Key Points
- Experts are urging emerging nations to develop their own “sovereign AI” to avoid being left behind.
- Current top AI models are based on English and don’t reflect the unique cultures and languages of other countries.
- Using open-source AI models enables local talent to build their customized systems.
- The young, tech-savvy population of Southeast Asia (ASEAN) makes it a perfect place to build this new AI infrastructure.
Experts at a major tech conference in Bangkok warned on Friday that emerging economies must develop their own “sovereign AI” or risk being left behind in the global technology race. Sovereign AI refers to a country controlling its own AI technology, data, and infrastructure rather than relying solely on systems built in the U.S.
A key problem, one panelist explained, is that the world’s most popular AI models, like those from OpenAI and Anthropic, are built around the English language and Western culture. “The way you think… can be very different” when you speak another language, he said. Simply translating an English-based AI isn’t good enough; countries need AI that understands their unique languages and cultures from the ground up.
So, how can they do it? The panelists agreed that using open-source AI models is the key. Open-source models, such as Meta’s Llama, provide local tech talent in regions like Southeast Asia with the foundation to build their own customized AI systems. This creates a “collective energy” that helps the whole country compete.
But software isn’t enough. Countries also need access to massive computing power. The panelists noted that cloud companies, such as Amazon Web Services, and local providers are making this more accessible, allowing even small startups to experiment and build.
The push for sovereign AI is part of a larger global concern. The United Nations has warned that the benefits of AI are currently concentrated in just a few countries. By building their systems, emerging nations can help ensure the AI revolution benefits everyone, not just a select few.
