Investment Opportunities in the Era of Large Models
-
Strong venture capital fosters thriving industries. The '2023 Wuhan Innovation Investment Ecosystem Conference' was held in Wuhan on November 3, 2023. The conference, themed 'Building an Investment Ecosystem to Drive Innovation,' aimed to synergize fiscal funds, industrial capital, social capital, and innovative forces to discuss new developments and advance Wuhan's sci-tech innovation initiatives. At the event, Mei Xianfeng, Deputy General Manager of Wuhan Innovation Investment Group, moderated a roundtable discussion titled 'Investment Opportunities in the Era of Large Models.' The panelists included:
Han Yi - Executive Director of Yunqi Capital
Ji Li - Managing Director of Huagai Capital
Ouyang Yanjun Partner at Harmony Capital
Wang Xuemei Partner at Inno Angel Fund
Wei Haitao Partner at Blue Lake Capital
Ye Zhigang Chairman & CEO of Wuhan Green Network Information Service Co., Ltd.
Yuan Hongwei Founding Partner & Chairman of Zhuoyuan Capital
Zhou Qi Managing Partner of GSR United Capital
The following is a transcript of the dialogue,
Mei Xianfeng: Before starting this roundtable discussion, I'd like to briefly describe the current state of this year's investment market. Recently, I attended a venture capital summit in Beijing where an industry leader summarized the investment market of the past two years with three keywords: 'Hard, hard, hard,' 'Down, down, down,' and 'New, new, new.' Let me borrow these terms here. 'Hard, hard, hard' refers to the increasing difficulty in finding good projects these past two years. 'Down, down, down' indicates that fund sizes have been shrinking compared to previous years. 'New, new, new' suggests that new industries and opportunities continue to emerge this year, offering hope. Of course, this year's capital market performance hasn't been great, with ongoing adjustments, but we've noticed one particular sector consistently attracting investors' attention, whether in primary or secondary markets.
This niche sector is the artificial intelligence and large model industry. On stage, we have partners from venture capital institutions as well as founders of related companies. Next, we will explore the investment opportunities in the era of large models.
First, I'd like to invite each guest to introduce themselves. Based on your experience and actual investment cases this year, please share your perspectives on the large model investment market and opportunities. Would you prioritize investing in large model infrastructure or applications?
Han Yi: Yunqi was established in 2014, primarily focusing on two major sectors: technological industrialization and industrial digitalization.
Yunqi has been paying attention to AI for a long time because this broad direction, from the earliest CV recognition to voice technology and then to the emergence of Bard, has been evolving for quite some time. Now, large models have optimized supervised learning scenarios as well.
We have already invested in several companies in the large model space: an early investment in MiniMax, a company focused on developing self-researched multimodal large models; and leading an investment in Zilliz, one of the top three vector databases globally. When it comes to the industrial implementation of large models, not only are models needed, but also vector databases to support application scenarios. We are investing across the ecosystem chain of models, and this year, we invested in an early-stage startup working on AI agents. The development of this industry is really just beginning.
Currently, the ecosystem around large models is very hot. However, having invested in AI from the early days, we’ve already gone through one cycle: initially, the hype was intense, but the development of companies has been gradual. Now, things are slowly cooling down, and companies are beginning to iterate. Our overall feeling is that the optimization and iteration speed of models isn’t as fast as expected.
Mr. Wang mentioned that the emergence of large models may replace some creative jobs like secretarial and marketing work. For tasks requiring high consistency and accuracy, models will need to rely on peripheral engineering capabilities to improve these aspects.
Overall, we believe China's large model industry presents many opportunities, as the upstream of large models requires GPUs and computing power cards for training, driving the iteration and growth of the entire data center industry chain, primarily IDC.
Ji Li: Huagai was established in 2012 and has been around for 11 years. Currently, we are mainly focusing on two major sectors: healthcare and technology. Healthcare may account for the majority of Huagai's managed capital, while we have also been investing in the technology sector since our inception.
