When digital transformation enters the era of big data and cloud computing, where is the new industry?

AI is considered the next blue ocean. The birth and popularity of ChatGPT made the industry reexamine the importance of data, algorithms and computing power. Behind the competition of many large models, the foundation software for intelligent analysis of data and intelligent decision-making has also become a strong demand for enterprises to move towards digitalization. In China, this trend seems more obvious. The digital subject matrix led by the government and large and medium-sized enterprises has a more eager desire for more "intelligent" data to enable business decision-making, which also makes a lot of enterprises who are deeply engaged in the "intelligent" track of data stand out and come to the center of the stage.

In the new wave, DataCanvas, regarded as the leader in the field of foundation software of artificial intelligence in China, is one of the best. In September 2022, DataCanvas announced the completion of the C+ round of financing, making the capital market begin to pay attention to the new dream of this company with the vision of "empowering global enterprise intelligence upgrade".

As an AI foundation software provider that continues to serve the government and enterprises to upgrade their digital intelligence, DataCanvas, founded in 2013, focuses on the research and development of AI foundation software, and is committed to helping users achieve digital intelligence upgrading and promote the large-scale application of government and enterprise AI through a series of self-developed platform software products and solutions required for enterprise level AI applications.

Under the impact of the epidemic, DataCanvas still bucked the trend and achieved rapid growth. In the past two years, its revenue has maintained a continuous 100% growth, of which software products accounted for more than 60%. In addition to increasing customer penetration in the financial industry, DataCanvas has benchmark customers in government, communications, manufacturing, transportation and other industries, demonstrating its sustainable profitability of healthy and healthy growth.

How can DataCanvas gain the favor and trust of the capital market when the data "intelligent" competition ushers in a historic window of opportunity? How to view the new trend of the industry? Recently, the reporter of China Business Daily interviewed Fang Lei, Chairman of DataCanvas.

AI foundation software enables enterprises to upgrade digitally and intellectually

China Business Journal: The latest round of C+ financing has made the capital market pay attention to DataCanvas again. In your opinion, what is the main reason why the company can continue to gain the favor of industrial investment?

Fang Lei: DataCanvas is an AI foundation software company, which should be one of the main reasons why the capital market pays more attention to us. For most companies in the industry, what they sell to customers is actually an artificial intelligence "model" - it can be understood that they have relatively strong model making ability and can quickly train "models" and sell them to customers, while we sell a whole set of software to customers and let them make "models" themselves.

From the perspective of the mainstream commercialization of IT in the world, for example, in the U.S. market, the most mainstream realization mode is to sell the technology platform to enable the end user to carry out independent development and self-service. But today we gradually realize that it is impossible to face the ever-changing terminal demand to sell the last kilometer directly, and AI demand is very long tail, so we are gradually beginning to realize the importance of this road.

China Business Journal: Can you briefly introduce what AI foundation software is and what role it plays in the AI industry?

Fang Lei: For example, like common office software such as Excel is a tool for many people. You can make all kinds of reports to meet your needs on it, instead of directly making a table or sending you the desired results. Therefore, a popular definition of "foundation software" in the industry is that if some underlying software can be re-developed to obtain applications that meet the needs of developers and support the operation of these applications, we call it foundation software.

With regard to industry positioning, the AI industry is hierarchical. The bottom layer is called the computing layer, the middle layer is called the technology software platform layer, and the top layer is called the application layer. The computing power layer is basically dominated by hardware manufacturers. For example, chip vendors provide hardware computing power while providing high-performance storage capacity. On this basis, the middle platform layer is the software platform provided by vendors like us. It can run on large-scale machine clusters, import a large amount of data, build models from them, and finally implement the launch and operation of models to support various business and scenario requirements.

China Business Journal: Machine learning platform has become one of the important tools for digital transformation of many industries in recent years, especially in the financial industry. From the service experience of DataCanvas, why can't industry digitalization be separated from machine learning? What pain points does it solve for financial industry digitalization?

Fang Lei: Digitization is a big category. After data precipitation, we need to make statistics on the data. Machine learning mainly solves the problem of prediction. Take the bank as an example. Statistics can tell us who failed to repay on time last month, but we may need to predict who may fail to repay in the next month and intervene in advance. This depends on some prediction algorithms of machine learning.

However, it should be emphasized that prediction alone is not enough. For a financial institution, it may be necessary to make decisions, which is why DataCanvas launched the causal learning open-source project. For example, in the bank credit repayment business, in many cases, the ordinary predictive algorithm of machine learning can only predict who will not repay in the next month. However, if a certain amount of repayment coupons is given or the repayment cycle is extended for a certain period, he may repay again, which needs to be deduced to facilitate the financial institutions to make decisions.

From the perspective of the development of the industry, the demand of many enterprises moves forward step by step in the process of statistics, prediction, deduction and finally decision-making. Therefore, there must be a systematic demand for corresponding prediction, deduction and final corresponding decisions. In the digital transformation of the entire enterprise, machine learning is indispensable for data processing. They must use foundation software based on machine learning to build more intelligent decisions or new applications.

