How China is embedding AI everywhere (even in nightclubs)

The world's second-biggest economy is embracing AI at breakneck speed – driven by state support, local models, and a hunger to scale up fast and beat the United States.

Jessica Sier
North Asia correspondent
Jun 23, 2025 – 1.41pm

Under the kaleidoscopic lights of INS Park, one of Shanghai's largest nightclubs, the dance floor pulses to more than just music – it responds to algorithms.

Beams of light slice across grass growing out of the walls, and screens show images of rollicking ocean waves that morph into an explosion of fast-growing mushrooms. Bodies writhe and jump under the flashing lights.

At the DJ booth, 28-year-old Chenxiao "Race" Li oversees a small cluster of flashing monitors. His start-up, Magipop, runs an AI system that absorbs the music inputs in real time and controls the club's visuals and lighting. It adjusts colours, patterns, and intensity based on beat, tempo, and even audience noise.

Li's application is both an example of China's race to embed artificial intelligence into every layer of daily life, and a neat microcosm of the country's co-ordinated campaign to supercharge productivity and economic growth by ceding routine processes to machine-taught models.

"The AI controls the structure of the show," Li yells over the thumping music. "We make the art in between. You don't need to do so much work to get the timing of everything right. You can just concentrate on making the music."

From entertainment to city infrastructure, mining, energy and telecommunications, the world's second-biggest economy is harnessing the power of superfast computing power and machine learning to automate decision-making and drive the country's next phase of economic transformation.

The national poster child of this revolution is DeepSeek-R1, a cheap, open-source large language model that earlier this year burst onto the global scene and demonstrated China's ability to compete with American technology. At its core, DeepSeek's engineering proved that data could be processed quickly with less computing power.

Within days of its launch, DeepSeek's technology was being embedded across AI models powered by China tech giants such as Alibaba, Tencent and ByteDance.

While the West has focused heavily on foundational breakthroughs such as OpenAI's ChatGPT and Anthropic's Claude, China is pushing ahead on wide-scale deployment.

This means getting AI into the hands of users quickly. Instead of waiting for cutting-edge game-changers, Chinese companies and government bodies are rapidly embedding existing models into real-world applications.

More than 3700 generative AI tools are now registered with China's cybersecurity regulator. More than 1100 of the tools offer APIs – mechanisms that enable two software components to communicate with each other – for commercial use.

This means businesses across China, from e-commerce platforms to manufacturing firms, can plug these tools directly into their own systems to automate tasks such as customer service, product recommendations, document translation and logistics planning.

An online retailer, for example, can integrate a generative model via API to run a 24/7 AI-powered chatbot that handles thousands of customer queries daily, without needing its own data science team or computing infrastructure.

Ivy Wang, Canva's China manager: "Chinese people are already so familiar with using AI as a companion tool in their lives." Sanghee Liu

"In China, the efficiency of a centralised government further helps the execution of big data and AI policies," says Curt Shi, founder of Imprint Capital Partners, a venture investment firm based in China and Australia.

"That's why in the last couple of years, we see a number of AI unicorns coming out of China. In general, Chinese entrepreneurs are good at business model innovations and AI is a large leverage."

'City Brain'
Magipop, with its reinvention of digital performance art, is typical of the new wave of nimble start-ups. But the beauty of China's often open-source model is that tiny enterprises like these are flourishing alongside mega-projects such as Hangzhou's "City Brain", an AI platform managing traffic flows and emergency responses in real time.

Hangzhou, the bustling capital of Zhejiang province, is widely seen as the testbed for China's most ambitious smart-city experiment.

Developed with tech giant Alibaba, the system collects billions of data points from traffic cameras, "Internet of Things" sensors, police reports, toll stations and municipal records, all in real time. The goal is to centralise and algorithmically manage the city's operations – from traffic flows and emergency services responses to public services – with a level of efficiency that only AI can provide.

The City Brain was originally rolled out to combat congestion, but has since helped transform Hangzhou's gridlocked roads.

In 2020, the provincial capital south-west of Shanghai was the second-most congested city in China. Within a year of AI-powered traffic light optimisation and policing decisions, it had dropped to 34th. The system didn't just ease traffic – it slashed accident rates and accelerated emergency vehicle response times.

During the COVID-19 pandemic, the platform was repurposed to manage quarantines and health code enforcement.

Few people would argue against such efficiency gains. In this same vein, however, AI applications are also at the heart of China's notorious surveillance state.

Beijing's "Sharp Eyes" program has stitched together CCTV footage from public and private sources into a nationwide network that now covers 95 per cent of rural China.

Although most developed countries monitor public movements in large cities and other population centres, China has been accused of using advanced surveillance technology to hunt down "state enemies" and profile others who Beijing might suspect are acting against it.

In Xinjiang in north-western China, facial recognition systems identify ethnic Uighurs with racial profiling and flag "suspicious" behaviour, triggering automatic police alerts.

Emotion recognition, voiceprint tracking, and social credit algorithms further deepen Communist Party control over the population at large.

"AI tools help the state to look for deviation from 'normal' behaviour," says Valentin Weber, a researcher at the National Endowment for Democracy.

"While some forms of AI-powered surveillance, such as traffic monitoring systems, can be benign tools for improving governance, the use of these tools to track people directly impacts civilian rights to privacy, freedom of assembly, and freedom of movement."

