10. Lost Hong Kong: Can AI Be Its Ticket Back to Prosperity?
Summary
- Hong Kong’s problem is not whether it has AI, but whether it can connect AI to application areas where it still holds an edge. Hong Kong’s overall compute capacity offers no particular advantage. Some want to build large models, but the guests see more viable entry points in finance, retail, and especially entertainment and AIGC. Web3 got an early boost from Hong Kong’s financial strengths but stumbled as the sector cooled; AI cannot simply chase the next hot trend—it must build an ecosystem spanning talent, research, capital, and commercialization.
- Hong Kong’s decline stems from the systematic replacement of its old functions, not simply from an aging urban aesthetic. WTO accession, the removal of foreign-trade quotas, the opening of mainland supply chains, and the rise of internet platforms such as Alibaba shifted the value of deep-water ports, trade intermediaries, and compradors to Shenzhen, Guangzhou, and the industrial belts. The result: “Hong Kong Island is 2024,” while Kowloon’s middle class and neighborhood vitality feel stuck in the 1990s.
- The talent scheme proves Hong Kong can still attract people, but high housing and living costs and limited growth opportunities mean it has succeeded at acquisition and failed at retention. Scarlett believes the first step is already complete: people are willing to apply for Hong Kong status. The harder task is filling in housing, visas, taxes, employment, and industry services so talent actually settles in rather than treating the status as a backup option.
- Hong Kong’s most differentiated path to revival is to combine its surviving entertainment network with AI-native creation. Scarlett asks: “Why did we give Korea those 20 years? Hong Kongers should have been the Koreans.”（为什么我们把这个二十年的时间都让给了韩国人？香港人应该是那个韩国人。）Hong Kong needs to draw back creative talent scattered around the world and attract new international migrants. 卫诗婕 hopes Hong Kong can become “a hybrid of Berlin and Silicon Valley, ideally with Los Angeles added.”
- AI’s value to entertainment is not just lower costs and higher efficiency, but the creation of work that old media could not express. 刘德华’s AI MV, Hong Kong Baptist University’s online concert, BIUBIU 小斑马’s collaboration with 大张伟—which drew more than 100M views—and LISA 404’s human-machine love story with 吴克群 all point to the same paradigm: don’t force AI into an old workflow; use a new medium to reimagine moving images, music, and virtual idols.
- Media and labor markets will both be reshaped, but “AI will make everyone unemployed immediately” is an oversimplification of what is actually changing. Meitu’s Gary says every creator will have “10 more clients” because of AI. 卫诗婕 agrees only halfway: companies may not lay people off directly because of AI, but they will拆解 which tasks belong to people and which to models. Customer service, basic sales, translation, and junior coding will come under pressure sooner, while some individuals will still “fall through the cracks of this era’s transition.”
- AI companions may work; the real open questions are whether they can carry the full weight of human emotion and how society will support a heavily consumer-oriented population. At lunch, Scarlett and the other 2 panelists agreed that people will fall in love with AI, because love for idols is often self-projection without reciprocal feedback. 刘擎 goes further, imagining a future in which 1% of people create and 99% consume; welfare systems, cultural norms, and the assumption that love must point toward carbon-based life would all have to be rewritten.
Deep dive
1. Hong Kong’s Old Prosperity Faded with the Value of Trade Intermediation
Scarlett moved from Wuxi to Hong Kong at 12 and witnessed the golden age of Hong Kong films, TVB, and Cantopop. The episode notes that she later ran Channel V. “On weekends, if we went to see a movie, we always picked a Hong Kong film; no one picked Hollywood.” Today, new directors, musicians, and works are scarce. 《九龙城寨》 revived memories of youth, but felt more like a long-delayed echo.
Her account of the industrial migration runs as follows: after 1949, a wave of factory owners from Jiangsu, Zhejiang, and Shanghai moved to Hong Kong; manufacturing later shifted to Dongguan and Zhejiang, while Hong Kong moved from property into finance. The era of “transitioning from finance to technology” happened to be the period when Hong Kong gradually lost its voice.
WTO accession, the removal of foreign-trade quotas, and the rise of internet platforms displaced Hong Kong’s old entrepôt and trade-intermediary functions, while Alibaba further eroded the value of trade information asymmetry. Diane notes that apparel companies once set up showrooms in Hong Kong to receive English-speaking buyers. Those functions have since moved to Shenzhen and Guangzhou, closer to the upstream supply chain.
