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

China's DeepZang AI model, the country's inaugural Tibetan-language large language model, is set to accelerate upgrades in order to foster ethnic harmony. Discussions on enhancing DeepZang were held in a seminar on Monday, attended by experts from research establishments, universities, and businesses, as reported by the Chinese news outlet Chinanews.com.

The event, located in Hohhot, Inner Mongolia's autonomous northern region, addressed matters such as refining Tibetan AI technologies and incorporating DeepZang into multilingual real-time translation. Tenzin Norbu, DeepZang's creator, committed to ongoing improvements and practical applications of the technology, highlighting how technological progress can "enhance ethnic interaction and integration" and boost digital growth in minority regions.

DeepZang was unveiled in March, approximately three years after the introduction of Monlam AI, a Tibetan-language large language model created by the non-profit Monlam Tibetan IT Research Centre in Dharamsala, home to the Dalai Lama and the Tibetan administration-in-exile since 1960. Tibet is a top concern in ethnic and religious matters for Beijing, which persistently strengthens domestic Tibetan studies to regain control of global discourse on the subject.

DeepZang is a component of this larger initiative to reshape global perceptions of Tibet by providing authentic data and technology. Beijing's initiatives in ethnic minority regions, including Tibet, have been a key point of discussion between Western governments and human rights organizations, who frequently accuse China of human rights violations.

China firmly denies these allegations. In March, China enacted its Law on Promoting Ethnic Unity and Progress, intended to counter Western ideological influence and establish a legal basis for assimilating minority groups. The law came into effect in July. Trained on over 70 million Tibetan texts and 30,500 hours of audio from the language's three main dialects, DeepZang powers AI chat, real-time translation, and speech-to-text functions in Tibetan, Chinese, and English.

Tibetan AI technologies have traditionally grappled with low-resource issues due to their unique linguistic structures, dialect differences, and scarce digital databases, complicating precision and versatility. Monday's Chinanews.com report stated that DeepZang has consistently overcome these obstacles, "bridging critical technological gaps in the field and strengthening China's technology sector for ethnic language AI".

Tenzin Norbu acknowledged that AI development in other Chinese ethnic minority languages also faced low-resource challenges, underscoring the importance of cross-regional and cross-linguistic collaboration. The seminar also outlined future plans to integrate DeepZang into public services, education, healthcare, local governance, and the dissemination of traditional Tibetan culture, ensuring all ethnic groups equally benefit from technological advancements.