State-owned laggards
China's AI push is not confined to just flashy consumer tech, smart cities and citizen surveillance. State-owned enterprises, traditionally seen as tech laggards, are also entering the race. For example, power giant State Grid is developing a model to dispatch electricity across the national grid.

PetroChina's 33-billion-parameter model – where the more parameters, or tokens, the more responsive the artificial neural network is to nuance and detail – is another example of big AI. Known as "Kunlun", the energy giant's system aims to optimise exploration and extraction in oil and gas.

Telecom behemoth China Mobile and others are integrating Deep Seek-derived AI across their huge infrastructure to process billions of data points, from customer service to network optimisation. The result of all of this is a nationwide AI acceleration unlike anything seen elsewhere.

In Hangzhou, a smart system developed in partnership with tech giant Alibaba collects billions of data points from traffic cameras. Bloomberg

While the US concentrates on foundational model supremacy, China is betting on scale, speed and sector-by-sector integration to rewire its economy with intelligent systems.

"We're seeing a scrum in foundational technologies by Chinese companies," says Kendra Schaefer, tech policy analyst at Trivium China, a Beijing-based research firm. "And they're not focusing on one sector; they're going across industries, making these models available to anybody anywhere [who] can use them to build things."

Canva's China bet
Australian design giant Canva, for example, has registered five AI models with Chinese authorities. Ivy Wang, Canva's country manager for China and a member of the local founding team, says AI is a natural sell to Chinese users.

"Chinese people are already so familiar with using AI as a companion tool in their lives," Wang tells The Australian Financial Review in the company's colourful office in Beijing's bustling Chaoyang District.

One of the most ambitious applications is in font generation. Chinese characters are notoriously complex. There are more than 20,000, and licensing even a single commercial font can cost upwards of 100,000 yuan ($21,322) annually.

"We have two font designers who manually create about 800 characters," Wang explains. "Then our in-house AI expands those into full character sets. The output isn't generic, we start with something that has heart and blood, then the machine fills in the rest."

In Canva's marketing division, AI is used to mimic the viral structure of China's short video ecosystem. The team analysed the top-performing miniseries on Douyin (the TikTok of China), identifying a formula where a plot twist or dramatic event occurred every three to five seconds.

"We then used AI to generate marketing videos using that same rhythm," says Wang. "It's not just about style, it's about matching the pattern that Chinese users respond to."

Like Magipop, Canva hasn't built its own large language models. Instead, the team has blended local LLMs such as DeepSeek and Alibaba's Quen to power text-to-image tools that generate backdrops, illustrations and layouts inside the Canva design tool. These models, Wang says, are good enough to rival offerings from Google and OpenAI, especially when customised for Chinese aesthetics.

"People here are used to AI, and they're quick to adopt new tools. There's not much friction," she says.

Still, there are limits. Strict government regulations require LLMs to run on Chinese cloud infrastructure and keep user data onshore, meaning Canva had to build and maintain a separate government-licensed platform in China. And there's another major constraint: computing power.

US constraints
Behind the explosive growth of China's AI sector lies a critical bottleneck: access to high-performance chips. Since 2022, the United States has tightened export controls on advanced semiconductors, specifically targeting graphics processing units made by Nvidia, which are essential to train and run LLMs.

The restrictions have forced Chinese AI companies to scramble for alternatives. Some, like Magipop, managed to purchase GPU clusters before the ban kicked in.

"We bought a lot just in time," says Li, whose nightclub visual AI depends on fast computing power. Other firms haven't been so lucky. Start-ups and university labs often rely on lower-grade chips or turn to domestic alternatives from Huawei. These chips, although improving rapidly, still lag those of Nvidia in raw power and software ecosystem compatibility.

"It's like driving a sports car versus a sedan," one AI engineer at a Beijing gaming company start-up says. "They'll get you there, but not in the same time."

The semiconductor chokehold is more than a technical hurdle; it's a strategic move in a broader tech war.

Washington views AI dominance as critical to national security and is seeking to "fence in" China's capabilities without cutting off trade entirely. By restricting the tools needed to train foundation models, the US hopes to blunt China's rise in advanced AI, especially in dual-use areas such as military surveillance and cybersecurity.

In response, Beijing is pouring billions into semiconductor self-reliance. The government has prioritised chip research in its national policy frameworks, offering subsidies to local fabrication plants and accelerating investment into firms such as SMIC, Cambricon and Biren Technology.

Huawei's latest chip breakthroughs, manufactured under the radar using older but still effective lithography nodes, have raised eyebrows in Washington.

Indeed, Nvidia chief executive Jensen Huang recently conceded that Chinese rivals were catching up fast. "China is right behind us," Huang said. "We are very close. Huawei is one of the most formidable technology companies in the world."

Back at INS Park, the crowd surges as another AI-generated lightscape washes over the room – neon vines crawl up the walls before shattering into digital fireworks. For Li, this isn't just a spectacle; it's a glimpse into a future arriving fast.

"Part of Chinese culture is to be more and more efficient," Li says. "It's one of the reasons why AI is taking off everywhere."

The country might lack access to America's most advanced chips, but it has embraced AI, unleashing a wave of adoption across industries and into everyday life that appears to be rolling faster than anywhere else in the world.

"In China, things move quickly when the government wants them to," Li says.

https://www.afr.com/world/asia/how-c...0250603-p5m4fu


di sana gak ada yg menggunakan AI untuk tanya soal Tiananmen

kalau di sini pemerintahnya masih fokus menghilangkan sejarah pemerkosaan massal dari buku sejarah