2. Hong Kong’s Real Missing Piece Is a Middle Class with Room to Move Forward
Walking through Mong Kok, Yau Ma Tei, and Sham Shui Po, 卫诗婕 felt that “the hour and second hands just stopped there.” Old brands still dominate the streets and malls; coffee shops lack variety and freshness. The street-level character remains, but the energy of young people and new businesses is gone.
Scarlett’s diagnosis is sharper: “You go to Hong Kong Island, and people are clearly living in 2024,” while the middle class remains stuck in the 1990s. The wealthy continue to move forward; Kowloon is home to large populations of Southeast Asian workers and mainland migrants. Hong Kong’s central problem is its inability to offer the middle class new paths and employment opportunities.
Diane argues that Mong Kok and Sham Shui Po should not simply be remade into shinier neighborhoods for the wealthy. They should look more like Berlin’s youth communities or Tokyo’s Shibuya: not necessarily upscale, but open to artists, musicians, and subcultures, with “Shibuya’s own energy.”
卫诗婕’s pushback is worth preserving: Hong Kong is not short of everyday vitality. What is missing is the next generation of content from its old shops, old brands, and “the 蔡澜 era.” Retaining a 1990s texture is part of Hong Kong’s local character; without new companies and young people entering, however, it becomes a sign of stagnation.
3. The Talent Scheme Brought Traffic, but Not Retention
Scarlett traveled to more than 50 mainland cities and saw local investment-attraction systems provide tiered support for residency, housing, taxes, employment, and factory construction. Hong Kong, by contrast, historically followed a “small government, big business” model in which the economy was driven by companies rather than policy. It lacks the full suite of support needed after a startup project arrives.
Invest Hong Kong has begun taking on a more active investment-attraction role. Diane also remembers it holding events in Silicon Valley in 2017 and 2018. Her reservation is that, after companies are recruited, the surrounding infrastructure—talent visas, housing, investment requirements, and other support—remains less mature than what mainland industrial parks offer.
Many of 卫诗婕’s peers have applied for or completed the talent scheme but are unwilling to live in Hong Kong because of housing costs, pressure, and limited business growth. Scarlett describes it as a product funnel: “The first acquisition step works, but the traffic isn’t retained downstream.” Hong Kong still has brand pull; the real problem is retention at steps 2 and 3.
4. Hong Kong’s AI Edge Is Industry Connectivity, Not Model-Building
Diane contrasts Silicon Valley with Hong Kong. After ChatGPT 3.5 launched, people who had left Silicon Valley for Texas and Florida in 2018 and 2019 began returning, while mainland investors reappeared at conferences such as GTC. Stanford, Fairchild, and Silicon Valley’s technology gene had never disappeared. Hong Kong’s stronger underlying gene, she argues, is entertainment.
Scarlett says Hong Kong is “not particularly advantaged” in compute and should not copy the mainland’s large-model frenzy. It should instead go deeper into AI for finance, retail, and entertainment. The question is not whether a product carries an AI label, but whether the technology can “connect with the industries that already exist.”
Web3 benefited from policy support because of Hong Kong’s financial base, and companies such as Animoca attracted attention. But Web3, blockchain, and Bitcoin-related sectors have cooled overall. Scarlett believes the sector may still need time, and that Hong Kong should take from the experience the importance of applications and ecosystem-building.
5. Entertainment and AIGC Are Hong Kong’s Most Distinctive New Cards
Scarlett’s central argument is to reactivate the entertainment gene built up through the eras of 梅艳芳, 张国荣, 刘德华, 张学友, 周星驰, and 王家卫. She asks: “We all watch Korean dramas and films now. Why did we give Korea those 20 years? Hong Kongers should have been the Koreans.”（我们现在都看韩剧、韩国电影，为什么我们把这个二十年的时间都让给了韩国人？香港人应该是那个韩国人。）
The paths presented at Future and Remix included a 刘德华 AI MV, VR/AI concerts, and an online concert from the Professor Poon team at Hong Kong Baptist University. 卫诗婕 was initially surprised by how open Hong Kong artists were to AI, but saw 刘德华 and the stars around him actively looking for ways to bring AI into future creative work.
卫诗婕’s ideal formula for Hong Kong is “a hybrid of Berlin and Silicon Valley, if possible with Los Angeles added”: Berlin for youth culture, Silicon Valley for the technology engine, and Los Angeles for the entertainment industry. Hong Kong’s international character could allow different immigrant communities and creative traditions to collide.
