About the “AI Chatbot Based on Xi Jinping” Story

This week, various English-language newspapers reported that China is launching its very own Xi Jinping AI chatbot. China’s top internet regulator is reportedly planning to unveil a new chatbot trained on the political philosophy of Xi Jinping. This Large Language Model (LLM) is humorously referred to as ‘Chat Xi PT’ by the Financial Times and in other foreign media reports.

The Times of India website headlined that “China builds AI chatbot trained on Xi Jinping’s thoughts.” News site Asia Financial reported that “China has unveiled a chatbot trained to think like President Xi Jinping.” Various outlets even called it a “ChatGPT chatbot based on Xi Jinping.”

The Financial Times calls the application “China’s latest answer to OpenAI” and notes that “Beijing’s latest attempt to control how artificial intelligence informs Chinese internet users has been rolled out as a chatbot trained on the thoughts of President Xi Jinping.”

Besides the Financial Times article by Ryan McMorrow, media reports were largely based on a piece in the South China Morning Post authored by Sylvie Zhuang, titled “China rolls out large language model AI based on Xi Jinping Thought.”

Zhuang detailed how Xi Jinping’s political philosophy, along with other themes aligned with the official government narrative, form the core content of the chatbot, which is launched at a time when China “tries to use artificial intelligence to drive economic growth while maintaining strict regulatory control over cybersecurity.”

News about the supposed “Xi Jinping chatbot” is based on a post published on the WeChat account of the Cyberspace Administration magazine.

The magazine in question is China Cyberspace (中国网信), overseen by the Cyberspace Administration of China (国家互联网信息办公室) and published by the China Cyberspace Research Institute (中国网络空间研究院).

“Cyberspace Information Research Large Model Application”

On May 20, China Cyberspace (中国网信杂志) posted the following text on WeChat, which was viewed less than 6000 times within two days (translation by What’s on Weibo):

“Recently, the Cyberspace Information Research Large [Language] Model Application developed by the China Cyberspace Research Institute has been officially launched and is being tried out internally.” ^[1]

“As an authoritative and high-end think tank in the Cybersecurity and Informatization field, the China Cyberspace Research Institute relied on the data of the “Internet Information Research Database” and organized a special tech team to independently develop the Cyberspace Information Research Large Model Application, to take the lead in demonstrating the innovative development and application of generative AI technology in the field of Cybersecurity and Informatization.”

“The corpus of this Large Model [LLM] is sourced from seven major speciality knowledge bases within the “Internet Information Research Database,” including the “Comprehensive Database of Cyber Information Knowledge”, the “Knowledge Base of Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era,” “Dynamic Cyber Knowledge Base,” “Internet Information Journal Knowledge Base,” “Internet Information Report Knowledge Base,” and more. Users can independently select different categories of knowledge bases for smart question-and-answering. The specialization and authority of the corpus ensures the professionalism of the content that’s generated.”

“Do you want to quickly make a summarized report on the current status of AI development? Are you curious about the differences between ‘new quality productive forces’ and ‘traditional productivity’? This Large Model application can quickly produce it for you!”

“The Cyberspace Information Research Large Model Application is based on domestically registered open-source and commercially available pre-trained language models. By combining Information Retrieval technology with specialized Cyberspace Information knowledge, it can do smart question-and-answering [Q&A chatbot], it can generate articles, give summaries, do Chinese-English translations, and many other kinds of tasks in the field of Cybersecurity and Informatization to meet the various demands of users.”

“The system used for the Cyberspace Information Research Large Model Application is deployed on a dedicated local server of the China Cyberspace Research Institute. Data is processed from this local server, ensuring high security. This application will become one of the embedded functions of the “Internet Information Research Database” and authorized users invited for targeted testing can access and use it.”

“The Cyberspace Information Research Large Model Application will also support users to customize and build new knowledge bases. Users uploading public data and personal documents can analyze and infer, further expanding the scope of personalised use by users.”

Although some Chinese media sources reported on the launch of the application, it barely received traction on Chinese social media.

At the time of writing, the only official accounts posting about the application on Chinese social media are those related to research institutions or the Cyberspace Administration of China.

Key Takeaways on the “Chat Xi PT” Application:

So what are the key takeaways about the so-called, supposed ‘Chat Xi PT’ application that various foreign media have been writing about?

