AI Singapore and Google launch new Project SEALD partnership to strengthen LLM for SEA region

AI Singapore (AISG) has announced they are kicking off the new Project SEALD initiative with Google that aims to enhance LLMs powered by languages of the South East Asia region.

The context begin with AISG wanting to support the training models under SEA-LION (Southeast Asian Languages in One Network) initiative so that it can serve the region better through better cultural contexts and linguistic nuances. Languages included right now are Indonesian, Thai, Tamil, Filipino, and Burmese.

Thanks to the partnership with Google Research Asia Pacific (APAC), they are working together to develop translocalized and transation model, enabling best practices for instruction tuning datasets, create tools to enable the job at scale, and publishing these pre-trained recipes in full open source motion.

They are hoping the results could aid and reach different types of important domains from redressal of worker grievances and extension of assistance schemes as well as ecosystem partners from academia to government.

In terms of details of the collaboration, AISG will be releasing SEA-LION LLMs on Google's Cloud Model Garden on Vertex AI for 1st and 3rd party access that enables enterprise-grade tools for customization and deployment. Publishing on Hugging Face will continue as well.

Notable collaborations include signed MOUs and LOIs between AISG and Indonesia + Malaysia + Vietnam-based entities, partner engagement in Thailand + Indonesia + Philippines for regional language syntax and semantics, and working locally with public sector and R&D stakeholders.

Interested personnel with relevant expertise and skills in LLMs and AI training can contact AISG via seald@aisingapore.org.