The key to sound development of AI industry lies in industrialization

China Business Journal: Compared with "digitalization", DataCanvas prefers to use the industry label of "digital intelligence" in many occasions. What is the reason for more emphasis on "intelligence"?

Fang Lei: First, we are positioned in the category of artificial intelligence, so we will highlight the concept of intelligence. Because "digitalization" in the original sense is more electronic or information-based process, we generally call it digital transformation, such as developing enterprise OA process system into software; Later, with the rise of mobile Internet, we regard online as digital; In the future, people said that digitalization also has the second half, which is to make it more intelligent.

Secondly, from informatization and online to today, it covers intellectualization. As an enterprise in the AI arena, DataCanvas also hopes to highlight the words "data" and "intelligence" more actively, especially "intelligence". Because our products, from platforms to real-time data, are basically built around what we call foundation software, which also better reflects our core characteristics, which is why we emphasize "intelligence".

China Business Journal: According to the experience of DataCanvas, where are the main pain points of AI industry clients at present, and how to meet these needs of clients?

Fang Lei: In the process of absorbing the underlying capabilities into the business, clients have several pain points. The first is the threshold, because it is difficult for most large state-owned enterprises to maintain a team of data scientists or analysts of the same size as Internet companies. Therefore, they hope to lower the threshold for the use of the entire AI algorithm. Let people in the business team or the information department be able to contact and use them skillfully. However, the model demand of different enterprises is completely different.

The second is the expandable landing capability, which is called closed loop, which is also a great challenge. For example, we can spend 1-2 million yuan to make a verification prototype for a scenario, but it may cost 100-200 million yuan to achieve a complete implementation, or even if it is not handled properly, it will not be able to achieve the goal.

The third pain point is that enterprises lack a group of people who can "translate" business problems into problems that can be solved by AI. For example, the problem of predicting who will not repay the loan next month is translated into a machine learning problem, which is called a dichotomy problem. This problem is reflected from the business to AI technology and is "translated" by a product manager or project manager.

Among the three pain points, the threshold challenge is the biggest, but as AI vendor begin to provide services, the difficulty of this problem is rapidly reducing. The rest is to do closed loop, which requires customers to reform their own processes. Personally, I believe that sharp clients are constantly adding intelligence to the business process and are reshaping their own business processes.

China Business Journal: At present, the domestic AI industry seems to be going through a cold winter. Many unicorns shining in the AI industry a few years ago have encountered development bottlenecks. How do you view and respond to this round of current challenges in the AI industry?

Fang Lei: I think the biggest problem now is that both the media and regulators have a comprehensive understanding of AI, which means that AI itself has become a company rather than an industry. If AI is really regarded as an industry, it must have a division of labor, which is why DataCanvas always strives to make its business field more detailed and specific, rather than all inclusive.

Today we go to see the ecology of some global companies. In fact, the division of labor is relatively detailed. What industry does the application do, what field does the foundation software do, and whether it does hardware or software, these are all very subdivided. In China, it is unified at present. Although this mode will enjoy the income dividend brought by integrated projects in the short term, it will have problems in the long run.

The launch of the "AI cloud in clouds" strategy is to seize new opportunities in the IT industry

China Business Journal: After this round of financing, how will DataCanvas conduct the next business layout in the AI industry?

Fang Lei: Our next strategy is called " AI cloud in clouds ", because we believe that China's cloud market will be more fragmented than that of the United States, and market players, mainly state-owned enterprises and central enterprises, will build their own clouds. As long as there are enterprises with such demands, they are our clients in the traditional sense. Because their IT department has AI capability needs, this is where we can serve them.

For example, when they build their own cloud companies, it is actually a process for their IT departments to go global. Then we can just embed our own products into the process of their going out to serve the outside on the basis of serving them internally. We call it entry, penetration and embedding, that is, we first enter into enterprises, then become their infrastructure, and then embed into their cloud services and the upstream and downstream of their industries.

China Business Journal: With the deepening of the digital transformation of the domestic industry, how does DataCanvas anticipate the opportunities and prospects of future industry digitalization?

Fang Lei: At present, the whole industry has reached a consensus on the overall prospect, which is very optimistic. But what is the final form of future digitalization? This is the direction that DataCanvas is most concerned about.

We first targeted the foundation software. Because in the past 30 years, Chinese software or software service companies rarely tell us what their products are because they are essentially service providers. In the current cloud era, domestic vendors have the opportunity to become product vendor, and also have the ability to make good products. In addition, localization is the trend of the times, so this market will have a lot of space.

This is precisely the purpose of DataCanvas' "AI cloud in clouds" strategy. We are looking forward to the expansion of cloud vendors, and the market will always have demand for service vendors, so the domestic market will certainly emerge a batch of very excellent new IT product vendors. And we want to be one of them.

