tech
Hong Kong’s Votee AI is taking on English and Mandarin’s AI dominance with a Cantonese model
The AI boom is being driven by two languages: English and Mandarin Chinese. The leading AI developers—OpenAI, Anthropic, DeepSeek, Moonshot AI, Z.ai, and others—are almost all based in either the U.S. or mainland China, and so their models perform best in their native tongues.
TL;DR
- Leading AI development is dominated by companies in the U.S. and China, resulting in models that perform best in English and Mandarin.
- Votee AI is retraining open-source models on Cantonese data to build AI capabilities for the language, used by over 80 million people.
- Cantonese differs significantly from Mandarin in grammar and vocabulary, and existing mainstream models struggle with its cultural and local nuances.
- Votee AI trains its models affordably using various data sources, including online scraping, community contributions, and synthetic data, with significantly lower costs than frontier models.
- The effort aligns with the concept of "sovereign AI," where nations and companies seek to own their AI technology to avoid external dependencies.
- Votee AI aims to expand its services to other under-resourced languages in Southeast Asia and beyond, and is advised by Hong Kong tycoon Allan Zeman.