AI+ Venture Capital Special Session: Changes and Constants in AI Investment丨CCF-GAIR 2017

From left to right: Gu Minman, Managing Director of ZhenFund, Han Yan, Founding Partner of Lightspeed China Fund, Yue Bin, Founding Partner of Gaorong Capital, Xu Lixin, Managing Partner of Bojiang Capital, Wang Huadong, Partner of Matrix Partners, and Gao Tianyao, Vice President of Lenovo Star Investment

On the afternoon of July 8, the second CCF-GAIR 2017 Global Artificial Intelligence and Robotics Summit hosted by Leiphone.com continued in Shenzhen.

In the AI+venture capital special session, hosted by Gu Minman, managing director of ZhenFund, Han Yan, founding partner of Lightspeed China Fund, Yue Bin, founding partner of Gaorong Capital, Xu Lixin, managing partner of Bojiang Capital, Wang Huadong, partner of Matrix Partners, and Gao Tianyao, vice president of Lenovo Star Investment, discussed the theme of "Changes and Constants in AI Investment". These investors have deep investment experience in the field of artificial intelligence. This time, they also talked about their views on this field, including what kind of artificial intelligence companies they are optimistic about, how artificial intelligence companies can improve their commercialization capabilities, and pointed out the minefields for artificial intelligence companies.

Gu Minman: The field of AI can be further subdivided, whether you are looking at it from the bottom up or from a vertical business perspective. Our fund, whether it is personal investment or not, but there is an AI startup company that we are very optimistic about. Let's talk about one, because whether you are optimistic about it or have already invested, you must have brought up corresponding views on this field. We would like to ask everyone to talk about your views on this project and your views on this field from a project perspective.

Han Yan: If we talk about one company, it is called eCheng Technology, which is a company doing big data in the HR field. What is interesting is that we invested in this company before the concept of AI became popular. The entrepreneur contacted us in 2011 and said that he had a dream. He wanted to use a tool for all HR managers. Because HR managers manage a lot of employees every day and have to submit a lot of resumes, the efficiency is very low. He wanted to improve their efficiency through big data and artificial intelligence.

So they had this simple idea in 2011, and came back to us in 2014 to say that they had made breakthroughs in both data and algorithms, so we invested in this company at the end of 2014. Now this company is very low-key. It has been established for more than two years, but its SaaS revenue has reached nearly 100 million, and the entire company is profitable. AI is now widely discussed, but when we invested in this company, we didn't think so much. We were optimistic about this company because it wanted to provide a product provided by SaaS. At the same time, they have accumulated data for a long time. They know where the data of HR's daily work is, and they have spent a lot of energy and money to attract many top talents in data analysis in China.

The point I want to highlight in this case is that Lightspeed has been spending a lot of time researching in the field of artificial intelligence. In my opinion, only two types of companies can impress us. One is that the product technology can truly improve the efficiency of the industry in which it is implemented; the second is that through its data and algorithms, the services or products it provides can be three to five times more useful. Therefore, the reason why this company can grow so fast and its SaaS revenue has increased so quickly, I think the biggest reason is that it has opened up the difference in efficiency and product experience, so that an HR who usually reads resumes may read 200 resumes at work can tell him through his technology that you have read 5 resumes, and the conversion rate of these 5 resumes is particularly high, and the matching degree with the talents you need is very high. This is a company we are very optimistic about, and it is also our experience of AI.

Yue Bin: Among the AI ​​companies we have invested in, I can see that there are no less than three companies with a market value of more than one billion US dollars. If I talk about one company in particular, it is Yitu, which has been in business for a relatively long time and can be seen in the media. ZhenFund also participated in the early stage. When we participated in this company, people may not have recognized them so much. The valuation was not cheap at that time. The company had only a dozen people and worked in a private house in the suburbs of Shanghai. One of the two founders has been in the United States for many years, and previously worked in places such as UCLA and MIT. The last place he worked was in Yann LeCun's laboratory. The other founder was his classmate and the first technical director of Alibaba Cloud. He built Alibaba Cloud from scratch. These two founders gave up a good opportunity at that time and came back to start a business very early. Today, in the field of vision, they have made many breakthroughs, and I believe these breakthroughs are not seen in these industry conferences or papers.

