NewCorperateCR

Cantonese AI Takes On English and Mandarin Dominance

· business

The Language of Power: Why Cantonese AI Matters in a Post-Globalized World

The dominance of English and Mandarin Chinese in artificial intelligence (AI) has significant consequences for non-native speakers. Millions of people who speak languages like Cantonese have been relegated to the periphery of the digital revolution, despite being crucial to understanding what AI can and cannot do.

Pak-Sun Ting, CEO of Votee AI, asserts that “AI is essentially useless” without local language support. As a Hong Kong-based startup developing Cantonese language models, Votee is part of a growing movement to address this issue.

A key challenge in building AI models for non-English and Mandarin languages is the lack of standardized written data. Although 80 million people speak Cantonese, comparable to those who speak Korean, it has not led to a deep pool of standardized written data, particularly for colloquial Cantonese. Companies like Votee are addressing this gap by working with existing open-source models and retraining them on Cantonese data.

Building a Cantonese language model requires technical expertise and an understanding of cultural nuances and linguistic idiosyncrasies that set Cantonese apart from Mandarin. Votee’s approach involves developing AI models for various languages, including Indonesian languages like Bahasa, in collaboration with government agencies, universities, and private companies.

This “sovereign AI” initiative aims to create local language support without relying on overseas models or infrastructure. Votee is also developing smaller, more targeted models that can be used for specific applications, such as automating administrative work or providing basic customer support. These models may not be as powerful as their English- or Mandarin-language counterparts but are sufficient for their intended purposes.

The notion of sovereign AI raises questions about the global flow of data and the role of governments in shaping their own digital futures. As Ting notes, “AI has become such an essential need, and so you don’t want to be tethered to anybody else who can turn it off.” This sentiment is echoed by state-backed initiatives in countries like Singapore and South Korea.

However, sovereign AI also raises concerns about data fragmentation, interoperability, and standardization. Votee’s clients include both governments and private companies, highlighting the tension between local language support and global connectivity.

As we move forward in a post-globalized world, it’s essential to recognize the value of local languages and cultures. By supporting initiatives like Votee AI, we can improve the lives of millions of people who speak Cantonese and build a more resilient and inclusive digital ecosystem.

Reader Views

  • DH
    Dr. Helen V. · economist

    The article highlights the importance of developing AI models in non-dominant languages like Cantonese. However, I'd caution that simply creating language support isn't enough; we also need to address the systemic issues that lead to uneven data availability. For instance, why don't educational institutions prioritize teaching coding and AI development in local languages? Without such a shift, we'll continue to perpetuate a model where Western or Mandarin-speaking countries dictate the global AI agenda.

  • MT
    Marcus T. · small-business owner

    This Cantonese AI development is long overdue, but I worry about the sustainability of Votee's approach relying on open-source models and government partnerships. How do we ensure that these sovereign AI initiatives don't end up siloing languages further? What happens when funding dries up or collaborations falter? Can we really afford to have multiple separate language ecosystems, each with their own technical requirements and maintenance costs?

  • TN
    The Newsroom Desk · editorial

    The push for Cantonese AI is long overdue, but let's not forget that linguistic diversity in AI also means navigating complex regulatory frameworks and data protection laws. Governments need to step up and provide clear guidelines for local language support, rather than just encouraging companies to "go sovereign". Votee AI's initiative may be a game-changer, but it will only succeed if it can overcome the significant technical and cultural hurdles of standardizing Cantonese data. That won't be easy.

Related articles

More from NewCorperateCR

View as Web Story →