China’s DeepZang AI aims to transform Tibetan language tech, promote ethnic unity

China's inaugural Tibetan-language AI model, DeepZang, aspires to advance linguistic technology and foster ethnic unity. Discussions about enhancing the large language model were held during a seminar on Monday, which brought together specialists from research institutions, academic institutions, and businesses, according to state-run site Chinanews.com.

The meeting, held in Hohhot, Inner Mongolia, debated matters such as enhancing Tibetan artificial intelligence (AI) technology and broadening the model's use in multilingual simultaneous translation. Founding member Tenzin Norbu pledged to keep refining Tibetan AI technologies and extending their real-world applications. He also underlined the role of technological innovation in promoting ethnic interaction and integration, as well as supporting digital advancement in minority regions.

DeepZang debuted in March, about three years after the launch of Monlam AI, a Tibetan-language LLM created by Monlam Tibetan IT Research Centre, a non-profit group based in Dharamsala, a town in northern India that has housed the 14th Dalai Lama and the Tibetan administration-in-exile since 1960. China maintains Tibet as a key focus in ethnic and religious policies, continuously fortifying domestic Tibetan studies to regain control over international dialogue on the subject.

DeepZang is an integral part of this larger initiative to redefine worldwide perceptions of Tibet through its own authoritative data and technology. These efforts are also mirrored in Beijing's campaign since 2023 to utilize the romanized Pinyin name "Xizang" in place of "Tibet" in English articles, projecting its sovereign narrative on a global scale.

Ethnic minority policies in regions like Tibet have been a central concern for Western governments and rights organizations, who repeatedly accuse China of human rights abuses. China staunchly denies any wrongdoing. In March, China enacted its Law on Promoting Ethnic Unity and Progress, designed to counter Western ideological influence and establish a legal foundation for assimilating minority groups.

The legislation came into force in July. DeepZang has been trained on nearly 70 million Tibetan texts and over 30,500 hours of audio covering the language's three primary dialects. It facilitates AI chat, live translation, and speech-to-text services in Tibetan, Chinese, and English. Tibetan AI technologies have long grappled with low-resource challenges due to unique linguistic structures, dialect variations, and limited digital sources, making accuracy and adaptability tough to attain.

According to the Chinanews.com report, DeepZang has persistently overcome obstacles, "bridging critical technical gaps in the field and strengthening China's technology ecosystem for ethnic language AI." As noted in the report, Tenzin Norbu added that AI development in other Chinese ethnic minority languages also faced low-resource constraints, making cross-regional and cross-linguistic cooperation crucial.

The seminar also detailed forthcoming strategies to incorporate DeepZang into sectors like public services, education, healthcare, and local governance, while "empowering the digital preservation of traditional Tibetan culture and ensuring equitable benefit-sharing for all ethnic groups from technological progress."