In addition, they also do a lot of things in the medical field. Today, many people are talking about medical AI, but I have hardly seen other companies that have real applications, a large amount of data, and can really help doctors, except for this company. Last year, CCTV made a documentary for them. In the recent financing process, this company is also a company at the $1 billion level. I will briefly introduce this company. In fact, I want to say that such companies are definitely not isolated cases.

Xu Lixin: We like to invest in entrepreneurs who are good at telling stories, or they can also talk about their dreams. CloudMinds Technology is the first cloud platform. For example, German robots pressed workers on the production line. Last year, a robot hit the glass at the High-Tech Fair, injuring a person who had to be taken to the hospital. In fact, everyone is worried about whether robots will hurt humans in the future or whether they will be controlled. This company built a cloud platform, which is invisible and intangible to ordinary hackers. It is an independent platform.

The founder of this platform is the president of China Mobile Research Institute. They believe that the financial sector is rich and will be the first to use it. They are shareholders of many banks, so they joined a group of shareholders to do angel investment, including Qualcomm, because cloud computing requires communication, and Taiwan's Foxconn, Terry Gou, who was the angel investor. We raised 100 million US dollars from Terry Gou. It is a very low-key company that does big things. Now they not only have dreams, but are also implementing them. They use secure networks for secure communications. For example, Apple phones have vulnerabilities, and some sensitive departments and government officials can use secure phones for communication. Some practical applications are also being promoted. I think these are companies that have both lofty dreams and can be implemented, so I hope to discover more of these companies.

Wang Huadong: Let me share a case. This project was the one we invested in last year, first in the angel round and then in the A round. It's called Moran Cognition. From a Jingwei perspective, we hope to invest in companies that have their own accumulation in the field of technology and can develop companies with certain technological leadership. For example, the human-computer dialogue system launched by Moran Cognition last year is one of the few engines that can achieve continuous dialogue and at the same time has very good commercialization. For example, it has many customers in cars, so it has certain technical barriers and very good commercialization. These are the two most important factors we make judgments about.

Meng Xing: I think we are all equal, and it is unfair to criticize any one company. In each field, we will invest in the best one, such as Mobileye, which does very well in visual perception in the field of autonomous driving. The limit of investing in artificial intelligence lies in our entire imagination of cognition and business judgment. The best company is always the next one. Known business scenarios are not as good as discovering the best one in the next one, not the ones we have already discovered. In hot industries, every company we invest in is very excellent, but the most exciting one is always the next one.

I'm talking about an early-stage company. I don't need to praise other more mature companies. We invested in a company called Owlii, which is a very early team from Tsinghua University. It is engaged in large-scale 3D reconstruction, and it is dynamic. If we use the only one I just mentioned, it is the only company in China that can do dynamic real-time multi-scene 3D reconstruction in multi-person scenes. It is also on the list of new intelligent manufacturing today, which is quite rare. The benchmark is holo, that is, I stand here, and all my movements in the room are reconstructed in real time in front of another person. He can see it through AR glasses, and all coordinate information is reconstructed in real time. Thinking further, their benchmark is not the visual problem, but the traffic. If I really don't know whether you are in front of me, your movements are realistic enough, and your angles are realistic enough, so that I can't tell whether you are really in front of me physically, it can replace the traffic and not face to face with you.

Gao Tianyao: I totally agree with what Mr. Meng said. The host has dug a hole, and it is not a small one. I believe that many institutions have invested in companies in many related fields, so it is difficult for me to say which one I am most proud of. We invested in Megvii and AISpex because we valued their imagination. Computer vision was not very popular with capital or the market at that time, so I thought about it for a long time. I will give an example of a project that I recently invested in, a relatively interesting project, which uses AGV to solve the parking problem.

In fact, parking is a big pain point now. There are many companies that are engaged in parking learning, and there are many equipment and software for finding cars, but there is a gap in the middle, that is, you still have to walk to your car or physically find your car. The company we invested in uses AGV carts. You can park your car in a fixed position, and the AGV can directly lift your car and transport it away. When you come back, there may be several carts running in the garage. Traditional garages are also doing related things. It is the DJI team. What I want to express is that we think we still need to see some imagination space for the future of this matter, just like we invested in LP a few years ago. Of course, this may be hardware-oriented, but if the efficiency of this node is improved, the entire car factory can park more densely, saving your car's waiting time. This space is still relatively large, so I would like to share a more interesting project.

