision, utilizing AI infrastructure and human capital readiness to drive innovation and contribute to socio-economic development.
Garuda, built on the principles of Tech Mahindra’s indigenous LLM Project Indus, will offer unique features facilitating applications across industries such as healthcare, e-commerce, education, and finance. The model’s capabilities include enhancing customer support, improving user experience, and aiding content creation. This collaboration underscores the commitment to Indonesia’s digital transformation by fostering digital inclusion and accessibility.
Tech Mahindra, a prominent digital transformation solutions provider, will leverage its technology expertise to collect and curate data in the Indonesian language. This data will be used to pre-train Garuda, providing a conversational model for Indosat’s customers. The project aims to preserve linguistic diversity, enhance accessibility, and promote inclusivity in the digital space. The collaboration aligns with Indosat Ooredoo Hutchison’s broader mission of connecting and empowering every Indonesian.
Garuda is set to be developed with 16 billion original Indonesian tokens, offering 1.2 billion parameters to shape the model’s understanding of the language. This beta version will undergo testing with Indonesian-language speakers, with further enhancements using reinforcement learning from human feedback (RLHF) techniques to ensure robust conversation capabilities. Additionally, specialized use cases will be developed using the less-is-more-for-alignment (LIMA) method.
The large language model market is anticipated to reach US$40.8 billion by 2029. Tech Mahindra believes that LLMs like Garuda and Indus can revolutionize online communication by enabling people and enterprises to communicate in local dialects and languages. The partnership between Indosat and Tech Mahindra signifies a strategic alliance, leveraging their respective strengths to drive technological innovation in Indonesia. The collaboration reflects a shared vision to advance language-understanding technology, contributing to linguistic diversity preservation and tailored communication solutions for the Indonesian market.