Huagai Capital currently manages assets totaling approximately 20 billion yuan and has invested in over 160 companies, with about 30 having completed IPOs. My role at Huagai primarily involves the establishment and investment management of digital technology-focused funds.
In the field of large AI models, Huagai's investment style and strategy primarily target growth-stage opportunities while maintaining a continuous focus on the large model sector. We specifically seek out relatively certain investment opportunities and directions. Previously, we have made strategic investments in areas like GPUs, which involve computing power infrastructure. Since the emergence of large models, the demand for computing power and the potential market size have grown exponentially. Therefore, in the computing power sector, we are actively pursuing more definitive investment opportunities.
Additionally, in the application layer, we have been closely monitoring the broad direction of artificial intelligence. With the emergence of this wave of large models, AI has evolved to the 2.0 or 3.0 stage. We have been observing this since the 1.0 stage, including the applications of AI in the financial sector and enterprise services. Currently, we are also exploring some definitive opportunities on the application side, such as a large model application incubated by iFlytek, which has shown promising development in IoT scenarios for enterprises.
Therefore, when examining the large model landscape and investment opportunities, we tend to focus more vertically on the underlying infrastructure and the top application layer. Horizontally, we are identifying two key opportunities: one is incubation by major companies or validated application scenarios, and the other is seeking technologies or products with the potential to become platform-based or ecosystem-oriented in both technical and product aspects.
Ouyang Yanjun: I am primarily responsible for hard tech investments, currently focusing on two main areas: healthcare and technology. Heyu is an investment firm that pays close attention to early-stage foundational innovation opportunities, and we operate as a dollar fund.
Established in 2015, we initially focused on internet+ investments, such as Boss Zhipin, Meituan, and Zhaogang.com, empowering the digital transformation of traditional Chinese industries. In recent years, we have specialized in hard tech. Starting around 2017-2018, we began investing in AI-related companies, spanning from foundational computing chips to mid-layer applications, including autonomous driving, medical big data analysis, and 3D vision robotics. In fact, we have invested in many so-called 'previous-generation AI' companies, particularly decision-based AI firms. Generative AI trains machines in an unconventional way, giving them greater generalization capabilities, which represents an epoch-making revolution. The emergence of a platform-level model will undoubtedly give rise to a vast application market. This trillion-dollar market is incredibly exciting for investors.
We focus on two main areas: the underlying computing power. Domestic chip manufacturing presents significant opportunities and is essential, regardless of profitability—everyone should support it. Additionally, there are numerous opportunities in applications, particularly leading applications or new vertical applications. Recently, we've been actively exploring but haven't made major moves in large model-related areas yet. We've invested in an AI company that serves as a comprehensive R&D platform and an AI industrial design platform, enhancing product efficiency for industries like healthcare, smart manufacturing, and new energy. We are gradually deepening our expertise in this sector. Heyu Capital's primary advantage lies in its strong presence in the Middle East, while we also invest in early-stage projects in the U.S.
Wang Xuemei: InnoAngel was founded in 2013 and has now been operating for 10 years. We have consistently focused on early-stage investments, managing a 5 billion RMB angel fund and investing in over 500 early-stage projects.
InnoAngel, with its roots in Tsinghua University, recently registered a new angel fund in Optics Valley, primarily focused on investments in artificial intelligence and next-generation information technologies.
In terms of investments, InnoAngel has made numerous investments across all layers, from the foundational to the middleware and application layers. At the foundational and computing power level, we invested in Yuanli Semiconductor, which develops core functional chiplets. In the model space, we invested in Shenyan Technology, whose model capabilities are among the best in China. Shenyan covers all text generation tasks with multimodal approaches and, in collaboration with Tsinghua University's NLP Lab, released the first open-source large Chinese-English model system called Yujing. Founder Qi Fanchao mentioned that they received significant help from the open-source community during their startup and growth phases, and now they want to give back by open-sourcing some of their technologies and data from training large models, welcoming everyone to use them. Additionally, we invested in two AI for Science projects: one focused on medical material R&D and another on lithium-ion battery cathode and anode material development.
