If you’ve used ChatGPT for local queries, you’ve probably noticed it struggles with Malaysian context. Ask about EPF withdrawals or local business practices, and the responses often miss important details that matter here.

Husein Zolkepli recognised this gap and started building AI models specifically for Malaysia through his company Mesolitica. His team creates AI that understands local languages, including the way Malaysians naturally mix English, Malay, and other languages in conversation.

What Makes Malaysian AI Different

Mesolitica’s MaLLaM (Malaysia Large Language Model) was trained on 349GB of Malaysian content, including social media posts, forum discussions, and government documents. The result is an AI model that better understands local context and communication patterns.

The difference is noticeable when handling Malaysian-specific queries. While international AI models might struggle with local expressions or cultural references, Malaysian-trained models respond more appropriately because they learned from local digital conversations.

Why This Matters for Malaysian Businesses

Malaysian businesses currently pay for AI services that aren’t optimised for local needs. Customer service chatbots trained on foreign data often provide responses that don’t match local communication styles. Translation services may miss cultural nuances important for Malaysian markets.

Open-source Malaysian AI offers an alternative. Companies can run these models locally rather than paying ongoing subscriptions to international providers. This approach also keeps sensitive business data within Malaysia rather than sending it overseas.

The local job market is adapting to include more AI-related roles. Workers with AI skills currently earn about 12% more than the national average, according to government employment data. Malaysian companies across sectors are looking for people who understand both technology and local business practices.

Economic Impact and Investment

The government projects that AI could contribute RM530 billion to Malaysia’s economy by 2030. This growth potential has attracted significant foreign investment, with Microsoft committing US$2.2 billion for data centres in Malaysia, followed by similar investments from other tech companies.

For smaller businesses, open-source AI reduces barriers to entry. Previously, building AI-powered applications required substantial budgets and specialised expertise. Now local developers can access Malaysian-trained models without licensing costs, making it easier to create solutions for local markets.

Getting Started with AI Skills

The government’s MyMahir platform provides information about in-demand AI skills and training opportunities. Basic skills like understanding AI tools and working with different models are becoming more valuable across various industries.

For business owners, experimenting with open-source Malaysian AI models is now accessible. Mesolitica’s models are available for testing without licensing fees. Companies can evaluate how these tools might fit their operations before making larger commitments.

Malaysia has moved beyond importing AI technology to developing our own. Husein Zolkepli and Mesolitica represent a new generation of Malaysian tech innovators who understand that building locally relevant AI gives us a competitive advantage in Southeast Asia.

These Malaysian AI models understand how we actually communicate – switching languages mid-sentence and mid-thought. This gives Malaysian companies an edge when expanding in the region where language develops organically at the speed of sound.

• TAGS:
• AI Model
• Husein Zolkepli
• Malaysia Large Language Model
• MaLLaM
• Mesolitica
• Open-Source AI