In recent years, artificial intelligence companies focusing on machine vision, natural language processing, and machine deep learning have been founded. Internet giants at home and abroad have also entered this track.
Lingzhi Technology is also a member of this competition. Since its inception, this company has focused on researching artificial intelligence machine vision solutions. As early as 1998, its core team began to study artificial intelligence, and successively achieved excellent results in the fields of face recognition, object recognition, emotion recognition, and gesture recognition. In 2015, the Lingzhi Technology team shifted algorithm research to the direction of edge learning algorithms, and formally established the company the following year. Their goal is to use artificial intelligence to build the future world and allow artificial intelligence to enter all aspects of people’s lives.
But on the other hand, today’s artificial intelligence industry is already a fiercely competitive red sea. Many well-developed start-ups have emerged in the subdivisions of machine vision, natural language recognition, intelligent robots, and intelligent interaction. At the same time, entrepreneurs have to face the impact from BAT and even larger international companies. How to break through the Red Sea is a question that every entrepreneur is thinking about.
Recently, Zhang Kejun, CEO of Lingzhi Technology, was a guest on Moike Planet, sharing his experience with us around the current industry and technical issues.
Lingzhi Technology is mainly focused on researching machine vision artificial intelligence solutions. However, when looking at your application scenarios, I found that there are smart door locks, smart home robots, driver detection systems, etc. The application directions seem to be very many, Lingzhi How did technology choose, and why so many application directions?
Zhang Kejun: This is actually related to the positioning of our company. It seems that we have many application directions, but in fact it is not the case. Our company is positioned as an artificial intelligence machine vision solution provider, and our core product is an efficient artificial intelligence machine vision solution based on edge learning algorithms. This solution itself can be applied in many fields, including door locks, robots, fatigue driving, etc., but our core point has always been on how to provide efficient solutions.
Between different application scenarios, such as smart door locks, smart home robots, driver detection systems, etc., will the requirements for machine vision technology be different? How does Lingzhi Technology adapt to different scenarios?
Zhang Kejun: There are differences and similarities. The same point is that they are all based on artificial intelligence vision processing. The difference is that there will be some subtle adjustments in different applications. Lingzhi Technology captures these similarities and makes subtle adjustments to the differences, with no more than 10% of changes in each project. Our algorithm can realize the application of face recognition, emotion recognition, gesture recognition, item recognition, etc., and then only need to adjust the algorithm in the corresponding scene.
In previous interviews, you said that “marginal algorithm will be the mainstream trend of AI algorithm in the future”. Can you tell us why?
Zhang Kejun: edge algorithm is an algorithm that only needs a small number of samples to train the machine. Its feature is that it does not need to upload the data to the cloud, and can complete the recognition only locally. Our company believes that if artificial intelligence wants to be implemented in the future and widely applied to people’s daily life, edge algorithm will be a great trend. There are two reasons: first, its real-time performance is very good; second, it can better protect personal privacy.
In terms of real-time, for example, if you want to open the intelligent door lock, if you still need to upload the data to the cloud for identification, then in case the network is not good, can’t you open the door. Driver detection is the same, if you are in the underground garage or tunnel, it is easy to disconnect the network, so you can not guarantee the safety of driving. Even if 5g or higher technology appears in the future, this kind of wireless transmission will have various problems.
In terms of privacy, when we swipe ID cards at airports and railway stations, the data will be uploaded to the cloud for identification, but we don’t need to worry about privacy issues, because these data are handed over to the public security department. But when using smart door locks or home monitoring systems, it is not appropriate to send data to the cloud. Because that is equivalent to having a pair of eyes staring at your every move at home, or stealing your data, you can open your door from the far end.
These problems are common both at home and abroad, so the edge algorithm has become more and more popular recently, and has been paid more and more attention by the industry.
Now there are many enterprises doing machine vision solutions in China. What is the competitive advantage of Lingzhi technology? How do you persuade customers to use your products? Is there any specific data?
Zhang Kejun: our advantages are reflected in many aspects. The first is the comprehensiveness of our technology. Lingzhi technology has been doing artificial intelligence research since 1998, so it involves a lot of applications, including medical, engineering, civil and entertainment identification. We have designs for all aspects. Through the accumulation of technology in these aspects, we will have a wider range of ideas than others when making new plans.
Secondly, we focus on the edge algorithm, which provides a lot of convenience in real-time and privacy. And the scheme we offer is more favorable in price, which is also an advantage for us. For example, we had a client in Zhejiang Province who was going to make a face recognition attendance system for kindergartens. This kindergarten, including children and teachers, is about 500 people. He has also found other enterprises before. Some of these schemes give him a starting price of 300000 yuan, and the highest one is more than one million yuan, which is a large amount of funds for a kindergarten. But with our edge algorithm, it only needs 80000.
Another point is the real-time performance of the edge algorithm, that is, there is no need for networking. We have a joint project with the Ministry of public security to monitor the grassland in Inner Mongolia. As we all know, Inner Mongolia grassland is boundless, and the signal is not good. What if there is a fugitive here? At this time, it is convenient to have an off-line version of artificial intelligence.
In addition, lingzhi technology also has great advantages in terms of professionalism. Our team now has more than 50 full-time employees, all of whom are from the world’s top 500. We always believe that quick response to customers is the way to survive and the foundation of the company. When a customer wants to have an AI solution, we can usually propose a very good solution to the customer within two weeks, and sometimes we can directly demonstrate it. This rapid response, I think, is also a great competitive advantage.