■ Focus on Cyberspace Administration and Digital Governance:
Contrary to some English-language media reports, the application is not primarily centered around Xi Jinping Thought but rather emphasizes Cyberspace Administration and digital governance. Its official name, the “Cyberspace Information Research Large Model Application” (网信研究大模型应用), does not even mention Xi Jinping.

■ Not a Rival to OpenAI’s ChatGPT:
Unlike what has been suggested in the media reports, this particular application should not be seen in the light of China “creating rivals to the likes of Open AI’s ChatGPT” (FT). Instead, it caters to a specific group of users engaged in specialized research or operating within certain knowledge fields. There are many others (commercial) chatbots in China that could be seen as Chinese alternatives to OpenAI’s ChatGPT. This is not one of them.

■ Modernization of Cyberspace Authorities:
Rather than solely meeting user demand, the application underscores China’s Cyberspace authorities’ modernization efforts by integrating generative AI technology into their own platforms.

■ Clarifying “Xi Jinping Thought”:
Various English-language media reports conflate “Xi Jinping Thought” with “thoughts of Xi Jinping.” “Xi Jinping Thought” specifically refers to “Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era,” the theories, body of ideas that were incorporated into the Constitution of the Chinese Communist Party in 2017.

■ Nothing “New” about the Application:
The ‘Cyberspace Information Research Large Model Application’ is based on existing LLMs and functions as a tool for navigating databases and information in the AI era, rather than representing a groundbreaking innovation or an actual ‘Xi Jinping chatbot.’ While it may have been written as a tongue-in-cheek headline, let’s be clear: there is no such thing as a Chat Xi PT.

By Manya Koetse

^[1]About the translation of the term “网信” (wǎngxìn): in this text, I’ve used different translations for the term “网信” (wǎngxìn) depending on the context of its use. The term can be translated into English as “cyberspace” or “internet information,” but since it is mostly used in relation to China’s Cyberspace Administration and digital governance, it is sometimes more appropriate to refer to it as Cyberspace Security and Information，like the term “国家网信部门” which translates to “national cybersecurity and informatization department” (Also see translations by DigiChina).

Attention!

For 11 years, What’s on Weibo has remained a 100% independent blog, fueled by my passion to write about China’s digital culture and online trends. Over a year ago, we introduced a soft paywall to ensure the sustainability of this platform. I’m grateful to all our loyal readers who’ve subscribed since 2022. Your support has been invaluable. But we need more subscribers to continue our work. If you appreciate our content and want to support independent China reporting, please consider becoming a subscriber. Your support keeps What’s on Weibo going strong!

Full Text by China Cyberspace:

“近日，由中国网络空间研究院开发的网信研究大模型应用已正式上线并内部试用。

垂直专业：聚焦网信领域

作为网信领域权威高端智库，中国网络空间研究院依托“网信研究数据库”数据，组织专门技术团队，自主开发了网信研究大模型应用，率先示范生成式人工智能技术在网信研究领域的创新发展和落地应用。

该大模型语料库来源于“网信研究数据库”的七大网信专业知识库，包括“网信知识总库”“习近平新时代中国特色社会主义思想知识库”“网信动态知识库”“网信期刊知识库”“网信报告知识库”等。用户可自主选择不同类别的知识库进行智能问答。语料库的专业性、权威性保证了生成内容的专业性。

便捷高效：实现多种功能

想快速列出关于人工智能发展现状的报告提纲？想知道新质生产力和传统生产力的不同之处？网信研究大模型应用能够迅速生成！

网信研究大模型应用基于已备案的国内开源可商用预训练语言模型，通过将检索增强生成技术和网信专业知识相结合，实现了网信领域的智能问答、文稿生成、概括总结、中英互译等多种功能，可满足用户的多种需求。

安全可靠：数据本地处理

网信研究大模型应用系统部署在中国网络空间研究院的专属本地服务器，数据由本地服务器进行处理，具有较高的安全性。该大模型应用将成为“网信研究数据库”的嵌入功能之一，获得授权的定点测试用户可以应邀使用该应用。

网信研究大模型应用还将支持用户自定义新建知识库，可通过加载用户自己上传的公开数据、个人文档进行分析推理，进一步拓展用户的个性化使用范围。”

Featured image by What’s on Weibo, image of Xi Jinping under Wikimedia Commons.

Follow @whatsonweibo

Spotted a mistake or want to add something? Please let us know in comments below or email us. First-time commenters, please be patient – we will have to manually approve your comment before it appears.

The post About the “AI Chatbot Based on Xi Jinping” Story appeared first on What's on Weibo.