In 2020, Shen Xiangyang established the IDEA Guangdong-Hong Kong-Macao Greater Bay Area Digital Economy Research Institute in Shenzhen, opening a new stage of life. He hopes to use the new mechanism to do something to promote the implementation of technology. Therefore, the positioning of this international innovative research institution is: to promote the development of AI technology, based on social needs, develop disruptive innovative technologies and give back to the society, so that more people can benefit from the development of the digital economy.

On November 22, 2022, the second IDEA conference was held. The main agenda on the first morning was the “IDEA Technology and Product Conference” presented by Shen Xiangyang. In the 2-hour press conference, Shen Xiangyang introduced the various achievements of IDEA in the past year, and the thinking behind each achievement, which involves both the current hot AIGC and the transformation of research methods… In addition In addition, from the perspective of the innovation-capital paradigm, he sorted out the four stages of technological innovation from the laboratory to the market, and IDEA’s positioning in it, and how to do it.

In the end, he concluded that IDEA has been looking for its own innovation “sweet zone”, that is, the process from mission-driven to risk-based innovation. Throughout the entire press conference, in addition to getting the dynamics of cutting-edge trends, you can also get inspiration on how to think about technological development.

This is the main content of Shen Xiangyang’s speech at the press conference that day, which was condensed and organized by Geek Park:

 

good morning everyone! Welcome to the IDEA Conference. We are very happy to be doing some really amazing work in this land of Shenzhen. The purpose of holding the IDEA conference is twofold. The first is to have an opportunity to report and showcase our latest achievements in IDEA. The second is to hope to have the opportunity to share with you our views on cutting-edge technologies and judgments on future trends.

 

 

My speech today is divided into two parts. The first part is my personal experience over the past year or so, what are the latest trends in the development of computer science and artificial intelligence. In this process, I will also intersperse some excellent work done by our research institute to sort out what our IDEA research institute is doing, what kind of people we are, how to think about such scientific problems, and why we do it. such a job. In addition, I would like to share with you the relationship between scientific research and capital investment using an innovation curve.

Regarding the development of artificial intelligence, I have always talked about two major reasons: big data and big computing power. It is remarkable that precise algorithms have promoted the development of the entire industry to the present. There are still new great models appearing constantly, driving a new upsurge in the development of artificial intelligence.

Today, I want to share four meaningful ideas with you.

 

01 From a word is worth a thousand pictures to a picture is worth a thousand words

 

 

The first idea to share with you, a word is worth a thousand pictures.The recent vigorous development of artificial intelligence is the development of pre-trained large models. Its overall scale and speed have reached the point where we cannot ignore it today. There are still many problems that artificial intelligence is far from solving, but it doesn’t matter. I want to share some examples with you today. You can generate very beautiful photos with a single sentence, and the results are very shocking and lifelike. I have been doing computer vision for so many years, and I would not believe it if I told me that such a result could be achieved three or five years ago.

There is a system called Open AI, which is actually made by a company funded by Microsoft. Tell me a word, there is a fox on the grass at sunset, can you draw it in Monet’s style? A year ago, DALL·E could make such a result: the appearance of a bush, the appearance of a fox, and the appearance of a sunset. After one year, the result is very fast.

 

 

There are three reasons why he was able to make such a result: huge amount of data, huge computing power, and a great new deep learning algorithm, especially the Diffusion model is used here. A large number of data pairs between images, annotations and languages ​​can be found on the Internet to make such results. Many people think that today’s deep learning produces such results, and the technical route is simple and crude, but you have to agree that such results are already very detailed and beautiful.

With such remarkable results, what kind of questions should we think about? When IDEA technology has reached a certain level, we wonder if we can make these technologies into products and tools. Because we focus on the digital economy, the bottom layer of digital productivity is the tools represented by artificial intelligence.The most important thing about this matter is that it can help artists. Everyone does something when they have the urge to create art.