What is the core competitive point of machine vision? How to create trade barriers in this industry?
Zhang Kejun: I think it is better to say what is the core competition point of artificial intelligence. After years of investigation and research, we believe that it should belong to the “Ai + industry”. That is to say, artificial intelligence itself has not formed an industry, and it must be combined with the existing industry to burst out vitality. In this way, two things are very important.
The first is the ability of enterprise algorithms. The development of artificial intelligence is very fast. Some fields may be innovated in one or two years, but artificial intelligence will be innovated in two or three months. Algorithm update is very fast, it is difficult for an enterprise to say that its artificial intelligence has always been at the forefront.
Second, the engineering capability of the enterprise. Different industries have different demands for artificial intelligence. Taking a lock factory as an example, how can a lock factory use artificial intelligence? This middle span is very large, it includes algorithm, scene optimization, electronics, and the mechanical part of the whole module. For a lock factory, the interface it can provide is only the mechanical part, so as an artificial intelligence industry, how should it be implemented? Lingzhi technology defines itself as a service provider. Since it is a service provider, how to provide customers with useful services is very important.
So I think if you want to build industry barriers, you need to do these two things well, and then you need to put your attitude right. The first two points require enterprises to do a good job in internal basic skills, constantly update and optimize their core algorithms, and at the same time, they should also think about how to interface with the industry. As for the correct mentality, it is necessary to have a service-oriented mentality, so that enterprises can do better and better, and barriers will become stronger and stronger.
Do you use 3D vision technology or 2D vision technology for face recognition lock? Which technology is more mainstream in this field? Why?
Zhang Kejun: we use 3D vision technology, and now it is the mainstream of 3D vision technology. The reason is that the door lock has many advantages in addition to recognition, such as living detection, and the adaptation of angle and light. In these aspects, 3D vision is better than 2D vision.
As for the smart home robot, it has many functions, such as security guarantee, housekeeper service, and can help take care of the elderly. What is the biggest difficulty in studying this kind of robot?
Zhang Kejun: the biggest difficulty is how to do a good job in human-computer interaction. We think that in the future, smart home robots will definitely enter the home scene, so the key point is the processing of machine vision. Our processing basically includes object recognition, emotion recognition, face recognition, human intention recognition and so on. These kinds of recognition have been realized in our robot, all of which are to strengthen the ability of robot and human interaction.
Of course, it is not enough to make products only with good technology, and the cost performance is also crucial. Now many robots use lidar, but the cost of lidar is very high. The cheapest one is several thousand yuan, and the expensive one is tens of thousands. But machine vision is not the same, as long as a few hundred yuan, so we think that in the future, intelligent robot based on machine vision should be a great development direction in the future.
How about the landing or popularity of Lingzhi technology’s smart home robots?
Zhang Kejun: at present, what we put out on the official website is our top-level version to show how far Lingzhi technology has achieved. However, when we are really landing and popularizing, it is generally a simplified version of this version, or a version for some specific application scenarios. For example, for home scenarios, we will provide some simplified versions to reduce costs.
Now domestic AI enterprises are competing for open source. How do you view this phenomenon? Is Lingzhi technology also involved in this aspect?
Zhang Kejun: the essence of open source is to let more people use the technology, and then continue to promote the development of technology. But to promote the landing and development of a technology, it is not necessary to adopt the open source method. If you can provide a stable and appropriate solution, it is also a way to achieve the same goal.
For Lingzhi technology, we are more concerned about how to reduce the cost of machine vision module. Our goal is to reduce the cost of machine vision module from algorithm to camera to $3 this year.
We have interviewed some investors before, and they have such a view: many Chinese AI enterprise teams are returnees, but in fact, there are only a few universities in the world that can cultivate AI professional doctors and talents, and most of the talents have been “ordered in advance” by international enterprises such as Google and Microsoft before graduation, so now some Chinese AI companies have been “ordered” in advance Start up companies do very little in terms of technology. What’s your opinion on this?
Zhang Kejun: artificial intelligence is a talent intensive industry, which is beyond doubt. However, talents are not born to be talents. It does not mean that you are talents after reading several books.
There is an old Chinese saying that the times make heroes. It is far more important for a person to meet a platform where he can give full play to his talents, and let him keep trial and error and iteration than to read more books or learn a major.
However, there are a lot of doctoral students coming back from the United States, or we will find that there are a lot of doctoral students coming back from the United States. The most important thing is to see your practical experience and your understanding of technology.
And artificial intelligence will certainly develop to all aspects of human life in the future, but now Google, Microsoft and other big companies have done all aspects well? For now, that’s not the case. There are still many blank areas, and the future is unpredictable. Even large companies will have many deficiencies, so there are still many opportunities.
What is the development plan of Lingzhi technology in 2019, or other new products to be launched, can you disclose it?
Zhang Kejun: in 2019, our goal is to continue to promote artificial intelligence in different directions, so that it can be better implemented and applied to more fields. It is also our dream to build a better world with artificial intelligence. According to the situation in 2018, artificial intelligence has been recognized by many people, and the degree of recognition is getting higher and higher, which is a good phenomenon.
For Lingzhi technology, we need to be more and more sophisticated in the interface part and make our existing technology better and better. I think this is the point we need to break through in 2019. In terms of products, we will apply artificial intelligence modules to the field of medical robots, such as disinfection robots, handling robots, etc.; we will also step into the field of automobiles, such as the fatigue driving monitoring that we are doing now, which we will make great strides forward this year.