Oasis Capital was invited to participate in Yicai's "AI Night Talk"

During the 2023 World Artificial Intelligence Conference (WAIC), Oasis Capital was honored to be invited to Yicai's Brainstorm AI Night Talk: Exploring New Opportunities in China's AI Industry. The livestream tackled the hottest topics in the current global wave of artificial intelligence technology, discussing development opportunities and ethical considerations with enterprises, experts, scholars, and investors. Fellow panelists included Xiaodong He, VP of JD.com and head of JD.com's research institute; Xiaqing Huang, founder, chairman, and CEO of CloudMinds; Renhan Li, member of the National New Generation AI Governance Committee and professor at Shanghai Jiao Tong University; Jian Ma, co-founder and CEO of XtalPi; Li Tian, VP of Yingfan Technology and general manager of Yingcai Zhiyun; and Haitao Song, executive director of the Shanghai AI Research Institute.

Over two hours, the guests shared diverse perspectives and engaged in spirited debate. We've selected highlights to share. Enjoy.

WAIC Impressions

Xiaqing Huang: Robotics and AI have finally converged. With large model support, whether for robots or digital humans, no one doubts anymore whether they can fulfill our expectations for intelligent robots.

Li Tian: Beyond the proliferation of foundation models, we're seeing large models begin to develop vertically. There will likely be specialized domain-specific models. The emergence of domain large models reveals new possibilities.

Renhan Li: This conference gave me significant insight: going forward, we need to create new value on top of large models. Large models aren't the end goal. Embodied intelligence and human-machine integration may arrive sooner than expected.

Jinjian Zhang: When we talk about AI and large models, it feels like impossibly advanced technology, but it's not. China has the world's most complete supply chain foundation. Every industry is exploring how to genuinely integrate this technology — whether in robotics, embodied intelligence, or creative fields like art, fiction, and entertainment. Fundamentally, it's all about technology landing in industries.

China's AI Advantages

Xiaodong He: China's greatest advantage is its highly complete industrial supply chain and society's high openness to open technologies — this lays the foundation. AI technology requires not just algorithms but strong data-driven development, and data only comes from scenarios and applications. When users enthusiastically embrace new technology, application barriers drop and data proliferates.

Xiaqing Huang: China's biggest strengths are industrialization capability and a massive future market. China ranks in the world's top tier for AI, first globally for 5G, first for manufacturing capability, and unquestionably first for population and service scenarios. These advantages compound: China can pioneer the AI robotics era, and this lead can drive both our market and global markets.

Jian Ma: China has enormous advantages on both demand and supply sides. This wave of AI empowering thousands of industries is essentially building new infrastructure, so demand is robust. On the supply side, China has significant quantitative advantages in talent reserves.

Li Tian: China's vast industrial chain foundation creates greater room for application. Our globally leading mobile internet, including digital infrastructure development, has generated massive data. Gen Z has natural personal-level acceptance of AI applications. And national strategic emerging industry policies provide crucial support for sustained AI development.

Haitao Song: The core advantage starts with population scale. China has the world's largest mobile terminal user base and usage across the entire PC market. The governance data, industrial data, and lifestyle data generated daily leads globally. This provides core raw material for developing both foundation models and industry vertical models. Second, domestic universities are continuously launching AI programs, supplying researchers and technical professionals. Third, there's strong enthusiasm across all of society and industry for embracing technology.

Challenges and Bottlenecks

Li Tian: ① Application scenarios ② Domain data.

Currently, research far outpaces real-world application. How to effectively translate R&D into enterprise and commercial use to generate business value is the most critical challenge. Opportunity and challenge coexist. We haven't done enough in truly improving efficiency, creating new value, and driving innovation.

Second, domain data. Take professional financial and economic fields — the expertise required is high. Market movements shift instantaneously, with complex underlying causes that require professional analysts to uncover truth. This isn't something current large models can directly answer. Accumulating domain data is key to whether large models can land in vertical industries.

These two problems are interconnected.

Xiaqing Huang: ① Chips ② Talent.

The lack of proprietary AI chips limits the effectiveness of large model training. We need computing chips that generate processing power.

Going forward, we face how to use large models, how to effectively clean and precisely tune data. We must cultivate talent specialized in large model research and fine-tuning, even prompt engineering research. Theoretically, everyone will need to learn prompting.