Gu Minman: Regarding the field of AI, I think there is a triangle. These triangles are the commercialization that everyone mentioned just now. In the overall AI, whether it is technical capabilities or algorithms, the ability to obtain data, and whether it is a vertical field or a general field, the ability to achieve commercialization is often a very important triangle for the success of a company. We have seen that in the companies mentioned by several guests just now, these three corners are often very strong, but the other two corners are also slowly being supplemented. When investing in the early stage, we cannot ask for a perfect company in front of you, and it happens to be the price you invested in. My next question is about trade-offs, or I want to help the people present understand preferences more. When facing the early team, how do we balance in this big triangle? Or priority trade-offs.

Gao Tianyao: There is another dimension here, which is the investment stage. Mr. Gu mentioned earlier that Zhenge is mainly an angel investor, so it is difficult to generate income, basically impossible. So when we look at projects in the early stages, we need you to have particularly outstanding advantages in terms of judgment. I think of the very good point the host just mentioned, that you can use a hammer or hammer a nail with a hammer. When we look at AI or AI-related fields, we also hope to find people who make hammers or people who can use hammers.

The biggest imagination of this wave of AI is that you still don’t know what can be solved. We all mentioned commercial implementation and efficiency improvement, which is definitely something that everyone is pursuing. However, what is not known is that at this point in time, you don’t know the evolution of technology or the emergence of new entrepreneurial talents. It can solve more problems in the industry. This is actually a big opportunity for us. Back to the host’s question, we pay more attention to two types of people. One is that you can make a hammer, and the other is that you can use a hammer. Otherwise, you have very good algorithms, including the one mentioned by Mr. Liu just now. Although the barriers to algorithms will decrease in the long run, there is still scarcity of algorithms at present, including the ability to combine software and hardware. This is that you can make a hammer.

Then you will use this hammer, and I hope this part can better reflect the commercial implementation. So we don’t have a special requirement that you are not good at something, and we will not invest if you are not good at something. What we value is what aspects you are particularly strong in, and we are willing to support you in this regard.

Meng Xing: Let me talk about the conclusion first. The conclusion of this project is that we don't make any trade-offs, we don't give up any point, we want everything, how do you understand it? Continuing with the topic just now, why are you excited to see investment projects that you haven't seen before, data volume, algorithms, commercialization, for example, when it comes to data volume, massive data is a prerequisite, everyone knows it, but there are too many ideas and algorithms to solve this problem, can I use very small data to achieve the same good way, these projects have a huge plus for us, because it is subverting our inherent model, thinking that big data must be used to solve problems. The second is that we don't have such a way. Can we find a way to say in commercialization, if I can't have such data for this kind of model, can I find it in the business model, can my product definition enable me to absorb more data, even if my business ability is not good, or I don't have natural data, or in a field that no one has touched, no one competes with me for the data source, I become a data developer and data analyst myself. Two things are superimposed on each other.

I think it may sound a bit vague, but this is how we think about it. We don't really want you to have commercialization or data. This triangle does not exist. We have a four-corner or five-corner. The three things that Mr. Gu just mentioned are innate based on our own technology and algorithms. Most of them may not be easy to change during the introduction process, so this part is inevitable as an initial team or early condition. There are ways to touch the data, but whether you change an algorithm, use less data, or look in the right direction, it must be solved.

Today we don’t focus too much on commercialization, but on the prospects of commercialization, not on how much money you have already made. If you can do the first two things well, but the ceiling is very low or it is unlikely to have a huge breakthrough, then naturally we cannot make a choice in this matter and cannot invest in the project.

(Gu Minman: You want everything, but you don’t choose anything. Then can you tell me about the stage of investment? Because Mr. Gao and I are very practical. We can’t have everything in the early stage. I know Shunwei is a company that covers all stages and is more comfortable in the field of AI. At what stage can you really enter the market?)

Shunwei invests mostly in Series A and B rounds. I believe entrepreneurs or teams must have their disadvantages, but they have the ability to solve their disadvantages in different ways, which is what I value most.