Some people have studied the great artists in history. When they were making works, why many artists can only paint a few paintings in a lifetime, while other artists can paint many paintings. The most famous person is Leonardo da Vinci. Art history researchers have come up with a possibility after research: Leonardo da Vinci did not paint so many paintings by himself. He brought many apprentices to paint. At least I, a layman, can’t see the quality of Da Vinci’s paintings and the quality of paintings drawn by his apprentices.

An artist has to bring a bunch of apprentices to be able to paint many things. After the launch of a system like DALL·E today, what is the impact on us? Artificial intelligence is your little apprentice. With artificial intelligence and some tools, everyone is a “Da Vinci”. Everyone has the opportunity to do such things and truly improve digital productivity. In fact great artists are very sensitive technology users. If Da Vinci is alive today, he must be the first to try a system like DALL·E.

After seeing such an effect, do you think that the problems of artificial intelligence have been solved? Of course not, there are still many problems here. Today, big data and artificial intelligence are still a process of memory and interpretation, and there is no real understanding and cognition process.

 

To give a simple example, you let artificial intelligence be an astronaut riding on a horse, and then put it another way, the horse is riding on the astronaut, and this is the result. Because there is no similar concept in the data accumulated by human beings before, it is normal. Such flaws do not prevent the continuous iteration and progress of technology and applications, because with feedback, we can close the loop, make progress, and innovate.

 

 

How popular is the big model thing? DALL E, a start-up company such as Open AI supported by Microsoft, claims to be valued at 20 billion US dollars. For example, the United States came up with a Stable Diffusion method, and a company used open source methods to do it, and it turned out to be a unicorn with a valuation of one billion dollars. The computing power of this company is far inferior to that of DALL·E, but if it is open source, many people will use it, and there will be a closed loop, and it will be possible to make progress.

Dr. Zhang Jiaxing from IDEA Research Institute did in-depth research on large models in IDEA two years ago.Today I am very happy to announce that our IDEA Institute is inThe “Gaia Project” in the field of artificial intelligence content generation, one is the model automatic productivity engine, and the other is the pre-training model system.

The entire project is open source. This project also independently developed the first Chinese open source Stable Diffusion model. The open source of this project also marks the advent of the Chinese AIGC era. We are very happy to be able to make some contributions to the Chinese AIGC community.

You can look at a few examples, enter our system and type a sentence: dream back to Jiangnan, and you can see such a result. Now our model has been on Hugging face, and it has ranked third this year among more than 100 Stable Diffusion. It surpassed 100,000 downloads three weeks ago.

 

 

 

02 From Divide and Conquer to Combine and Solve

 

 

The second idea, I spent a lot of time thinking about it, after the big model came out, artificial intelligence deep learning practice.Today’s impact on research methods ten years later, so I talk about “from divide and conquer to combine and solve”. Those who have studied computer science will definitely agree with my statement. In the past, a very important research method in computer science was divide and conquer. If a problem is too big, it should be divided into small pieces, and then put together after the small pieces.

Ten years ago, people who studied computer vision and natural language processing were basically inseparable, and no one knew what the other was doing.But today with deep learning, the large model brings together people from all links, and everyone uses the large model method to do this today.

Recently, there is a very good paper called “Image as a Foreign Language”, Image as a Foreign Language. A picture, imagined as a foreign language, is also a language. If you have such a perspective, the rest is easy. All the work, methodology, and results in NLP can be applied to computational vision.

This is indeed the case. Google Research published a very remarkable article two or three years ago called “Vision Transformer”. Transformer is the most remarkable achievement of natural language processing in the past two years. The image is cut into small images of 16×16, and then connected together, like a string of characters, and then the method of Transformer is used. On this basis, we apply the Vision Transformer system to the most important problem in computer vision, called target detection.

Let’s look at the following picture, two objects are detected, one is called a motorcycle and the other is called a person. There are many downstream tasks for computer vision to understand images. Once the target detection is achieved, target tracking and target segmentation can be done. There are many industrial applications for these problems, such as medical testing and autonomous driving, and there are huge application opportunities.