Diane says foreign friends often hear Chinese pop music as “music from the 1980s.” The issue is not only production tools; it is also a relatively pure culture lacking new blood. If Hong Kong can attract global creators and new migrants, AI could help local culture mix again and start moving forward.
6. AI-Native Creation Is Not About Imitating Humans
The key to 刘德华’s AI MV was not simply using AI as an assistant inside a traditional MV workflow. It was acknowledging that AI imagery can be creepy and fall into the uncanny valley, then treating that imperfection as a new language. Diane summarized the shift this way: “The point isn’t to fold AI into your workflow; it’s to use AI to rethink what future creation itself should look like.”（不是把 AI 融入到你的工作流里面，而是你用 AI 整个去想未来的创作是什么样子的。）
Scarlett divides current practice into 2 camps. One uses models to “resurrect” existing stars, as in a 邓丽君 concert. The other creates entirely new idols, such as 洛天依 and 初音未来. She is more interested in the creative potential of the medium itself: creators should use new media to produce effects that oil painting, photography, and live action cannot.
Her team’s BIUBIU 小斑马 collaborated with 大张伟 on an MV that built a complete world from virtual 3D environments, virtual characters, and dancing building blocks; it has drawn more than 100M views. LISA 404’s MV with 吴克群 visualized music through a story about human-machine emotion.
Scarlett has been working on “how to visualize music” since her Channel V days. Using a Beethoven melody as an example, she says live action can flatten the imaginative dimension of music, while an AIGC generator can give abstract sound a visual form unconstrained by physical settings. Dance Diffusion, mentioned by creators at the event, was also seen as a tool for improving precision and detail.
7. Hong Kong’s Entertainment Gene Lives in People and Chemistry, Not Nostalgia
卫诗婕’s challenge is that mainland China and South Korea have built entertainment industries at high speed: is Hong Kong’s so-called entertainment gene still an advantage? Scarlett does not locate the answer in the city brand, but in its people. 梅艳芳 was once called “梅廿八”—28 productions built around her in a single month—with a huge network of lighting, cinematography, screenwriting, wardrobe, and makeup professionals behind her.
Those workers and their apprentices later scattered around the world, but Scarlett believes they can still be brought back together. The gene is the chemistry formed through long-term collaboration between directors such as 徐克 and their regular makeup artists, costume designers, and camera operators. “Maybe 王家卫 says one sentence, and everyone knows exactly what he means”; with a new team, explaining the same idea could take a week.
The collaboration between 王家卫 and Chris Doyle shows why that chemistry cannot be substituted. The texture of 《重庆森林》 and 《花样年华》 came from a particular cinematographic relationship; after changing cinematographers, it became difficult to reproduce the feel of 《花样年华》. Hong Kong’s opportunity is not to remake old films, but to make its old networks, animation, and design capabilities react with AI-native media.
8. Universities Should Underwrite 10-20 Years of Technology Risk
Scarlett compares technology R&D with pharmaceuticals: it may require 10 or 20 years, while a VC fund typically has to close in about 8 years. The time horizons are structurally mismatched. Universities and research institutions should cover frontier exploration; once the technology is ready for commercialization, VC and PE can enter in stages.
Diane uses Silicon Valley’s early development to illustrate the need for a “giant engine.” NASA and the military first absorbed the cost of long-cycle research; only after the results were no longer limited to internal use did they gradually become commercialized. Hong Kong needs the government to place long-cycle R&D within academia, with VC and PE entering at the application layer.
Hong Kong has established a number of laboratories in recent years, while academia has made visible investments in AI, media, and information research. At Future and Remix, participants included 方可成, Hong Kong University PhD students, and their supervisors in addition to Yolanda, showing that the research base is not empty.
卫诗婕 asks whether Hong Kong will fully embrace AI after Web3, and whether the government still needs to pay comprehensive attention to the sector. She says AIGC carries “very large implications” for Hong Kong’s strategic direction.
9. AI Is Rebuilding Media Production and User Operations at the Same Time
Hong Kong Baptist University’s Yolanda repeatedly stressed at the forum that “the old business model of news is collapsing.” She cited News Tracer on Reddit, which can identify breaking news; US newsrooms use its data to guide story assignments. AI is therefore changing not only how media is produced, but also the business model around it.