Wang Huadong: Jingwei mainly invests in rounds A and B. From our perspective, let me first tell you the conclusion. The most critical thing for us to look at this kind of project is whether he has thought through the usage scenarios of this thing. In the usage scenarios, we will pay attention to the composition of the team. You can understand it as paying attention to the usage scenarios is paying attention to the future business potential. If something has business potential, it needs a team to make it. At this time, we don’t want to just invest in a pure technology team. We hope that someone in this team can do a very good job in technology and someone can do a very good job in operations, that is, a very good product, so this is what we pay more attention to.

Xu Lixin: We are mainly focusing on the A round in terms of project stage, that is, it has its own uniqueness, a bit like Tao and art. The current outlet is AI. This is a dojo that has been built. Everyone has their own tricks. You do face recognition, he does voice recognition, he does mobile devices and algorithms, and each has his own strengths. It depends on whether this art is a dragon-slaying art or a narrow chicken. For example, it may take a long time to achieve 100% image recognition. If you can't find a practical direction, you may need to investigate. If image recognition or voice recognition reaches 99%, it is still a little far from perfection, but you can find my application scenario, such as in the security or financial fields. I think it is very down-to-earth. This art is not a dragon-slaying art, it is useful and in line with my investment direction.

Yue Bin: In the past, when investing in AI companies, I was lucky enough to invest in companies that were very commercially sound and the best in the world in terms of technology. Every time I had the opportunity to come across such a company, even if I only came across one a year, it would make me laugh in my dreams. But I was very lucky to have come across some of them. Such companies are indeed rare. There is another characteristic of investment in the AI ​​field. The slightly better companies actually have very high steps. The requirements for VCs in this field are very high. Almost all VCs are very difficult. With such a high valuation and so few good teams, the chances of launching a mobile phone in a year are extremely limited. Therefore, in this case, to have a good return, you must be very accurate.

Back to these points, we also need to look at specific fields. For example, if we use deep learning algorithms for training and hope to get very good accuracy and results, the amount of data is indeed very important in this case. Or if there is no data today, the way to obtain data tomorrow is also very important. But if in other fields, for example, everyone is doing Go, but in fact, everyone's Go data is the original 30 million games. Tencent's Jueyi and AlphaGo may be very different. Even under the same conditions, different sub-fields have different requirements for these. If we combine these points together, what is the most important? Cognition is the most important. There are many industry conferences and AI conferences now, and there are many people who come out to do PR. There are all kinds of opinions. Are they the most valuable opinions? There are very few most valuable opinions among them. Where are the people with these valuable opinions?

I think it is possible that he is not in a conference room like ours, and he may still be trying to test his next result in front of the computer. If we have the opportunity to meet such people who really have a deep understanding of the cutting-edge development of this industry, I think everyone can help with commercialization. If the amount of data is not enough, everyone can also work together to find a solution. Back to the points just mentioned, if you summarize it, the requirement for VC today is that if you really want to invest in the best companies in the field of AI and make a lot of money, on the one hand, you have to invest in the best companies, and at the same time, you have to ensure that each company you invest in does not make others think that this company is actually very ordinary. Both conditions must be met.

Han Yan: I was deeply touched by what Yue Bin said, because Yue Bin and I are very familiar with each other. It seems that everyone looks at so many cases every year and is looking for disruptive cases. From the perspective of investors, we feel that the entrepreneurs who can impress us the most may not necessarily understand the industry as Yue Bin said, that is, experts in a certain industry. He may be an outsider, but he has deep insights into a certain industry, or has some crazy ideas. At the same time, he has a very good sense in business. Such people are often very impressive to standard VCs. Therefore, I think their ability to perceive the future, their imagination, and their business sense are invaluable. We will definitely seize such a team.

Looking back, there are not so many opportunities to build a big company every year, so entrepreneurs should not think about becoming the next Didi or becoming the next hot spot all day long. I think it is still necessary to analyze the team's own strengths. From my observation, although AI is so popular, investors have been filling in the team in the past two or three years. This team has a lot of accumulation in this industry. Maybe his accumulation is at the algorithm level, maybe his accumulation is at the data volume level. So investors are not gods and cannot predict the future. What are they looking for? Find the team's strengths. So from the perspective of entrepreneurs, don't follow the trend, don't read PR articles, and see what your strengths are, is it technology?

