Key facts
- Taiwan is developing a local generative AI model called Taide.
- Taide uses traditional Chinese characters and local data to reflect Taiwanese perspectives.
- The initiative aims to counter political bias and preserve language and culture from mainland Chinese influence.
- A previous AI chatbot project exhibited pro-China political leanings due to biased training data.
- Taiwanese companies are increasingly adopting closed-loop AI systems for data privacy and security.
Taiwan is accelerating the development of its own generative artificial intelligence model, Taide, to safeguard its language, culture, and democratic values against the influence of mainland Chinese content prevalent in global AI systems. This initiative is a response to concerns that existing models reflect Beijing's political narratives, as highlighted by a previous incident where a beta version of a Chinese-language AI chatbot from Academia Sinica incorrectly identified China's president as Taiwan's leader and cited China's national day.
Taide is being built on open-source foundations but will be trained on local data, including content from Taiwanese media and government sources, and will utilize traditional Chinese characters. This focus on cultural authenticity and democratic principles distinguishes it from models prioritizing scale. The project is seen as a critical step towards digital sovereignty, especially following incidents like the Samsung data leak, which underscored the risks of using foreign AI platforms for sensitive operations. Companies like Asustek are exploring closed-loop AI systems to keep hardware and data within Taiwan.
Developing a competitive AI model with limited resources presents challenges, but Taide's localized approach is viewed as an advantage for customization. Professor Jyh-shing Jang noted that companies can use Taide as a foundation for specific applications, deployable even on mobile devices without external dependencies. This strategy aligns with Taiwan's broader AI governance framework, which emphasizes responsible innovation over rapid expansion, contrasting with development approaches in mainland China. The initiative is part of a wider regional movement in Asia towards developing indigenous AI capabilities, with countries like Singapore, South Korea, and Japan also working on their own language models.