Wei Haitao: Our fund primarily focuses on two investment directions. One is enterprise digitalization, which corresponds to enterprise-level application software, and the other is industrial digitalization, essentially smart manufacturing, including digital and intelligence-driven industrial upgrades. We don't have many restrictions on stages; some companies are invested in from the angel round, with the main focus on Series A and B rounds.
This past year has been quite insightful. We've invested in many product-focused companies in both enterprise software and manufacturing sectors. Previously, during shareholder and board meetings, it was often the investors asking the entrepreneurs: 'Have you looked into this new direction?' However, with the rise of AIGC and large models, the trend has reversed. Many of the CEOs and CTOs from our portfolio companies are now asking us if we're using large models—they're already using them. This wave has garnered more attention in the industry than in the investment community. They're already applying these large models to practical tasks like quality monitoring in production processes and data collection and processing.
I believe every investor here hopes to find their own unicorn in this wave and invest in it at an early stage. Every industry's development follows its own patterns, always starting with technological progress. This current wave began in the United States, where technology evolved into products. As these products gradually matured, they entered a market-oriented era, using their technological foundations to target specific markets. Currently, large language models are undoubtedly in a phase of technological development. Looking back at the previous AI era around 2013-2014 when AI was just emerging, investors were also constantly searching for directions. At that time, the AI field was divided into numerous specialized segments.
During that AI era, after thorough research, we ultimately focused on autonomous driving. In this field, we invested in several angel-round projects, which have now grown into highly successful companies.
Looking back, we reviewed all the promising AI companies of that era, but our high valuation standards led us to miss some investment opportunities, such as Fourth Paradigm, when the timing wasn't right.
From today's perspective, large models are proving to be a tremendous driver and enabler for the future development of the entire industry. Whether in B2B or B2C sectors, many professionals are already using ChatGPT to solve work-related problems, demonstrating immense value.
At the commercial level, we still return to the fundamental question: what specific problems does this technological advancement solve for customers? I believe this aspect will differ between China and the United States.
We are very confident in China's innovation and development at the application layer. In China, especially for B2B clients, people may not care much about what advanced technologies you use. What they care about most is whether your solution actually solves the problems faced by enterprises or in production and supply chains. We hope to see it bring significant innovation to the industry and create real value for customers. At the same time, in China, we expect to see a strong willingness to pay.
Ye Zhigang: I am the Chairman and CEO of Wuhan Green Network Information Service Co., Ltd., and also serve as the CTO. In fact, I am an engineer.
Wuchuang Investment recently invested in our pre-IPO round. Donghu Development Zone, Wuhan City, Hubei Province, and the National Investment Group have all invested in our company. Our company operates in the infrastructure sector, with clients being the three major telecom operators. We focus on the traffic of operator networks and the underlying technologies involved in large models and artificial intelligence, such as machine learning and deep learning, which we have been using for several years.
More importantly, the AI direction ignited by ChatGPT and OpenAI is clearly a massive market. Since our founding in 2020, our traffic has grown 2,000-fold. Our prices have decreased primarily because the underlying growth has been exceptionally strong, leading to a very high growth rate. This year, we anticipate another substantial increase.
The reason for this success ultimately comes down to selecting the right foundation and direction. In the field of AI, I've seen data projecting a market worth $2.6 trillion to $4.6 trillion by 2040. Not every company can choose its own path, but if the broad direction is correct and you become the leader in a niche area, you can certainly seize the opportunity. While our overarching direction remains unchanged, the specific focus has evolved. Even with these changes, we continue to capture opportunities. This is indeed a significant opportunity, but we remain committed to deepening our expertise in our field rather than following trends. We constantly ask ourselves: when the tide recedes and the wind dies down, will we still be standing?
Yuan Hongwei: When the pandemic started, our team ventured out from another platform to start our own business. To this day, our strategy remains unchanged—regardless of the conditions or external environment, we continue to focus on our core approach: investing in the most cutting-edge technologies.