Xiaodong He: ① Talent ② Innovation ③ Data

Talent quality is key — we can't simply cover this with quantity.

Much domestic innovation is incremental; original, risk-taking innovation remains scarce. Innovation pursued with persistence, even conviction, often yields revolutionary breakthroughs. This is the spirit we need to cultivate.

How to improve quality and build a data ecosystem is also a challenge.

Haitao Song: ① Internationalization/global expansion ② Cross-disciplinary integration

First, internationalization and going global. AI technology reshapes entire industries. Many traditional and specialized industries can benefit from AI. In this global AI wave, many Chinese enterprises and research institutions are at the forefront, but we need broader global users and data. How to internationalize and go global is something the entire AI industry must solve and break through. This involves not just technology, but culture, international infrastructure, and international discourse power — factors that will long-term affect national competitiveness in AI.

Second, cross-disciplinary integration. AI isn't a single industry that empowers all sectors. This process involves cross-industry, cross-technology, cross-talent integration, including data interoperability.

I believe these two points urgently need breakthroughs.

Renhan Li: ① Governance ② Philosophy — industry and technology

For China to lead in AI, three issues matter: safety, human resources, and mutation.

When AI is everywhere, how do industry and technology integrate? How is talent deployed? Solve these and our problems are solved.

Jinjian Zhang: Governance

The core of governance isn't "control" but "principle" — reasoning, legal principles, systematic understanding. Only when we clarify the "principles" can entrepreneurs better know where to focus, while also helping them avoid risks.

China's AI Opportunities

Xiaodong He: I believe the greatest opportunity lies in the digital industrial upgrading happening across industries. This opportunity itself brings unlimited possibilities for AI application.

How to bridge the last mile of industrial application? We need to solve data silos. Once digital infrastructure is built, how do we apply large models and AI technology to extract intelligence from data, then optimize every industrial link?

Take JD.com's logistics: after digitization, we can use digital and intelligent technologies with optimal human-machine interaction to complete delivery. Or consider intelligent elderly care services — automatically calling elderly people living alone through intelligent robots, with automatic alerts to community workers if issues arise.

Jian Ma: The greatest opportunity is the shift from "human-centered" to "human-oriented." As the economy develops and society progresses, we're in a new demographic phase. In the past, when solving problems, people came first; in the future, machines may be the solution. "Human-oriented" means liberating hands and unnecessary work, letting people focus on what long-term development requires. "Human-centered" meant everything required human doing; "human-oriented" emphasizes human creativity.

Li Tian: AI + scenario-based services (finance and consumer)

In implementing digital growth services, we've found that applying AI technology to transform production processes — similar to production technology applications — can dramatically reduce costs and increase efficiency internally, while generating new commercial value externally.

Renhan Li: ① Data ② Technology + rules ③ Professional talent — enterprises

Opportunities are everywhere. How to convert them into productive forces: first, data authenticity and quality; second, technology plus rules, an inherent characteristic of AI; third, enterprises must possess their own AI talent, otherwise costs will be substantial.

Jinjian Zhang: New globalization

Every revolution in human productivity has accompanied a new wave of globalization. This globalization puts everyone on the same stage again, and may bring new rules and ways of playing.

For AI entrepreneurs, this represents not just exploring Chinese industry, but global exploration in this industry. At this stage, Chinese companies may well become future global leaders. So I see this as a new globalization opportunity for China's future enterprises.

Mainstream Application Scenarios

Haitao Song: I believe it's exploring how to develop vertical applications of large models in relevant domains. Take energy: for traditional energy, from recognition through production, how to use existing knowledge and rules to create better safety supervision system models. Or financial investment — PE and VC, including secondary market broker partnerships, to build professional domain digital financial analysts. And the metaverse: city-wide spatial digitalization foundation data models.

Renhan Li: Every industry has needs; it depends on benefits to people. I believe medical large models have strong prospects. AI for Science in drug research can improve efficiency and quality while reducing costs, highly impactful for human health. The key is organization and how social forms evolve. Intelligent development brings social order reconstruction, and how this reconstruction shapes future needs requires research across all industries.

AI plus robotics is, in my view, the best future industry. This industry may reach hundreds of trillions in output, not merely trillions.

