Workflow
DeepZang
icon
Search documents
西藏人工智能从场景应用走向藏语大模型研发
Zhong Guo Xin Wen Wang· 2025-11-30 03:41
Group 1 - Tibet has made significant progress in the development of Tibetan language large models, marking a transition from application scenarios to systematic research in AI development [1] - The "Sunshine Qingyan" V1.0 model, a billion-parameter Tibetan language base model, was recently released, showcasing the advancements in AI technology in Tibet [1] - AI technology has been widely applied in various sectors in Tibet, including government affairs, community governance, public services, and ecological research [1] Group 2 - The "e-KeKao" app, which allows for rapid holographic interpretation of research subjects, has gained attention and can upload data in real-time to a cloud database via satellite networks [1] - The Chinese government has provided clearer direction for AI development in Tibet through the issuance of the "Artificial Intelligence+" action plan in August [1] - The "Sunshine Qingyan" V1.0 model was trained on approximately 28.8 billion tokens of high-quality Tibetan language data, covering multiple fields such as news, law, medicine, education, and technology [1] Group 3 - The team led by Tibetan entrepreneur Danzen Rob has been dedicated to the digitization of Tibetan language corpora since 2018, developing multilingual intelligent translation, speech recognition, and OCR systems [2] - The team has constructed nearly 70 million Tibetan-Chinese parallel corpora and a multilingual speech database covering over 140 countries [2] - The Tibetan language model DeepZang, developed by Danzen Rob's team, has achieved high-quality semantic understanding, generation, and interaction across multiple languages, with plans for public release by the end of this year [2]