Indosat & GoTo: How AI is Bridging Global Linguistic Gaps

The development of language models has been predominantly focused on major global languages, particularly English.

This trend has left many countries with diverse linguistic landscapes struggling to harness the full potential of AI technologies.

The gap is particularly pronounced in regions with multiple local languages and dialects, where the nuances of communication and cultural context play a crucial role in everyday interactions and business operations.

As the global AI market continues to expand, with projections suggesting it could reach US$1.8tn by 2030, there is a growing recognition of the need for localised AI solutions that can cater to specific linguistic and cultural needs.

This shift is not only about language translation but also about creating AI systems that understand and respond to local contexts, idioms and cultural references.

Innovating to address these linguistic gaps, Indonesia's leading telecommunications company Indosat and the country's largest tech firm GoTo have announced the development of Sahabat-AI, a large language model (LLM) ecosystem specifically designed for Indonesian languages.

Bridging the linguistic divide

Sahabat-AI, which translates to "AI Friend" in English, is an open-source project aimed at empowering Indonesians to build AI-based services and applications in Bahasa Indonesia and various other local languages.

The initiative is designed to address a critical gap left by global AI models, which often struggle with the nuances and context of local languages.

According to Vikram Sinha, President Director and CEO of Indosat: "Sahabat-AI is not just a technological achievement, it embodies Indonesia's vision for a future where digital sovereignty and inclusivity go hand in hand”.

This statement highlights the project's ambition to democratise AI technology and make it accessible to Indonesia's diverse population.

The development of Sahabat-AI is supported by AI Singapore and India's Tech Mahindra, utilising Nvidia AI Enterprise software, including Nvidia NeMo, to train the model and enhance its general language understanding.

This collaboration highlights the international nature of the project, bringing together expertise from across Asia to create a solution tailored for Indonesia.

Technical specifications and deployment

In its initial phase, Sahabat-AI will launch with large language models featuring 8-billion and 9-billion parameters.

While these figures may seem abstract to non-technical readers, they represent the complexity and capability of the AI model.

For context, larger parameter counts generally indicate a more sophisticated model capable of understanding and generating more nuanced language.

The model has been trained using Nvidia's full-stack AI platform, which provides the computational power necessary for processing vast amounts of linguistic data.

This training process is crucial for ensuring that Sahabat-AI can accurately interpret and generate text in Indonesian languages, including local dialects and cultural references that might be missed by more generalised AI models.

Patrick Walujo, Chief Executive Officer of GoTo, emphasises the inclusive nature of the project: "Our vision for Sahabat-AI is to put the power of AI into the hands of everyone in Indonesia".

This vision aligns with the broader trend of democratising AI technology, making it accessible to a wider range of users and developers.

Indosat & GoTo collaborations 

The project also involves collaboration with leading Indonesian universities, including the University of Indonesia, Gadjah Mada University, Bandung Institute of Technology and Bogor Institute of Agriculture.

Additionally, media groups such as Republika and Kompas Group are contributing to ensure Sahabat-AI is optimised for local context and cultural relevance.

This collaborative approach reflects Indonesia's culture of 'gotong royong', or mutual collaboration, demonstrating how industry, researchers and the public sector can work together to advance AI development on a national scale.

The impact of Sahabat-AI

The impact of Sahabat-AI is expected to extend beyond the technology sector.

By enabling AI-powered services in local languages, the project has the potential to accelerate digital literacy and drive growth across various sectors of the Indonesian economy.

For instance, Hippocratic AI, a startup focused on developing safety-oriented AI models for healthcare, plans to incorporate Sahabat-AI models into its services for Indonesian residents.

As the project progresses, Indosat Group will provide continued support for the ongoing development of Sahabat-AI's family of models using its GPU Merdeka sovereign AI cloud service, which features Nvidia Hopper-based accelerated computing.

This commitment ensures that the project will continue to evolve and improve over time, adapting to the changing needs of Indonesian users and businesses.

The launch of Sahabat-AI comes at a time of growing interest in Indonesia's AI sector, with international companies like Microsoft committing to establishing data centres in the country.

This development positions Indonesia as a potential hub for AI innovation in Southeast Asia, with Sahabat-AI serving as a cornerstone for future advancements in localised AI technology.

As Jensen Huang, Nvidia founder and CEO, notes: "Sahabat-AI launches Indonesia's AI journey and showcases how LLMs can be tailored to serve unique linguistic and cultural needs".

Make sure you check out the latest edition of Technology Magazine and also sign up to our global conference series - Tech & AI LIVE 2024

Technology Magazine is a BizClik brand