 

 

In the past six months, Dr. Zhang Lei led a group of young students to make the IDEA Research Institute continue to dominate the international rankings. It has been amazing for half a year. Not only the result is the best, but also the size of the model, the cost of training, and the training data required are stronger than most other peers.

This is just the result, to give you a feel for what it looks like visually. Here is a very complex video scene, and it can detect moving objects and stationary objects very well. We are still continuing to promote the project. Zhang Lei also open sourced the best model on DETR, which can be said to be the most comprehensive Transformer detection open source framework. Compete with peers and show results that we can do.

What shocked me about this incident was not only the change in scientific research methods, but also the impact on the fields of artificial intelligence and computers.This kind of research methodology and deep learning will have a greater impact on scientific research.

 

 

There has been a lot of talk about AI for Science recently. We should choose some directions, make more tools, and help scientists do better work. I am working with Professor Cai Zheng of Tsinghua University on astronomy. I believe that there will be some very good results to report to you soon.

 

03 From emphasizing calculation results to emphasizing calculation process

 

 

The third idea, from emphasizing calculation results to emphasizing the calculation process.In the process of technological development, it is necessary to constantly look at the feedback from the market and social needs. For example, after 5G comes out, which applications promote 5G, and what new things should be done in 6G. In the development of computer science, the market is very driven. As long as there are great applications, these smart people and smart money will rush in.

In the past so many years, starting from the von Neumann structure, computing is a tool, and the things completed are the tasks assigned to you by the person who masters the tool. When we were still in college, the most amazing thing was scientific computing, and later we could do other jobs.

There have been huge changes since the Internet came out. The APPs that everyone uses are in the hands of the Internet platform. The platform makes your use more convenient, but at the same time it brings some problems. The platform permeates our work and is a black box, opaque. If you want to be transparent, credible and explainable, you need a new computing system to solve this problem:Why do I see such results, why do you give me such results.This has become a rigid need, which can help us solve the problem of the calculation process, not only the problem of the result.

Many smart people are already solving this problem, returning to the essence of this problem, such as the solution of cryptography; for artificial intelligence machine learning, there is federated learning.

At IDEA Institute,We chose a different technical route to make hardware.

We made a thing called SPU last year, it was just a prototype last year, and it has been mass-produced this year. Many bank partners are using it. The meaning of SPU is called Secure Processing Unit. The idea is to get some design concepts from GPU. For example, if you want to play games fast without getting stuck, GPU will appear. Today’s security, it’s not that everyone doesn’t know about it, so they don’t do it. Intel has already done it, and a piece of physical isolation is drawn in the CPU called SGX. But our opinion is that it should be taken out to make a dedicated chip. If you want to do this, security is reflected in all aspects, including security and trustworthiness. It must be safe when it is started, and the container that runs safely must be safe during the running process, and it must also have a safe virtual operation.

 

 

The core advantage of using hardware is that the original algorithm is not changed. What was done in the past is supported by the hardware today, and it can be used out of the box without changing the code. Software engineers who have written programs will appreciate the availability of such tools. Because changing the code is such a painful thing.

With such hardware, we can also combine it with many existing software solutions. For example, our cooperation with Weizhong can greatly improve the performance of federated learning.

I personally think that SPU will definitely be a disruptive technology in the future. Of course, there is still a long way to go. Whether this judgment is accurate and whether IDEA can really be realized will need to be verified next.

 

03 From using language to creating language

 

 

The fourth idea is about language.Doing computers and artificial intelligence is using language, and more importantly, language should be created.

People all over the world speak six to seven thousand languages, and there are thousands of computer languages ​​so far. Because there needs to be communication between humans and between humans and machines, it needs to involve different languages. In the past so many years, what artificial intelligence wants to do is whether the machine can learn human language, such as NLP.