This team has been working together since 2016-2017, a time when artificial intelligence was just beginning to gain attention in China. During the early stages of AI's development, we were fortunate to adopt a strategy of comprehensive investment in leading companies across related fields. In autonomous driving, we chose Pony.ai; in AI vision, we selected SenseTime; and in robotics, we invested in Ubtech, along with Fourth Paradigm—all strategic moves made during this period. For China, this decade to two decades represents the most promising era for technology startups, offering the best opportunities and attracting the greatest societal attention and resource allocation. With such significant resources concentrated in this sector, I firmly believe it will give rise to truly remarkable companies.
In the realm of cutting-edge technology, we have long been strategically positioned in the field of large models. In the AI digital human sector, we have invested in Tuoyuan Intelligence, while in the large model domain, we have chosen the well-established Baichuan Intelligence. Although we are an individual platform, our team primarily consists of graduates from Tsinghua University's artificial intelligence program, focusing on many entrepreneurial teams founded by Tsinghua alumni. We frequently collaborate with familiar institutions like InnoAngel and Baidu Ventures.
Regarding our focus on large models, we have developed our own approach. Firstly, China will undoubtedly have its own large models—this is an opportunity that cannot be missed. Moreover, these large models essentially represent the super-brains of the future. With our foundation in AI talent and the nation's current stage of development, it is imperative that we possess our own core technologies. We are committed to making significant strategic investments in frontier technologies.
The three key infrastructure directions we’ve identified are computing power, storage capacity, and transmission capability. These present excellent opportunities for domestic innovation companies. Of course, this won’t happen overnight—it will take considerable time. However, this is a tangible and existing market, one that we must address through our own efforts, and one that will attract societal resources to support its development. As such, we continue to invest in infrastructure related to computing power, including semiconductor chips.
In summary, large-scale models will bring many opportunities, and we remain highly optimistic, committed to ongoing attention and investment in this field.
Zhou Qi: GSR Ventures currently manages over 7 billion RMB in assets and has invested in more than 100 projects. We primarily focus on 'two hard and one soft' directions: intelligent manufacturing, next-generation information technology, and digital enterprise services. Our investments are mainly industry-focused with a B2B approach. Regarding large models, we understand them through two fundamental aspects: first, content generation (AGI/AIGC), and second, the transformation of human-machine interaction.
In recent years, the investment pace of dollar funds has slowed significantly. Large models align well with the style of dollar funds - had their investment not slowed, the funding in large models might have been ten times greater. However, this doesn't hinder the development trend of large models in the industry. Meanwhile, we must clearly recognize that China still lacks many original technologies. Large models weren't invented by Chinese researchers. In recent years, the AI field focused on NLP until discovering the 'brute force leads to miracles' approach of large models. It's like practicing various martial arts only to find opponents using cannons - rendering all techniques useless with one blast. We now hope to see more original AI technologies, which requires long-term investment and visionary entrepreneurs to drive innovation.
In terms of the application side, the large model field can be broadly divided into two major aspects. First, there's the model development side, which is primarily an opportunity for large companies or startups with substantial financial backing. Second, there's the application side, which is vast, so the focus should be on these two key areas.
Mei Xianfeng: Most of the attendees here are angel investment institutions, as well as venture capital firms like Huagai Capital and Mr. Ye's Green Network Co., Ltd. Everyone has discussed investments in the large model market—some have been covering it early on, some are looking at the infrastructure of large models, and others are focusing on their applications.
From the previous discussion, it is clear that the large model industry has a high ceiling and promising future trends. New applications developed based on large model technology not only drive the intelligent upgrading of various industries but also foster further economic development. In recent years, Wuhan has listed artificial intelligence as a key industry. Considering Wuhan's advantageous industries and your understanding of the city, what suggestions do you have for the development of large models in Wuhan? Additionally, how should investors promote and guide the development of large models through equity investments to support technological innovation?
