China develops AI model trained on 'Xi Jinping Thought'
A research institute at the Cyberspace Administration of China, the country's internet regulator, has developed a 'safe and reliable' large-scale language model (LLM) based on the political philosophy of Chinese President Xi Jinping.
China's latest answer to OpenAI is 'Chat Xi PT'
Sure, why not: China built a chatbot based on Xi Jinping
https://www.engadget.com/sure-why-not-china-built-a-chatbot-based-on-xi-jinping-155828456.html
China rolls out large language model AI based on Xi Jinping Thought | South China Morning Post
https://www.scmp.com/news/china/politics/article/3263530/china-rolls-out-large-language-model-based-xi-jinping-thought
ChatGPT, Meet XiGPT: China Unveils New AI Model That Speaks Xi Jinping's Thoughts, Spouts Official Party Line - Benzinga
https://www.benzinga.com/markets/asia/24/05/38956682/chatgpt-meet-xigpt-china-unveils-new-ai-model-that-speaks-xi-jinpings-thoughts-spouts-official-party
The Cyberspace Administration of China requires that generative AI 'embody the core values of socialism' and 'contain no content that subverts state power,' so major Western AI models such as ChatGPT and Google Gemini cannot be used. Chinese companies that offer generative AI services, such as Alibaba, must strictly control the content generated, and have introduced mechanisms to stop chats when asked about sensitive topics.
In this context, the China Cyberspace Institute has trained a new LLM as a step towards catching up with Western generative AI technology.
According to a government post on the Chinese social networking site WeChat, the model was trained on texts that align with the government's official views, including more than a dozen books allegedly written by President Xi himself, and embodies an ideology called ' Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era .' The post stated that 'the expertise and authority of the corpus will ensure the professional quality of the generated content.'
The model can answer questions, create report summaries, summarize information, translate between Chinese and English, and meet the needs of a wide range of users. It is built on a carefully selected knowledge base and is closed source, making it 'safe and reliable,' the government said.
The WeChat post also included a demo, suggesting that the AI's answers were generated from certain sources, including official Chinese documents and news reports.
At the time of writing, the 'safe and reliable' LLM has only been deployed on the servers of the China Cyberspace Institute, and internal testing is being conducted by invited users only.
Related Posts:
in Software, Posted by log1d_ts