But over the years, the country has not paid enough attention to language. Among the scientists we know, not many actually created language. Two months ago, I invited Zhang Hongbo to join the IDEA Research Institute after leaving Facebook. Hongbo was also my doctoral student at Tsinghua University. He is a rare talent among Chinese scientists who has in-depth research and practice in computer programming languages. Two months ago, the Technical Software Center of the IDEA Research Institute was established. Hongbo has published a very popular language since 2015 called ReScript. After Hongbo came, he quickly opened a practical open class online to teach the ReScript programming language. Currently, the Chinese version of Rescript has been released. Some foreigners said that if they want to listen to Hongbo’s class, they must learn Chinese, because Hongbo uses Chinese to teach. Looking forward to the future IDEA conference, Hongbo will release some new language work for you.

 

 

Today I also want to introduce another aspect of language. Dr. Wang Jiaping led the team to do a lot of work in AI finance and blockchain. Jiaping is now developing a PREDA parallel blockchain smart contract language. Its purpose is that you don’t necessarily need to know what the underlying work is, but you can also write effective parallel blockchain programs.

 

05 The Four Stages of Innovation

 

 

In the second part, I want to talk about innovation, how to organize it in the research institute, how to do it, and why I choose such a topic.

I think there are at least three levels of innovation, one level is technological innovation; the second level is product innovation; the third level is business model innovation. This is all very important. But technological innovation is the most fundamental and subversive.

Basic innovation is inseparable from capital. To innovate, you need investment and capital.

To understand innovation, it is necessary to analyze and describe the paradigm of innovation-capital. In actual operation, what we see are some specific technological breakthroughs, specific commercial successes, which company is listed again, and so on. There are many uncertain principles in it, which are gradually integrated into the real operation experience of this organization and precipitated into the organization’s culture. Today I try to explain the paradigm of innovation.

I want to introduce the innovation process with an innovation curve, which also maps to what we are doing in IDEA. The vertical axis of this graph is the resources invested in the project, and the horizontal axis is a timeline, which also includes a process. A technology starts from the laboratory, then goes to the big market, and finally reaches the Mass Market.

 

 

I think it can be divided into four different innovation processes,Basic innovation, mission-based innovation, risk-based innovation, and product-based innovation.Basic scientific research done in university laboratories is basic innovation. There is not much money spent on each project, and a teacher leading a graduate student may not have a lot of money at the beginning. Some people have some crazy ideas and want to change the world. This is the so-called mission-driven innovation. He has a Mission. For example, what must be done in medical treatment is mission-driven innovation, and the capital required for this single project is far higher, and a lot of money is spent on mission-driven innovation. The number of risk-based innovation projects is less, especially in the later stage, the amount of money that can be invested is already very large. The real innovation with the most money is the R&D innovation, research and development in companies, especially large companies.

Let me give you an example of my brother. Speech recognition is very common today. I have a big brother James Baker who is very amazing. In his doctoral dissertation in 1972, he proposed a mathematical model Hidden markov Model. To solve the voice. After graduation, he did not become a university professor. He had a crazy idea, and he thought that if this thing could work, it should be able to be used as a dictation machine, so he was the first in the world to create a dictation machine. People, the error rate at that time was very high, and the application scenarios were relatively limited.

He and his wife founded a company called Dragon Systems, which is the first company in the world that actually does speech recognition. Once these small companies have cleared the way for this matter, big companies will flock to it. Apple, Microsoft, etc. have all gone up.

Throughout the process, we should think about the cooperation between capital and projects. In the process of innovation, we need to figure out who are the real participants and who are the real contributors, and what kind of return on investment and capital model are used to connect these participants and contributors together.

In the end, money is still needed to support the overall scientific research process. There are many capital models, government and market; short-term and long-term; profit-oriented and non-profit. There are also many subjects, including government agencies, market-oriented companies, small workshops, and even one person working hard.