Han Yi: I graduated from Wuhan University and spent four formative years here, so I have deep affection for this city. In my previous understanding, Wuhan's economic structure relied heavily on heavy industry, steel, and chemicals, supporting a GDP of 1.8 trillion. Another characteristic of Wuhan is its abundance of universities, which represent talent and scientific research. We have many invested companies collaborating with various enterprises in Wuhan. For example, our invested company Yuanrong Qixing is partnering with Dongfeng Group to implement autonomous driving projects.
When I graduated, the Optics Valley was gradually rising, representing the growth of the electronics industry. Wuhan's core advantage lies in its talent pool, and under the framework of a research-oriented talent system, more technology industries will emerge. Additionally, with the high labor costs in first-tier cities, many digital companies are relocating talent to Wuhan and Chengdu, presenting an opportunity for Wuhan. The government is also providing strong support, offering a favorable financing and investment environment.
Secondly, how can funds help Wuhan achieve tangible results in the future? Wuhan's smart manufacturing sector is developing well, and factories in mechanical processing and precision engineering industries can relocate to Wuhan, where the supporting infrastructure in the middle and upper reaches of the Yangtze River is already well-established. In the future, there will be more opportunities to help some of our invested industries establish a presence in Wuhan, driving the city's development.
Ji Li: In the past two years, I haven't had many opportunities to visit Wuhan, but I've been to Hefei quite frequently. People often compare the industrial foundations and government styles of Wuhan and Hefei. Coincidentally, we've been looking at AI-related startups in Hefei, particularly those emerging from iFlytek, while also examining a promising startup in Wuhan focused on computing power and photonic computing. I believe each city has its own unique industrial foundation.
Hefei has cultivated several successful listed companies in the artificial intelligence sector, boasting a solid industrial foundation. Wuhan, on the other hand, has deep expertise in optical computing, chips, and intelligent manufacturing. Therefore, if offering suggestions to Wuhan, it would be advisable to focus on strategic layouts in cutting-edge technological industries like large models. There are two key recognitions to make: First, recognize which aspects of the industry can be government-driven, such as identifying certain investment opportunities and attracting businesses to establish local operations. Additionally, early-stage projects from universities and research institutions that allow for long-term planning should also be considered. The second recognition involves identifying Wuhan's unique advantages within the large model industry chain—areas where breakthroughs can create standout successes and leverage a single point to build a more comprehensive ecosystem or industrial chain. It's not about being all-encompassing but about developing distinctive strengths.
The final issue is how to promote industrial development from a fund perspective, aiming to identify promising enterprises and secure appropriate funding to sustain long-term industry growth.
Ouyang Yanjun: I'd like to make three points, continuing with the previous topic. Regarding computing power, I believe our country is indeed lacking in this area. Government support is essential, and we need to increase investment in chip-related technologies. We've studied the support policies for large models in various local governments, and it seems cities like Beijing, Shanghai, and Shenzhen offer more favorable policies in this regard. Wuhan could consider emulating these cities by increasing investment in computing centers and data trading, establishing a smart computing hub that could play a coordinating role. Wuhan might already have such initiatives underway, which could serve as references.
Regarding fund investments, Wuhan has also made some early-stage arrangements, such as angel investments. As I mentioned earlier, whether it's computing power or large models, investments become increasingly difficult to sustain in later stages. There needs to be a comprehensive plan covering early, middle, and late stages, with a concept of relay investment. This is an architectural issue—if early-stage institutions complete their investments but no one follows up, it creates problems. We've seen some chip companies facing this exact challenge. While many governments are providing support, I believe it's crucial from a governmental perspective to foster some long-term business initiatives.
Wang Xuemei: Wuhan is poised to become the fourth major city in new energy development, following Beijing, Shanghai, and Shenzhen. Therefore, I believe there are significant opportunities for vertical industry models in the energy sector here in Wuhan. We are currently introducing an energy model and working to implement it locally. However, we may encounter some challenges during this process, as the energy industry and other vertical sectors possess specialized know-how and face issues of data fragmentation. These factors may affect the depth, sensitivity, and timeliness of the models. Beyond securing top-tier entrepreneurial teams and necessary funding, we also require investors with deep industry knowledge to contribute their expertise, networks, and industrial resources, along with support from various levels of government. This is particularly crucial for vertical industry models.
