Stay up up to now with free updates
Simply register for Artificial intelligence myFT Digest – delivered straight to your inbox.
Beijing's latest attempt to regulate the way in which artificial intelligence informs Chinese web users has are available in the shape of a chatbot trained to take heed to the thoughts of President Xi Jinping.
The country's latest large-language model relies on the political philosophy of its leader, referred to as “Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era,” in addition to other official literature from the Chinese Cyberspace Administration.
“The expertise and authority of the corpus ensures the professionalism of the content generated,” said a social media post by CAC Magazine on Monday in regards to the recent LLM.
The effort to make sure AI understands Xi's philosophy comes at a time when Chinese officials try to balance the country's draconian controls on free speech with promoting AI development and creating competing products like Open AI's ChatGPT.
Currently, the brand new model is getting used at a research center overseen by the powerful web regulator, but it surely could eventually be released for wider use, in keeping with an individual involved within the project. The recent model can answer questions, produce reports, summarize information and translate between Chinese and English, the post said.
The creation of the LLM followed extensive efforts by Chinese officials to disseminate Xi's ideas on politics, economics and culture in quite a lot of formats.
More than a dozen books have been published in Xi's name, and his bestsellers normally take center stage at book fairs within the country. Popular news apps from corporations similar to Tencent or NetEase reserve spots at the highest of users' feeds for articles from official media, which mostly feature Xi.
Officials have also required schoolchildren as young as 10 to check his political philosophy. They developed the “Study Xi Strong Nation” app to show and test the knowledge of the country's roughly 100 million party members. In 2018, his ideas were incorporated into the state structure.
CAC, which has taken a lead in setting rules for generative AI and introduced a licensing system, requires that generative AI providers “embody fundamental socialist values” and that generated content must not contain content “that undermines state power.” Companies are liable for their AI outputs.
This poses a selected challenge for model developers because there are relatively few Chinese language datasets available for training their LLMs. Most groups also train with English language information, raising the chance that generative AI may produce responses that violate Chinese language norms.
Tech giants like Baidu and Alibaba have ensured that their models tightly control generated content related to Xi or other potentially sensitive topics. Both groups' generative AI chatbots typically ask users to restart the chat when asked about sensitive topics.
To help developers tackle the issue, the Cyber Security Association of China, a nonprofit organization affiliated with the CAC, in December released the primary public database of 100 million entries of “high-quality and trustworthy data” that groups can use for model training. The training set relies heavily on government regulations and policy documents, state media reports and other official publications, in keeping with excerpts reviewed by the Financial Times.
One of the handfuls of text documents in the information package accommodates 86,314 mentions of Xi Jinping. “Let us unite more closely across the Party Central Committee, with Comrade Xi Jinping because the core,” reads one line.
We must “be sure that our pondering, policies and actions are at all times in close alignment with the Party Central Committee, with General Secretary Xi Jinping at the middle,” says one other article.