In all investments, the market is the biggest subject. The money invested by companies and enterprises is the largest part of R&D innovation. Let me show you a piece of data, which is that the R&D investment of the entire country from 1956 to 2020 in the United States is getting higher and higher. In 2020, companies and the market will invest 75% of the national R&D budget of the United States, and the government will invest 9.4%. Is it possible to draw a conclusion: the government is not important? Absolutely not. Although government-led research and development is far lower than market-led research and development in absolute terms, it has a very important guiding and directional role. Many future innovations would not be possible without the main body of government.

Let me show you another figure, the R&D comparison of the three countries (Figure). Israel ranks first in the world, and its national R&D expenditure will account for 5.4% of GDP in 2020. The fifth is the United States, and the thirteenth is China. China’s emphasis on science and technology can be seen very clearly from this picture. In a short period of time, China’s overall investment has increased rapidly. Don’t underestimate 2.4%. If the purchasing power of the entire scientific research is also included, including the cost of personnel and shopping, many domestic devices may be 1/3 of the price of foreign countries.

 

 

The conclusion is very clear.Technology is the root cause of social progress, the market is the main force for technological development, but the government is the initial driving force for sustainable development.Both the market and the government are the main body of investment, and all these innovations are made with money. After investing money, where is the return he expects? My opinion is very simple. It tends to be early-stage basic innovation, including mission-driven innovation. The funder is mainly the government, and of course some non-profit organizations. He cannot and should not appeal to the return on capital invested in technology. This is very important, so that you can have a long-term perspective. When you invest in it, you should not think about the money coming back immediately or when.

An example of basic innovation is the NSF (National Science Foundation) model, and there is also NSF-C in China. Mission driven is the DARPA (Defense Advanced Research Projects Agency) model. There is also the VC model. When it comes to the company, it is product-based innovation.

NSF is to spread pepper noodles and support more professors and students. The funding for each project is very small, generally 1-2 graduate students, with a scale of 50,000 to 100,000 US dollars. NSF’s 2020 budget is only 1 billion US dollars in the computer industry and the information industry. Of course, it is not a small amount of money, but Google spent 600 million US dollars to buy Deepmind. Therefore, although the number of NSF is not large, it can support many people to try early basic scientific research. The most important thing is that NSF never said how much money we made from this thing, he just talked about how many amazing students we supported and how many amazing things we did later.

The second is the DARPA model. The rise of the United States as a superpower is closely related to their mission-based innovation. The United States did not do this before World War I. After World War I, the federal government invested money in public health, medical care, and public safety. Slowly, a so-called mission-driven innovation path emerged. The particularly successful one was DARPA, which did global satellite positioning. These were all mission-driven back then, and later produced huge value for society and business. The failure rate of mission-driven innovation is very large, and all we hear are a few examples of success. Once successful, the influence will be very large.

The book “The Brain of the Pentagon” has interviewed many researchers, and everyone agrees that DARPA is at least ten or twenty years ahead, and there are a group of very smart people who really make things. The DARPA model is still very well recognized by everyone so far. President Biden has proposed to increase the research funding of the National Institutes of Health by 9 billion, of which 6.5 billion is dedicated to the establishment of an Advanced Health Research Program modeled on DARPA.

Risky investment has high risks and high returns, and it is a game for risk takers. Entrepreneurs are more likely to die. More than 90% of startups fail. Only with commercial success can there be more money, whether the government collects taxes or other people invest money, let innovation continue to be done. At this time, we are talking about profit. Whether it is a large company, a laboratory, a small company, or your own garage, when you start, you have to think about the profit of this matter. But there is another important thing. Without the foundation of basic innovation and mission-based innovation, risk-based innovation is also very difficult. The importance of government input is here.

Finally, briefly talk about product innovation.