Boss' Secret Script

1. What are the main reasons why DataCanva can continue to gain the favor of industrial investment?

DataCanvas is an AI foundation software company, which should be one of the main reasons why the capital market pays more attention to us. For most companies in the industry, what they sell to customers is actually an artificial intelligence "model" - it can be understood that they have relatively strong model making ability and can quickly train "models" and sell them to clients, while we sell a whole set of software to customers and let them make "models" themselves.

From the perspective of the mainstream commercialization of IT in the world, for example, in the U.S. market, the most mainstream realization mode is to sell the technology platform to enable the end customer to carry out independent development and self-service. But today we gradually realize that it is impossible to face the ever-changing terminal demand to sell the last kilometer directly, and AI demand is very long tail, so we are gradually beginning to realize the importance of this road.

2. How will DataCanvas conduct the next business layout?

Our next strategy is called "AI cloud in the clouds", because we believe that China's cloud market will be more fragmented than that of the United States, and market entities dominated by state-owned enterprises and central enterprises will build their own clouds. As long as there are enterprises with such demands, they are our customers in the traditional sense. Because their IT department has AI capability needs, this is where we can serve them.

For example, when they build their own cloud companies, it is actually a process for their IT departments to go global. Then we can just embed our own products into the process of their going out to serve the outside on the basis of serving them internally. We call it entry, penetration and embedding, that is, we first enter into enterprises, then become their infrastructure, and then embed into their cloud services and the upstream and downstream of their industries.

Introduce

Fang Lei, graduated from Tsinghua University, holds a master's degree in Electronic Engineering and Computer Science (EECS) of Syracuse University, and a doctor's degree in Electronics and Computer Application (ECE) of Virginia Tech. The former data scientist of Bing search department of Microsoft participated in the development of Microsoft cloud computing platform: Windows Azure as an early team member. In 2013, he returned to China and founded DataCanvas Ltd., committed to creating the most advanced AI foundation software services. With 20 years of experience in big data analysis and management, he is the leader of advanced technology in deep learning and machine learning. He has published 19 papers in the field of distributed systems, design verification and algorithms, and has been cited for more than 700 times. In the fields of artificial intelligence, data science, big data and so on, he has been employed as a think tank expert, director, executive director and other positions for many times, and has been rated as the annual artificial intelligence industry innovation figure for many times.

Insight

A pragmatic thinker standing on the "cloud"

Like any new technology industry, the initial stage of AI in China was also a time of recklessness, rushing forward, and achieving a paper prosperity in the early stage of the industry. However, when the tide receded, many star companies that had been placed high hopes revealed their true colors. The unsustainable performance punctured many illusions of the outside world, but also because of the screening of the industrial boom and decline cycle, more pragmatic people emerged.

DataCanvas, which is rooted in the field of foundation software of artificial intelligence, is obviously one of them. From the beginning of its establishment, this enterprise seems to have never had the ambition of dominating the whole industry. Instead, it has been subdividing the racetrack, correcting the direction while observing. In Fang Lei's words, it is closely following the change of the industrial situation and taking advantage of the strong to complete a higher leap.

The so-called "transformation", on the one hand, is to adapt to the change of the industrial environment itself. After AI application scenarios cross the incubation period and enter the large-scale application stage, the whole industry is more targeted in choosing AI services. On the other hand, the rise of the cloud era has also allowed the AI industry to smell more business opportunities. Taking advantage of the "cloud" has become a new option for DataCanvas.

In Fang Lei's view, as governments and enterprises have launched into the cloud, building AI infrastructure on the cloud can not only effectively support business scenarios on a more autonomous and controllable basis, but also be more economical. At the same time, relying on the friend circle and ecological chain of thousands of clouds, DataCanvas can enter the upstream and downstream of the industry to obtain more customers, to cover the whole process of AI foudation software developed by DataCanvas and achieve automated, low threshold AI agile delivery.

This is also the essence of the syllogism of "entering, deepening and embedding" in DataCanvas' new strategy. In the process of communicating with Fang Lei, he spent most of his time explaining his understanding and thinking about digital transformation and cloud industry. Its core point is "the upgrading of software infrastructure in thousands of clouds will become a historic opportunity for enterprises to accelerate the transformation of digital intelligence". While explaining the changes in the IT industry, it also made the outside world understand the reason why this company achieved a growth of more than 100% against the trend under the influence of the cold winter of AI capital and the epidemic of COVID-19.

With keen insight and precise grasp of user pain points, DataCanvas has successfully implemented AI in more than 10 industries, including government, finance, communication and manufacturing, and has cooperated with leading enterprises such as China Mobile, Bank of China, Shanghai Pudong Development Bank, CITIC Securities, China Baowu, Hisense Group and CRRC to create a successful practice of digital intelligence upgrading for multiple AI scenarios. The initial effect of the "AI cloud in clouds" strategy is gradually showing, which also verifies Fang Lei's insights.