Hongtai Jin Haiyan: AI+ industry is slow work and hard work, there will be commercialization results

Hongtai Jin Haiyan: AI+ industry is slow work and hard work, and there will be commercialization results in multiple tracks in 2019 (Figure 1)

On the afternoon of June 26th, the 15th issue of the Unknown Creative Star Forum-AI Industry Application Scenarios and Future Development Trends hosted by the Peking University Youth CEO Club was held in Beijing. China Youth Daily and Peking University Innovation Review were supported by the media. Guests attended.

Jin Haiyan, an executive director of Hongtai Fund and an alumnus of Peking University's 03-level information management department, was invited to attend and shared her four thoughts on the artificial intelligence industry and investment.

Artificial intelligence + industry as a whole is slow work;

In the artificial intelligence venture capital circle, it is definitely not the winter of capital;

The technology-driven innovation opportunities brought to various industries are the foundation of capital favor;

Many industries are accelerating commercialization, and 2019 will see some commercialization results.

The following is the speech record:

Good afternoon, Peking University alumni. My name is Haiyan Jin. I have a master's degree of 03 in the Department of Information Management. I have 10 years of experience in search engines. I started to join Hongtai Fund for investment in 16 years, mainly responsible for investment in artificial intelligence big data. Here to share with you some of my observations and thoughts on artificial intelligence venture capital.

My first point of view, artificial intelligence + industry as a whole is slow work.

The speed of China Mobile's Internet, from the conception to the end of the team competition, can be won and lost within three years. For example, Didi Chuxing can acquire 200 million users in three years, and has acquired the most important competitive products on the track. It basically determines the pattern of online rides; like Yingke, it also takes less than three. In the year, more than 200 million users, Hong Kong stock IPO; social e-commerce, Pinduoduo spent three years to obtain nearly 300 million paying users, to achieve US stock IPO. The speed of China Mobile Internet is called rapid compression development.

The reason behind it is that after 10 years of Internet stage, entrepreneurs and investors understand the underlying logic, that is, to rapidly scale users, and with users there is a way to monetize; at the same time, users are also educated. So once a team enters the track, more teams will enter, and the capital will stand up quickly and fight quickly.

This is not the case in the AI era, especially in the direction of the combination of AI and industry. I think that in the AI+ industry, it can't be fast. If we still use the speed of the mobile Internet in the past to measure AI+ industry projects, I think this perception is wrong and the expectations are wrong.

The reason behind the slowness is that the AI-powered industry needs to solve several problems:

❶Artificial intelligence application should do data preparation in the direction of combining with the industry. Data is in some industries, such as industries with better digitalization, finance and advertising. These industries are faster, but there are also data cleaning processes and Claim. In many other industries, it is actually a process of data labeling, cleaning, and governance while digitizing. Medical care is a typical example. This is the case in many fields such as agriculture, security, and industry. Therefore, the data preparation stage is actually time-consuming.

❷It takes time to polish algorithms and products. In this field, it needs to prepare long enough time before chip taping. Many AI and industry combinations need to constantly tune algorithms, which itself takes time.

❸ The integration of cross-border talents within the team, the thinking methods and language systems of people who engage in artificial intelligence and technology and industry management are different. They are not teams in the same team, and need a process of integration, learning, and integration.

❹ There is also a process in the cultivation of industry customers. Unlike C-end users, after the Internet stage and the washing of the mobile Internet, today's uncles and aunts have been able to use mobile phones smoothly, and Chinese netizens are relatively mature user groups around the world. But enterprise customers and industry customers today have a process of value recognition and gradual acceptance of product services.

❺ In addition, there will be license admission links in certain areas. For example, like unmanned driving, even L4 needs to get some government licenses to allow you to test run on certain roads; if it is AI medical direction, if you want to land in the hospital as a medical device, you need CFDA certification. This certificate itself needs to be applied for one year or even one and a half years after your product is ready.

Therefore, the combination of various factors has led to the fact that in this direction, it is not like the company in the later stage of the mobile Internet stage. It can quickly grow a large company within two or three years. The business of the big finger here is large, not superficial. The valuation on the market is large. Under such circumstances, some investors who originally looked at the direction of 2C have turned to look at this direction. If they have the mentality that they want to quickly invest in it, they will quickly produce results, which may be disappointing.

The second point of view, in the artificial intelligence venture capital circle, is definitely not the winter of capital. We can look at the data.

This is based on the statistics of business cards. According to the year, the number of established artificial intelligence companies and the number of invested companies were selected. In 19 years, the data as of yesterday was selected. We can see that the amount of financing obtained each year from 2014 to 2018 has been rising. Even in 2018, there were more financing projects obtained than the number of projects established this year, which shows how easy it is to get money for artificial intelligence projects on the market. The number of completed financing in the first six months of 2019 (as of yesterday) seems to have decreased. Compared with the number of established companies, it is also a lot. This shows that the investment in artificial intelligence is, at best, a return to rationality, without excessive hustle and bustle.