Usually, I don’t hear too much about it, except that occasionally Apple holds a press conference, and actually spends a lot of money. what is the reason? Usually, most of the R&D money in the company must be maintained to maintain the continuous improvement of existing products and the new needs of these users. He must use new products to ensure that the company is profitable, and to ensure that the performance of each quarter is accountable to shareholders, so that Employees can continue to pay wages, so most of the money is spent here, and this money is very huge.

This is the number of Google in 2020. Google ranks first and Microsoft ranks third. Huawei is amazing. Huawei’s R&D figures are huge and the investment is huge.

Data source: EU investment scoreboard

 

How big companies innovate is an eternal topic. In the end the company was unsuccessful, ceased to exist, and had nothing else left. To be a century-old store, you must innovate, and only by subverting yourself and actively embracing disruptive innovation can you do this.

 

For IDEA, where should we invest our limited resources, and where do we stand on the innovation curve?

IDEA Research Institute is an international new research institution, we connect mission-driven research results with the needs of future industrial development. More specifically, we are a second-class public institution supported by the government. We hope to connect the innovative seeds sown by the government’s scientific research investment with marketization, especially in the Greater Bay Area and Shenzhen’s market-oriented capital mechanism.In the picture we aresit two look three。

 

 

 

 

06 IDEA’s crazy ideas

 

 

I would like to give you a detailed explanation of what these crazy ideas are doing through a series of scientific research products IDEA is doing.

First introduce the first crazy idea.Today, when artificial intelligence is developing, algorithms can generate algorithms, and even models can build models. Then our crazy idea is whether artificial intelligence can create artificial intelligence in the future.

This is our favorite Professor James Simons. He was the head of the Department of Mathematics when he was 30 years old. After doing data research for a period of time, he suddenly had a crazy idea. Renaissance, which has a fund called Medallion, has a quantitative rate of more than 60% in the past year. He used to be Professor Chern’s Ph.D. at UC Berkeley.

Another example is Black, Merton and Scholes, three Nobel Laureates in Economics, whose research results have also influenced the field of financial quantification. There are also many computer scientists, such as UC Berkeley’s Turing Award in 2011. They won the award because they did a very good job of causal reasoning, which also affected quantification. In 2018, the theory of the three musketeers of deep learning also affected the development of the entire industry.

 

 

Professor Guo Jian of the IDEA Research Institute leads a very strong young team and does a very good job. Professor Guo helped me summarize the three stages of quantitative investment experience in the past 40 years. In 1.0, a few smart people patted their heads and came up with some models. In 2.0, a small factory has become a large workshop, and more people are looking for such a Factors can be made into a pipeline. Now after machine learning, use machine learning to make models, in the 3.0s.

Even today’s 3.0 has not really achieved the level that deep learning can do. We propose that we can do better and reach the 4.0 stage. It has several characteristics, from manual modeling to automatic modeling, from the original black box to explainable artificial intelligence, from just data-driven to data-driven plus self-driven. We have done a lot of work on this.

Next, let’s talk about the second Crazy ideaI would like to introduce the medical and health aspects. Xie Yutao is doing a mission driven (mission-driven) project in IDEA. He hopes that in the future, any disease will have a “spectrum” that can be “medicated”, so that medical knowledge can benefit the public.

Last year at the IDEA conference, I briefly introduced to you that we have cooperated with Harvard and have been working with Professor Yu Sheng of Tsinghua University. What we want to do is a data-driven, artificial intelligence-assisted super knowledge map, which can make medical knowledge from all over the world into a super knowledge graph. Our target is UMLS (unified medical language system), the world’s largest system made in the United States. We have been doing it for less than two years, and we have already achieved initial results, and we have made it public online.

 

 

Today, the entire BIOS system is already the best in the world. Our number of entries far exceeds that of the UMLS system in the United States, and the number of concepts far exceeds the original one. The reason is also very simple, because we are data-driven, digging every day, and after new medical papers come out every day, we continue to find new entries. The accuracy rate and coverage rate are far more than competitors.