The discussion covered using AI to verify sources, identify story ideas, detect fake news, and reduce noise. The paradox is that the information explosion and content waste created by generative AI will itself require AI to filter.
生动活泼 demonstrated its internal workflows for finding story ideas and handling post-production, along with Google’s newly launched NotebookLM. Feed it under a couple hundred characters and specify 3 analytical directions, and it can generate an extremely natural English conversation between 2 people. Diane joked, “We could actually lose our jobs too,” because the tool had moved from assisting with a task to completing 2 steps in a single pass.
Ultra turned LISA 404 into a chatbot that answered questions about the festival’s history, tickets, scalpers, toilets, and finding a companion. Diane says AI can do more than customer service: it can help organizers understand user demand, run deeper marketing, and operate the event. Even event businesses that previously had “no user data” now have a new operating entry point.
10. AI’s Startup Opportunity Lies in Data, Organizational Friction, and Niche Demand
卫诗婕 frames the question through the mobile-internet era: will this be a case of “Sina turning into Douyin,” or will Douyin emerge from outside? Diane responds that giants such as Google and Microsoft have the capital and resources but cannot cover every need. Internal politics and inefficiency are precisely where new companies can quietly grow during a paradigm shift.
Diane believes content companies and big tech are, in some respects, starting from the same line. 生动活泼 has its own program data, while big tech may lack it. Why should OpenAI or Google spend “tens of millions” buying data from the content industry? She also notes that some large-model companies have begun discussing future partnerships with content providers.
Diane expects technology to enable greater atomization, with more individuals able to wield it and create value that was previously unimaginable. Value created by individuals could become the smallest unit in the future commercial world and its business models.
11. The Employment Shock Is an Organizational Rewiring, Not a Binary Event
One creator admitted feeling guilty because AI meant he no longer needed his original designer. Gary from Meitu pushed back, saying the designers he encountered had not lost their jobs to AI and arguing that “every creator will have 10 more clients because of AI.”
卫诗婕 agrees “halfway, with reservations.” Many companies have not laid people off directly because of generative AI, and efficiency gains may instead come from larger strategic adjustments. But AI is objectively breaking down existing workflows and reallocating responsibilities between people and models; the organizational reshape is already underway.
Diane believes creators and designers will not be the first to go; basic roles such as customer service will feel the impact first, and sales could eventually “be cut in half.” In some countries, outsourced translation workers may lose their jobs entirely, while Cursor can handle basic code. The calculator did not eliminate mathematics, but it changed which work had to be done by hand. AI may repeat that process.
卫诗婕 cites a Silicon Valley headhunter’s analogy: “AI technology is like the ox used to plow fields: once the Industrial Revolution arrives, it will inevitably be replaced.” At the macro level, society will eventually move to a new division of labor. At the micro level, some people will fall through the cracks of the transition. The question is not whether to deny the losses, but whether those people can pass through the period with less pain.
12. People Will Love AI; Society May Not Be Ready for the Consequences
Before the emotional roundtable, Scarlett, the other 2 panelists, and the host had lunch together. All 3 panelists independently concluded that love between humans and machines “will definitely happen.” Michael used people’s love for cars to show that love can be directed toward nonliving objects. 陈秋帆 argued that machines’ physical sensations will increasingly approach those of humans. Scarlett pointed to the fan-idol relationship: an idol may never respond, yet fans can follow one for decades.
Scarlett asks: “Do you think your love for an idol is love?” If the answer is yes, then love contains a powerful element of self-projection. 卫诗婕 extends the metaphor by comparing intimacy to a mirror: the medium was once limited to the opposite sex, then expanded to the same sex, and may in the future move from carbon-based life to silicon-based life.
Character.AI is already handling role-play and companionship. But Diane says friends who chat with such systems over the long term may not be willing to pay. Users of otome games, by contrast, continue spending and may even buy building-sized advertisements. The commercial question is not whether virtual relationships can satisfy emotion, but whether they can satisfy it “100%,” and what level of interaction is enough to convert into payment.
刘擎 offers a more radical social vision: perhaps 1% of people will create in the future while 99% become consumers, with emotions also supplied mainly by AI characters. Whether welfare systems can support the majority and whether cultural norms will be rewritten are much harder questions than whether people will fall in love with AI. Scarlett believes the AI era will break people’s dependence on existing cultural norms and force them to accept the world through a new worldview.