If the technology is OK and valuable, and the technology has highlights, then sell the company, which is also a successful startup. So from the perspective of entrepreneurship, you must pay more attention to your strengths and make full use of them. At the same time, don't forget that if you want to make this thing bigger, you also need to have a very deep understanding of business. For example, I recently read about medical AI. The company that impressed me the most was not from a medical company, but it was able to dig out the experts who understand medicine best in the medical industry, learn their knowledge, and absorb those experts into its own team. I feel that this kind of person is crazy from a medical perspective. This is my view .

Gu Minman: Thank you, guests. The reason why I asked the question just now is that two months ago, ZhenFeng brought a group of top AI companies in China to the United States and Silicon Valley to communicate with the world's top AI laboratory. During that communication, we found that the founders were frank with each other, and everyone had a lot of confusion. No matter what the scale of the company is, no matter what level of international peers they face, how can AI technology achieve greater commercial breakthroughs, including for a startup company, how to get a really large amount of data to make AI technology play its due value. So on this issue, I think it is more of investors who bring their views and the development path of the projects they invest in to everyone very selflessly share and observe.

Next is the last question. An investor made a very interesting observation before. He said that if starting a business is like chasing a girl, you should not chase the pretty one, but the best friend of the prettiest girl. How to say this? It leads to the fact that startups should view themselves as a very realistic starting point from entering this market, but there is inevitably competition between them and the existing large companies in this market, and there are definitely shortcomings that need to be made up in terms of data and technology. So the last question is, if you want to point out a minefield to companies in the field of AI entrepreneurship and suggest that they should not do it, what would it be? Because time is running out, I will go first. I suggest that everyone stop doing autonomous driving. I guess there are still many minefields, but I think everyone doesn't need to do this. Everyone is very practical and briefly point out a clear path.

Gao Tianyao: Don’t make platform-level things, especially for startups in the AI ​​field. Of course, the platform is very valuable, but it is difficult. We have seen it. Don’t expect to make a very general platform-level thing. The field of AI is very difficult.

Meng Xing: I think you shouldn’t start a business with the mentality of selling technology. Whether it’s your product or your final exit method, China does not have such a soil.

Wang Huadong: I agree with Meng Xing’s point of view. Don’t think that just because your algorithm is so great you can use it to start a company. Algorithms have cycles.

Xu Lixin: Artificial intelligence is just in its infancy, and personally I’m still not sure where the holes are. But you shouldn’t do things that you are not familiar with. You don’t know where the pitfalls are and where the pitfalls are. You may run into a pit yourself, so the team must have familiar fields to do the work, implement the technology in the corresponding scenarios, and give yourself confidence, otherwise you will get very frustrated.

Yue Bin: I have done several things this year. Last year, I invested in a company that was still in its early stages. I liked it very much. At the beginning of this year, he told me that he was considering a new round of financing. After he told me that the financing was worth it, I said that I didn’t need to go out to raise funds. I invested directly. What are the characteristics of such teams? The value of their talents can be said to be extremely scarce in this field. They have a very deep understanding of the development of the industry and what they do, and there are too few such people.

What advice can I give you? The current domestic media environment, including Leifeng.com and self-media, is very good. If there is any new progress in China, it is basically guaranteed to be seen the next day. If we consider doing something in a field, we don’t have our own cognition and understanding of the progress of the entire industry. You don’t even understand what others are doing and how they are thinking about doing this. I think the biggest possibility is that you don’t need to do it. Because today in the AI ​​industry, just like smart hardware and many other fields are very popular a few years ago, how many smart hardware companies are really making money now? How many VCs have made money from smart hardware? Very few.

Which companies will be successful in the end? Why them? What if I have done it and the big companies have done it? I think people who really do investment will not think about these questions, including China's BAT. They have great respect for big companies, but they do see what path to take to make it. If we do not stand at such a high level when we do things, it is very likely that we really don't need to do it.

Han Yan: The previous speakers have made very good points. I agree that you should never enter a field you are not familiar with. Startups in the AI ​​field should never imagine that Didi will invest so much money in it, so you must be prepared for five to ten years and fight a patient battle. If you can foresee that this startup will take enough time, patience, and focus, then you must have a strong perception and cognition of this field. Don't touch other unfamiliar fields. Don't start a business just for the sake of starting a business. In the past, O2O left two or three companies, e-commerce left two or three companies, and AI may also leave some companies. This is my point of view.

For more wonderful speeches by big names, please continue to follow Leifeng.com’s follow-up reports.