Investment Observations and Logic

Jinjian Zhang: The AI revolution is a productivity-level revolution. Large models can be compared to compression algorithms, abstracting previously unmappable data into outputs translatable to other dimensions, expressing what couldn't be expressed, standardizing knowledge and experience we couldn't previously professionalize. Compared to China's rapid growth, the professional talent cultivation cycle far exceeds the demand cycle. Today demand may be exploding, but the professional talent pipeline hasn't caught up. I believe whether digital humans or large models integrated with industry, many previously unstandardizable and unscalable professional problems can now be solved — medicine, law, accounting among them. There are massive opportunities to explore. Industries' embrace and enthusiasm exceeds our imagination; everyone is remarkably open, willing to experiment and try new things.

From fundamental investment logic, we're not ultimately investing in large models or AI, but in the digital twin or new digital world humanity is constructing. We're investing in infrastructure for the future digital ecosystem, including large models, data privacy, and security governance.

AI Governance and Vision

Xiaodong He: We should focus more on universal and inclusive AI technology, hoping everyone, especially vulnerable groups, can truly use AI to help improve their capabilities, participate in productivity growth, and even benefit from redistribution to enjoy quality of life improvements from technology. This matters for AI technology design too.

Xiaqing Huang: Humanity must have ultimate protection — a kill switch to shut all robots down. I believe this ultimate protection is essential.

Haitao Song: Pull the plug and there will be backup batteries. I believe two keywords matter: boundaries and symbiosis. How to define boundaries? First, boundaries for programmers in development; second, boundaries for enterprises launching applications; third, boundaries for government in rulemaking; fourth and most important, human-machine boundaries. As humans become more robotic in increasingly efficient thinking, and intelligent robots become more human-like, the boundary between human and machine blurs.

Symbiosis matters because we're embracing technology. Intelligent robots will be our new smart home appliances. In ten years, we may measure every country and society's development by the number of intelligent robots owned by families, communities, or society.

Li Tian: Intellectual property attribution may become a major obstacle to further AIGC application. Large models' inherent limitation is inability to discern truth behind data, closely tied to data quality, with no good solution. Also, large model discrimination and bias — training data is affected by data annotators' emotional factors.

Renhan Li: First, governance with new technology emergence. AI governance should embed into governance for the new era — this is crucial. Safety has two types: human-caused, and unknown, uncontrollable ones. My advice to entrepreneurs: have social responsibility. For social governance, safety, and ethics — you must be accountable. With this responsibility, you can embed what you believe won't cause problems into products. Follow international standards and you'll continuously improve. With such governance capability, self-responsibility, plus national governance keeping pace, your enterprise becomes world-class. I'm using past thinking to understand the future; understanding history guides the future.

Many uncontrollable technologies will emerge — we call them mutations. How to handle the uncontrollable? I suggest researching an evaluation platform system, with evaluation platforms for different scenarios. Whoever masters this evaluation platform system may master the future.

Next Phase

Li Tian: Looking forward to more commercial value emerging from AIGC applications. Hope to see more practical, commercially valuable AI robots better serving society.

Jian Ma: Hope for greater consolidation next year, and larger innovative AI platforms emerging.

Xiaqing Huang: Looking forward to the ultimate application scenario of large models plus robotics: humanoid bipedal robots.

Xiaodong He: Industry truly generating value is still one last mile away. Looking forward to more industry large models that truly solve industry problems and create closed-loop industry value.

Haitao Song: First, hoping humanoid robots truly apply and land next year; second, hoping brain-computer interfaces truly achieve breakthrough deployment and application, serving patients truly suffering from diseases.

Renhan Li: First wish is for AI governance to gain deeper public understanding; second hope is for medical large models to bring real benefits to ordinary people.

Jinjian Zhang: Looking forward to seeing Chinese enterprises truly become global leading explorers and leaders in certain domains through this new globalization process.

Oasis Capital is a new-generation Chinese venture capital firm dedicated to discovering the most vital entrepreneurs of China's next decade and growing with them to create long-term value. "Vitality" is Oasis's vision and mission. This vitality represents both the direction of era-defining structural transformation and the resilience and evolutionary power of entrepreneurs. Oasis Capital focuses on early and growth-stage investments, with individual investments ranging from $3 million to $30 million, concentrating on robotics, artificial intelligence, and technology services to support China's technology-driven new service upgrade.