Additionally, biopharmaceuticals represent another key industry in Wuhan. I currently have a protein model that we're working to establish in the Optics Valley area. I believe this presents fertile ground for development.
The third aspect is that we are also waiting for some killer applications to emerge from language models. The current environment in the capital market is different, and the funding structure has also changed. For entrepreneurs, the days when angel investments amounted to a few million and Series A rounds reached tens of millions are over. In the past, entrepreneurs needed to focus on both the moon and the sixpence in their early years, but now they must balance both. The environment is simply different.
During the morning report, the mayor mentioned that Wuhan now has a population of 15 million. In terms of applications, Wuhan itself is a massive market, offering great opportunities for AI assistants and medical services. When Douyin (TikTok) first started, it wasn’t about short videos or headlines—it later grew into a giant. Similarly, Meituan initially focused on group buying, not food delivery. Therefore, we need to identify top-tier teams, leverage all the resources from investors, and empower them with our insights to grow together.
Wei Haitao: Wuhan possesses exceptionally high-quality university resources, so there is no shortage of talent. In recent years, there has been a significant trend of talent flowing back to relatively inland regions such as Wuhan, Chengdu, and Xi'an. These cities are currently experiencing the fastest net population growth in China and have relatively younger demographic structures. Therefore, Wuhan's talent base is facing a rare historical opportunity.
I personally strongly believe in market-driven approaches—any industry's market-driven development is inherently healthy. As a very new industry, we first need to identify where the opportunities lie. Wuhan has many clients with the financial capacity to pay for large model applications. I believe that if these clients open up their application scenarios and budgets, they will undoubtedly attract high-quality startups and entrepreneurial teams to come here and serve these clients, thereby fostering business growth.
Once these outstanding leading startups arrive, talent will naturally settle here to focus on R&D and product development. I think we should consider opening up application scenarios and finding strong teams to establish their presence here.
Additionally, there are certain endeavors better suited for government investment rather than private capital—particularly long-term infrastructure projects that reduce entrepreneurial costs and attract innovators. Such strategic, far-sighted initiatives require government planning to sustain industrial ecosystems, and Wuhan exemplifies strong foundational conditions for this approach.
Ye Zhigang (from an entrepreneur/engineer perspective): While I agree with all investment-related viewpoints expressed, my technical perspective differs regarding computing power concerns. What truly worries me is whether we can develop genuinely effective, specialized, and diverse applications. China's engineering workforce advantage is undeniable—I recognized this 'engineer dividend' decades ago. While the Pearl River Delta capitalized on migrant labor advantages, Wuhan's concentration of universities and research institutes creates a 6-7x cost differential compared to Silicon Valley. We must acknowledge our shortage of truly innovative talents despite having numerical superiority in workforce size.
For this reason, what China has done well—now we talk about GPC, previously about CPU, storage, and networking—are the applications that consume these resources, whether it's Tencent, QQ, WeChat, Alibaba, JD.com, Meituan, or Toutiao. Overseas youth are crazy about using these, all stemming from China's massive application market. After achieving this, the technological sophistication is indeed quite high. However, from our perspective, it’s not technology-driven but market-driven. At this level, I believe China’s reversal may hinge on chips, and overall investment is also crucial. Without getting this business operation right, it would be very difficult. Regarding business operations, I agree with everything Mr. Wang from Kunzhong said, except he missed one point: from a macro perspective, the combination of multiple technologies could potentially create killer applications that are incredibly powerful. For example, large models and robots like Boston Dynamics—when we first saw them, we were utterly震撼ed. Although the most cutting-edge technologies aren’t open-source, and many core AI technologies remain proprietary, our growth is possible because some technologies are open-source. This includes the metaverse from recent years, where composite innovations combining these technologies with networking present industrial opportunities.