 

We are still working on it, and I hope everyone will try to see a system like ours. The space for future applications is very broad, and we hope that English, Chinese, and even other languages ​​​​in the world will be expanded.

The third Crazy ideawe have some ideas on education, I shouted a slogan “Let the world have no difficult papers”.

Doing scientific research today spends a lot of time reading papers, which are very difficult to read, because most of the papers, including our own, are poorly written and have to be read. How to read the papers quickly, I call it rough reading, How to read it in is called intensive reading, and how to understand these things and use them for yourself. There are many ways here, including the inconvenience of finding papers and inconvenient management tools.

Today I am very happy to release readpaper2.0 version, our system is open and free. Readpaper version 2.0 tries to solve four major problems: finding papers, intelligent reading, document management, and academic discussion.

Finally, I would like to share with you the mission-driven direction.Talk about the application of artificial intelligence in enterprises. The title is “Auxiliary Decision-Making in the Fourth Dimension”.We want to make a business map of affairs.You have heard a lot of things today, and the most important thing in it is the addition of a latitude, which is time dimension information. This is different from the traditional knowledge graph, you can know the causal relationship and timing relationship between events. Why is it important? As an enterprise, you must care about what kind of decisions should be made in such a turbulent business environment.

 

 

 

07 Find innovation “sweet spots”

 

 

Finally, I would like to summarize again, in the paradigm of innovation capital, where does the IDEA Research Institute focus. In fact, we have always wanted to find an innovative “sweet spot”. With the support of such funds, we can do some mission-driven innovations without staying here, and we can move forward and implement many things.

 

 

If our goal is to focus on such an innovation “sweet zone”, what kind of abilities do players need?First of all, you need to have the ability to choose the problem.It is not easy for first-rate masters to select problems, second-rate masters to solve problems, and third-rate masters to copy problems. The ability to choose problems is the most critical, and then you must have the ability to allocate resources, and then make unremitting efforts to really have the ability to do this thing from beginning to end. Most of IDEA’s projects are selected in this way. If you have a scientist’s mind, entrepreneurial quality, and entrepreneurial spirit, then IDEA is your best choice.

Finally, I want to share with you a very exciting project.We have just started, but have been planning for a long time, which is the “low altitude economy”. Many aircraft, such as helicopters and drones, generally fly at low altitudes below 1000 meters. The entire airspace is available. The simple definition of low-altitude economy is:In the low-altitude airspace, it is a comprehensive economic form that is driven by manned and unmanned low-altitude flight activities and radiates the integrated development of related fields.The low-altitude resources that are being wasted now are natural resources, and we can turn them into very valuable economic resources, just like roads on the ground, which become economic resources after the road is repaired.

The difficulty is going from airspace that is usable and accessible today to airspace that can be counted. Airspace is a very complicated matter. Today’s drones fly in the sky and cannot be seen, managed, or used. In the future, it will all need to be done. I am also very grateful to Academician Li Shipeng for leading the team to develop an intelligent fusion low-altitude system SILAS, to conduct in-depth research on all scene digitization issues, and to promote the development of this matter. Today I am also here to announce that I am grandly releasing the “White Paper on Low Altitude Economy Shenzhen Plan”. We welcome your valuable opinions.

 

 

Finally, I would like to end today’s speech with a few words. Why are we so keen on the low-altitude economy, why did we come to Shenzhen to innovate and start a business? What exactly are we trying to do? In fact, the reason is very simple. If we look back at the brief history of the entire human race, it is a history of bravely entering no man’s land. From our ancestors walking out of the African continent, to the discovery of the New World of America by Columbus in 1492, to the development of the western United States, and even to the space travel that everyone is talking about recently, human beings are looking for such “no man’s land”.

Today, we opened the window, and the low sky outside the window is the “no man’s land” closest to us.So I think we are in Shenzhen, we work together to unite resources from all parties, transform low-altitude into economic resources, realize low-altitude calculations, create a new paradigm of work and life, and open up new space for economic growth. thank you all!