We should look at the capital cycle behind the term capital cold winter. The supply and demand relationship reflected by the term capital cycle is the supply and demand relationship between capital and projects. As for the supply side of capital, when I started watching this track in 2016, I also saw fewer foundations on this track. Many funds did not look in this direction. But this year, I see that almost all funds that are still on the market have this investment direction. Although the overall fundraising difficulties have led to a reduction in the number of funds on the market, the proportion of funds that include the direction of artificial intelligence technology in the investment direction has increased significantly. From this dimension, the total funds invested in the track of artificial intelligence today are not necessarily Is down.

At the same time, on the entire artificial intelligence track, we see today that the number of good projects is actually not that much, so we see that it is not so difficult to finance good projects on the market.

Looking at the financing data for the first six months of 2019, we see that this year's financing is characterized by a U-shape, which is easier on both sides and difficult in the middle. The capital effect of the head project is still very clear. In the first half of this year, several companies have received large amounts of hundreds of millions of dollars in financing, such as disregard, horizon, Tucson; the other end is early, angel, Pre-A, In the round A projects, the financing ratio is actually high enough to account for more than 70%; on the contrary, it is the middle segment. For the projects after the round A+, when the investors at this time node look at you at this time today, they must ask about your The commercial landing situation and your income situation last year, because the valuation premiums given in previous years are quite high, if today the commercial landing situation is not ideal, and the valuation does not come down, it will be more difficult to melt. This is a defoaming process.

On the whole, the AI investment circle is definitely not in the cold winter, but is returning to a normal and rational state.

The third point of view, technology-driven innovation opportunities for various industries, is the foundation of capital favor.

I still firmly believe that the technology with artificial intelligence as the core is the new engine of China's economic development in the next decade. Although relatively slow, this is a major trend.

Everyone knows that the general trend is that the dividend period of the Chinese population has run out, and the economic cycle is in the downward period, and it is necessary to use good technology. Technology is the most fundamental and core factor in promoting the development of the entire economy, and policy strongly supports the development of technology.

When the economic situation is good, it is easy for companies to make money, and there is not such a strong motivation for the application of technology. However, during the economic downturn, the entire industry is not easy to live, but it is forcing companies to improve production efficiency and improve market The speed of the response reduces costs.

I voted for a company called Shulan Technology in Hangzhou. After several years of perseverance, this year has ushered in a high-speed growth period. The CEO of Shulan used to be the core backbone of Alibaba’s internal data center. He came out to start a business, positioned as a data infrastructure, and built a data center for the enterprise. This year is really the first year of the data center. More and more industry customers recognize and proactively propose consulting, services and products that require data center.

Briefly introduce the two-sentence data center, which is the infrastructure for data applications. It is an operating system similar to data. It connects all the data owned by a large enterprise at this level. It can exchange data and realize data assetization. When a department or a subsidiary makes data applications on this, the efficiency can become very high. The acceleration of this process is not only related to the process of informatization for more than ten years, but also related to the economic downturn. On the other hand, enterprises are more motivated to seek change, and use data to drive the optimization of business models and business models.

For the next ten years, for startups in the technology field, earning money is slow, but the money earned is very valuable and has a very high threshold. The process of earning money in enterprises and industries is the process of using technology to drive innovation in various industries and improve the efficiency of the industry. It has a multiplier effect on China's overall economic development.

The fourth point of view, multiple industries are accelerating the commercialization of landing, 2019 will see some commercial results.

In the commercialization stage of AI, 2019 should be a year that will be remembered. The field of AI technology landing, the earliest and fastest commercialization, security, and many of the top projects have good revenues this year. If the players who cannot earn more than 100 million this year may not be competitive enough. In addition, according to what I have observed, there are several industries where the revenue will really start this year. Some have been investing before, and the commercialization of the first year of this year generates revenue, such as AI medical imaging.

I am very optimistic about the application of AI in the medical field. I think there is a chance in this direction to produce great companies such as Internet BAT and mobile Internet TMD. The fundamental contradiction in the medical field is the imbalance between supply and demand. With the changes in the age structure of China’s population, the proportion of elderly people is getting higher and higher, and the disposable income is constantly rising, there is more and more demand. At least 10 years of clinical practice experience development model has not changed, this contradiction will become stronger and stronger.

We know that during the period of the Internet and mobile Internet, there were no very large and very powerful companies that applied technology in the medical direction. Although there are actually many teams exploring this direction, it is because the technical nature of the Internet and mobile Internet technology is in By connecting to eliminate the information asymmetry between the supply and demand sides to bring about the improvement of industrial efficiency, the problem of insufficient quality supply in the medical direction cannot be solved. And the characteristics of AI technology, as long as there are enough "correct answers", machine learning can take advantage of his "learning fast" and quickly train the skills of previous doctors.