Wuhan's opportunity lies in its high-quality mid-level innovators, not top-tier or globally leading talent—this point must be clearly understood. Many come from Tsinghua University, and Peking University and Tsinghua undoubtedly represent the top tier. If we can effectively utilize this segment of talent with precise government guidance, I believe there is still opportunity. Competing head-on would be very challenging.
Yuan Hongwei: Previously, I didn't have an in-depth understanding of Wuhan, but through some opportunities, I invested in several companies and experienced the local atmosphere. In 2019, I invested in LianTe, and during its IPO, I truly felt the strong regional support for this industry. Additionally, I compared it to other cases—it wasn't a leading enterprise in the optical communications field, but it had deeply integrated with the local industrial chain. I've also invested in some leading companies that, surprisingly, haven't succeeded in going public yet.
My perception may not be entirely accurate, but I do agree with Mr. Ye's point that Wuhan possesses a very strong foundation of engineers with relatively low costs. What's truly lacking are genuinely capable leading talents who can guide them. Whether we look at the experiences of Hefei or the Bay Area, it's essential for the government and capital to collaborate in creating such an ecosystem. Attracting more visionary and innovative talents who can become true tech entrepreneurs to local areas, and having them lead local engineers while integrating with the local environment, I believe, can create truly great companies.
Zhou Qi: I previously gave a presentation on AIGC titled 'A New Era of Cognitive Intelligence' rather than 'A New Era of Artificial Intelligence.' This is because artificial intelligence is divided into computational intelligence, perceptual intelligence, and cognitive intelligence - and we have now entered this third new era of cognitive intelligence.
From the perspective of large models, where exactly does China's opportunity lie? I believe China's development of large models primarily stems from the opportunities created by the technological decoupling between China and the US. So how do we cultivate large models? I don't think this is about leveraging engineers—large models are actually disrupting the role of engineers. For instance, the recently released Full Self-Driving V12 by Elon Musk doesn't contain a single line of manually written code; it's entirely machine-generated. Thus, artificial intelligence is disrupting engineers, whether we admit it or not—that's the reality. To truly develop large models, what's lacking isn't engineers but top-tier scientists. I believe their emergence requires the right environment. Companies like OpenAI and Midjourney, with just over a dozen employees, have achieved annual revenues exceeding a billion dollars—this is a complex and long-term endeavor. Additionally, having long-term thinking and capital support is crucial, and I believe Wuhan has such an opportunity in this regard.
I believe that investments in the large model field can be divided into two ends: one is the technology model itself, and the other is the application side. In the past, we have invested in many AI companies. After the AI wave receded, we found that although the AI industry is very lively, there are problems with application and implementation, and none of them are profitable. They only rely on some low-tech income to maintain company profits. I think when AIGC or the large model industry was just developing, we should also be soberly aware of this situation. In fact, some of the companies we invested in, including digital humans and AIGC, could not find application scenarios in the early years. On the contrary, some companies with very good application scenarios and high profitability were more willing to collaborate with AI, using AI and large models as small tools to provide their services, forming such a state.
Therefore, I believe that in China, we should not regard AI and large models merely as tools but rather as services to sell.
Mei Xianfeng: First, I’d like to respond to Mr. Zhou’s excellent presentation. I believe what you shared does not conflict with Mr. Ye’s earlier remarks. Mr. Ye focused on the development of enterprises and the industry, highlighting that the progress of large models and AIGC requires the efforts of numerous engineers. As the industry grows, its applications will also drive talent to reach higher levels.
Just now, the distinguished guests provided valuable suggestions on the development of artificial intelligence in Wuhan and large models from various perspectives, including government investment, energy models, investment stages, Wuhan’s talent advantages, and capital attraction.
The industry is rapidly evolving, and investment institutions are increasingly recognizing its potential. They have invested in many promising companies, though some opportunities have been missed. Moving forward, I hope our peers can engage in deeper offline exchanges at the project level to support industry growth and share the fruits of its development. Let’s once again thank all the guests for their insightful contributions with a round of applause. This concludes our roundtable forum.