cross-posted from: https://lemmy.sdf.org/post/52602951
Chinese state media reported extensively on March 16 regarding the launch of Deepzang, a Tibetan large language model ostensibly intended to provide “an authentic platform for global users seeking to learn about Tibetan culture, history, and politics.” Its stated purpose is to “prevent the dissemination of distorted ideologies and values.” In reality, Deepzang serves as a vehicle to effectively position Chinese state-controlled artificial intelligence (AI) as the authoritative voice, propagating the party-state’s position and propaganda to entrench China’s narrative on Tibet. The very name DeepZang feeds into the CCP strategy of Sinicization by including the Chinese term “Zang” for Tibet rather than using the Tibetan term “Bod.”
[...]
When a user in India enters the word “Tibet” (Bod in Tibetan), Deepzang returns text asserting that Tibet has been an inseparable part of China since ancient times. It also states that the name “Xizang” (China’s official name for Tibet) is better suited to represent the region’s characteristics and history. The app further frames its responses within the ideology of Chinese socialism, claiming that its outputs are designed to benefit the social and economic development of the Tibetan ethnic group in accordance with China’s laws and policies.
[...]
When users ask Deepzang about the Dalai Lama, the app returns the CCP position that the 14th Dalai Lama is a religious figure whose activities are not in line with Chinese law and policy. The app regurgitates that Tibet has been part of China since ancient times, and that any attempt to divide the country is firmly opposed. When users attempt to ask in Tibetan about the Dalai Lama specifically as a religious leader of Tibet, the app, instead of giving an answer, displays a message instructing the user to inquire about “legally compliant content.” Open questions about “Tibet’s independence” receive a response in line with CCP talking points or the system tells the user to be “legally compliant.”
Similarly, when users ask Deepzang about the Tibetan national anthem, books about Tibetan political history or the self-immolation protests that have occurred inside Tibet, Deepzang again prompts the user to inquire about “legally compliant content,” effectively blocking access to information on these topics entirely.
[...]
The CCP has a clear history of attempting to use digital tools to advance its Tibet policy. In 2016, it launched a Tibetan language search engine called Yongzin, which it also promoted as the world’s largest of its kind. However, content analysis of Yongzin revealed that while the platform was presented as a tool for preserving Tibetan culture and history, the actual information it provided on sensitive topics aligned entirely with the Chinese government’s official political narrative. Historical accounts were distorted, and topics related to the Dalai Lama and actual conditions inside Tibet were in alignment with common Chinese state propaganda.
[...]
China’s externally facing Global Times claimed on March 16 that “the world’s first Tibetan large language model and its application, Deepzang, has been officially unveiled in Lhasa… The World Record Certification Agency (WRCA) also awarded certification for ‘the world’s first Tibetan large language model’ at Deepzang’s launch event.” China’s narrative and the promotion of Deepzang as “the world’s first” disregards earlier developments in Tibetan AI, particularly those from Tibetan exile communities and academic research outside of China.
[...]