AI images are relatively easy to cut in. This track is favored by many entrepreneurs and investors, so dozens of projects have been financed in the past two or three years. We chose to vote for Shenrui Medical in early 17th, and now belongs to the first echelon of the track. At the earliest on this track, everyone looks at the technology, the team, the process indicators to see the effect of the trained products, how many hospitals are using it, and whether the products combined with medical devices have obtained the CFDA type 2 certificate and type 3 certificate. This year has reached a stage where a certain income scale can be started.

In the direction of commercial landing, we see several wave of technology of informationization, automation, and intelligence are superimposed together. For example, Brother Chen mentioned the industry drone Yunsheng Intelligent they voted for. We invested in the next round. Because it uses several technologies together at the same time, the scope of application has greatly expanded. Everyone knows that consumer-grade drones, such as DJI aerial cameras, have a very weak battery life. The 40-minute battery life can only be used as entertainment and cannot be a production tool because the efficiency is not enough. So this company solved this problem through hardware automation technology. He made a drone field, which is a cubic iron box about 1.5 meters long, wide and high. When the plane flew to no power, he ran back to himself. Replacing the battery is very fast. It can be replaced in 3 minutes, so there is no need to charge it.

At the same time, the company's artificial intelligence video data analysis ability is very strong. The data collected by the drone can be transmitted to the cloud or calculated and analyzed directly on the end. In this way, the drone is no longer just a hardware, but a flying data collector. It can also collect a variety of data, such as ordinary cameras to collect data during the day, and the drone will fly back to the airport when it is almost unclear in the evening. I automatically changed a night vision device, and then ran out to continue to record video collection data. When I needed to collect air components, I ran back to change a gas detection sensor. Therefore, the combination of multiple technologies has greatly expanded the scope of its application. The industry can be expanded to multiple industries, including inspections in the power grid, smart cities, etc.

Commercialization has landed, and this year saw a very unique phenomenon. In the early stages of intelligent projects, internationalization was considered in parallel. In the history of Internet development, only overseas companies, such as American companies, consider global users when designing their products. China basically only considers serving Chinese users. Now our projects also have this, this is a trend, China began to export its own technologies, talents and products to other countries.

Looking back at a few points I just said: the artificial intelligence + industry is actually slow work, it will not explode very quickly; today this time node capital is actually very friendly to the entire industry of artificial intelligence, today is definitely not the winter of capital, just remove it Some bubbles should be gone; I am still very optimistic about the opportunities of technology-driven innovation for various industries; in 2019, we will see that there will be commercial improvement on multiple tracks. To sum up, the core is not to underestimate the long-term nor to overestimate the short-term. Both entrepreneurs and funds need patience.

Hongtai has deployed about 30 AI companies in the past 3 years, from angel to B round, including the underlying computing power and algorithms, including the middle-tier big data infrastructure company, and the upper-tier Industry applications, including security, smart finance, advanced manufacturing, marketing, medical, industrial, and new retail projects. 70% of our projects have received the financing of the last round or more, and most companies are still relatively stable. Next, we will continue to invest in this track to accompany entrepreneurs who use technology to change the world.

We judge whether a project should be invested, mainly from the three dimensions of situation, things and people.

Early projects will value people very much. We define whether the founders in the direction of AI entrepreneurship are good enough:

First of all, the founders must have good technical Insights. This does not necessarily mean that the founders must be the most technically talented in the team. It is also possible that the most technically powerful is the CTO, but we believe that good founders must have technical Insights, know The future development trend of technology, where is the market and where is the value;

Second, look at his talent appeal. When the founder came out to start a business, there were no good partners willing to follow. We basically did not vote;

Third, the founders have a relatively high pattern, hoping not to do a small business but to have a long-term and valuable thing;

The last point is that there is entrepreneurship and enough patience in the face of difficulties. We foresee that entrepreneurship in this direction is more difficult and needs to experience 1981 difficulties.

This is human. In addition, regarding the potential, we have been doing various macro analysis and industry research internally, so when a project comes, we can easily make judgments instead of doing analysis every time. The choice of whether to do the homeopathic thing actually reflects the founder's ability.

Growing-up projects will look at the situation and things in specific industries, and will judge whether the time of application of the technology has arrived.

I have shared before, judging the turning point of technology landing, is to see whether the ability of technology has exceeded the basic ability of the average person. Judge whether the product enters the inflection point that can be promoted, single point to assess whether the machine cost is lower than the labor cost. If it is a systematic gain, evaluate the future impact on the industry. Of course, whether the CEO is excellent, we will continue to look at this. In addition, at this stage, we will also compare the care valuation is reasonable, after all, there is currently a bubble. This is our basic judgment of several dimensions of technology projects, for the reference of entrepreneurs present.

thank you all!