_China_Chatbot_11
What happens when Chinese AI meets Chinese media
Hello, and welcome to another issue of China Chatbot, where I take a look at the latest attempt by CCP elites to influence international AI governance, examine an industry alliance that all of China’s heavy-hitting AI companies are part of, and do a reality check for a new AI tool much-puffed by state media.
Enjoy!
Alex Colville (Researcher, China Media Project)
_IN_OUR_FEEDS(3):The Summit of Their Ambitions
From November 19-22, major industry leaders and government officials gathered in Wuzhen for the World Internet Conference, the “highest-level” event in China’s digital industry. The summit focused on global AI governance and developing the technology as a productive force. Xi Jinping showed his favor by congratulating the summit, while Politburo Standing Committee member Ding Xuexiang (丁薛祥) gave a keynote speech calling for greater international cooperation on AI innovation and governance. Elsewhere, Alibaba Group CEO Eddie Wu Yongming said his company was ready to deploy their base model Qwen for China’s developers to use for applications, building China’s AI productivity. The summit saw the launch of a “professional committee” of over 100 experts and professionals from around the world. According to the People’s Daily, the committee will “build international consensus” on AI. It will create promotion plans for AI standards, security and governance, and industry development. Global Times hailed the conference’s attempts to build international cooperation on AI, writing that Western actions like US chip sanctions on China risked extending an “unjust international order” that keeps rich Western nations on top.
Pushing Back Against Push Notifications
As part of their “Clean and Bright” (清朗) movement to reform the internet, the Cyberspace Administration of China (CAC) has launched an inspection campaign to overhaul internet platform algorithms. The readout from the CAC says that, until mid-February next year, the agency will combat social problems caused by algorithms. These include addiction and “information cocoons” created by recommendation algorithms pushing similar content to users based on their preferences, as well as algorithms that “compress delivery time” for delivery drivers and risk accidents by making them rush to meet impossible deadlines. The CAC said users must be given greater control over the content they are shown, urging recommendation algorithms to help identify online rumors and “actively spread positive energy.” This could be bad news for Douyin, whose “secret sauce” algorithm has hooked millions of users to the platform.
Sex and the City Councillors
Several members of Hong Kong’s legislature have been targeted by an unknown blackmailer using AI to graft their faces onto an image of a couple engaged in sexual intercourse. Legislators were sent emails from an individual claiming to be a private investigator — emails that included a screenshot of what they claimed was a compromising video of the lawmaker and a demand for money to prevent the video from going public. One target told reporters that he believed AI deepfake blackmail cases are common in Hong Kong, but originate from outside the city. This case is similar to a deepfake scandal in South Korea from September, when sexually explicit fake images of women were shared on private WhatsApp university groups.
TL;DR: Party elites continue their drive to harness AI for domestic productivity and international influence, now merging it into older narratives of how China is the champion of the Global South vs. a US-led exploitative Western bloc. Elsewhere AI continues to be a disruptive force — some parts (like big tech algorithms) the authorities can control. Others (like our HK blackmailer) elude them
_EXPLAINER:
China AI Industry Development Alliance (中国人工智能产业发展联盟)
Which is?
AIIA for short. An industry forum assembled under the CCP of the best people and companies in the private sector working in Chinese AI, bent on pushing it to the cutting edge.
Who’s in it?
It’s actually hard to find someone who isn’t in it. All the big fish — Huawei, Baidu, Alibaba, Tencent, Qihoo 360, iFLYTEK, Kuaishou, et al — are in the Party’s posse. AIIA boasts over 1,064 member companies, and the leader of their expert committee is Gao Wen (高文), who led the Politburo’s only study session on AI in 2018.
Huh, they been around long?
No. It was set up in 2017 by a plethora of important government bodies, including the National Reform and Development Commission (NRDC), the Ministry of Science and Technology, and the Cyberspace Administration of China. It came a few months after the State Council issued a plan for China to become a world leader in AI innovation and development by 2030.
So how is this alliance going to bring that about?
Good question. In his speech at the opening of the alliance, the then-deputy head of the NRDC said the Alliance must:
Fully support government policy decisions
Promote innovation and development
Collaborate and share
Strive to create a “brand effect” (品牌效应) — I’m interpreting this as making AIIA a known, respected brand that new companies want to be part of.
So have we been seeing much of this?
Well, according to their own press release, in 2024 AIIA helped implement government policy, platformed “various AIIA tasks as planned,” held 77 meetings and seminars, and “connected over 230 companies” on the supply and demand side.
Any examples of these?
Most seem to be focused on AI safety, ethics, and governance. Their expert committees have been trying to solve key problems like a Chinese-language data shortage. They’ve released a few industry standards, including an “AI Risk Management System” to forestall and prepare for AI calamities. They’ve created a think-tank network, an LLM benchmark system, datasets for companies to test models on, and lists of Chinese AI success stories.
Some of these just sound like people trying to look busy. Why are you wasting my time with this?
There are certainly questions about this alliance, such as how much use it would be for well-equipped industry leaders like Huawei, or to what extent they will “collaborate and share” with their competitors. But it’s clear the Party is encouraging private enterprise (which the leadership sees as the dynamo for Chinese AI innovation) to unite under AIIA, using it as a centralized forum to coordinate and achieve consensus on best practices. Perhaps, like TC260, the AIIA will be a source of new AI standards going forward.
_ONE_PROMPT_PROMPT:With so much AI hype around, it goes without saying you should take a company’s reports on their new tech with a pinch of salt. Make that two, if it ends up being featured in PRC state media.
On November 18, both Xinhua and China Daily pushed (in English) the successes of Chinese-owned AIGC platform Vidu. The new version, Vidu-1.5, generates videos that merge three or more images. “Look how consistent is that suit,” one admiring “AI product aficionado” on X is quoted as saying about a Vidu video depicting a man in a mech suit crossing an American street. Vidu’s CEO told Xinhua this was the company’s “greatest breakthrough,” painting it as one-in-the-eye for the consistency problems plaguing many AIGC models.
It’s true that neither Western nor other Chinese open-access AI generators are able to create videos from multiple images. Here’s one I created merging stock photos of a man with London’s Trafalgar Square and Chinese traditional clothing:
The man, the clothing, and the background are recognizably the ones from the pictures I uploaded. However, Vidu confused some of the details in the generation process. The clothing, for example, isn’t very consistent with the image (or at least, not consistent enough to warrant a tweet about it):
When I generated a second video with the same images, I got a very different result. Trafalgar Square now has some grand East Asian-style buildings I certainly don’t remember from my last visit.
So although this tool is definitely a new one for mass audiences, it isn’t quite there yet. China’s national news agency is not in the habit of cross-referencing a company’s self-promotion when it aligns with what the Party wants foreigners to hear. But hallucinations and inconsistency remain a problem.