Shulou(Shulou.com)06/02 Report--

Doing AI in China has something to do with Tsinghua University.

Author | Zhang Dong

In the traditional security market, high achiever of Tsinghua University made their debut one after another and pressed the key to thorough reform.

In 1978, when the Cultural Revolution was just over and reform and opening up were brewing, Tsinghua University set up a teaching and research group of "artificial Intelligence and Intelligent Control" within the computer Department.

That year, Tsinghua University enrolled six master's degree students, namely, Zhang Guoxuan, Zhang Xuecai, Wang Jiaqin, Zheng Xuezhong, Tang Jianbang, and Yan Junwei, and the instructors included Xu Wanyong, Huang Changning, Xu Yaorui, Shi Chunyi, and Liu Zhi frame.

These instructors became the pioneers and mainstays in the field of artificial intelligence in China in the following two or three decades, and the six first master's degree students have also become key figures in various fields.

In 1990, Tsinghua University established the State key Laboratory of Intelligent Technology and Systems, which has been rated as a Class A laboratory in the evaluation of the State key Laboratory for three times in a row.

The characteristic of the key laboratory is not only to study the basic principles and methods of artificial intelligence, but also to study the application technology and system integration technology. It not only reflects the forefront of the research, but also reflects the characteristics of multi-disciplinary intersection and the synthesis of different technologies.

According to the authoritative CS Ranking index in the computer field, during the 10 years from 2010 to 2020, Tsinghua University has ranked second in the world in artificial intelligence, second only to CMU.

If Tsinghua is the cradle of AI talent in China, in fact, there are many heroes in the market segment of AI security.

Chairman of Haixin Kejin: Liu Xiaochun

Liu Xiaochun is one of the first entrepreneurs engaged in pattern recognition research in China. He graduated from the Department of Automation of Tsinghua University in 1989 with a master's degree in pattern recognition and image processing, and later worked in a subordinate enterprise of the Naval equipment demonstration Research Institute.

In 1998, he resigned to start a business and founded Haixin Kejin.

They have successively developed a number of best-selling products, including Alipay "fingerprint payment", and the portrait recognition system with the largest storage capacity of 1.1 billion in the country is also provided and jointly built by Haixin Kejin.

It is such an once brilliant public safety information enterprise, which has encountered the growth challenges and technological transformation problems that almost all established enterprises have encountered in the process of rapid development.

Fortunately, Liu Xiaochun and his Haixin Kejin are also making continuous efforts on AI.

A few years ago, they established the Haixin Kejin AI Research Institute, which focuses on big data services and artificial intelligence applications. At the same time, we are also actively cooperating with enterprises including Intel, Ali and Huawei to create Haixin AI ecosystem.

"since its establishment, Haixin Kejin has been doing things to meet the rigid needs of society, and we will not spend too much energy on icing on the cake."

Liu Xiaochun said that any technology must be used to solve practical problems. Haixin Kejin will not do research and development like the Eiffel Tower, and all technologies must go to the market to verify its success.

In the eyes of others, such a development idea may be a bit insipid, lacking stories and highlights, but in the eyes of Liu Xiaochun:

Steady development is not a lack of progress, nor is it unwilling to take risks. It is based on years of understanding of recognition technology and better understanding of making choices.

Speaking of the future of AI security, the Tsinghua native spoke eloquently:

"AI is not a paper technology. The advancement of hardware, platform and network will certainly bring about qualitative changes in the industry."

He believes that in today's AI entrepreneurship, the technical threshold is getting lower and lower, but the application threshold is getting higher and higher, and the requirements are becoming more and more refined.

Because of this, each company must find its own value at a later stage.

For technology companies, the value of the algorithm will gradually decline: on the one hand, it will continue to open source, the threshold will be lower; second, there is no production cost, and replication sales will be faster than printing money.

In other words, the value of the algorithm lies in matching: either supporting projects, supporting equipment, or supporting application software.

"the AI app is a marathon with no end, and everyone has a chance."

Liu Xiaochun is full of confidence in the future of AI, at the same time, he also believes that even so, we should be cautious about starting a business.

"I do not quite support the mass entrepreneurship and innovation proposed by the country. There are few winners and losers everywhere, but in fact, any entrepreneur will consume a lot of social resources, family resources and personal resources."

Personal sacrifice of some time and effort is secondary, some mass entrepreneurship and innovation directly exhausted the limited funds of the entrepreneur family, bringing the plight of family life should not be.

Starting a business must be something that a few elites do.

Ignoring the three swordsmen: Yinqi, Tang Wenbin and Yang Mu

It is ignored that the three co-founders of science and technology, Yinqi, Tang Wenbin and Yang Mu, are undergraduate students in the "Yao class".

Yinqi joined the Yao Qizhi experimental class at Tsinghua University in 2006. Since then, he has worked in MSRA (including internship) for four years. He has studied Microsoft's face recognition system with Sun Jian and others, and has been widely used in Microsoft products such as X-box and Bing.

Inky then went to Columbia University to study for a doctorate in 3D cameras, and then returned home to start a formal business, founded Lazio Technology, and served as CEO.

Tang Wenbin has participated in informatics programming competitions since junior high school. He has won many championships in ACM,Code Jam and other programming competitions. He was sent to Tsinghua University in his second year of high school, and then entered the "Yao class."

Tang Wenbin also served as the head coach of the National Informatics Competition for seven years and was responsible for selecting gifted children from high school to participate in the International Information programming Olympiad (IOI).

Tsinghua graduate stage, focusing on social mining and image retrieval, has won ECML best student paper runner-up and Google topcoder target.

Yang Mu graduated from Yao Qizhi Experimental Class of Tsinghua University and won the gold medal in the International Information programming Olympiad (IOI).

When it comes to absenteeism and security, people in the industry should have heard of it.

At present, the overall valuation of more than $4 billion is the first batch of unicorn enterprises in China to apply AI technology to the traditional security industry.

Nine years ago, AI was not yet popular, there were few start-ups, and the VC circle did not recognize it. At that time, it was necessary to have a rare feeling of putting all your eggs in one basket as an AI. Not long after entering the venue, Xianyang aimed at the industry's recognized hard bone: the security market.

People were confused, puzzled and even ridiculed. After all, there were more views in the industry at that time that AI was just a new pot of old wine, a clich é, landing application and even commercialization would have a long process, and the technology application bubble was huge.

There are also some industry customers who think that since they were fooled by face recognition ten years ago, they will no longer believe this stuff.

Ignoring it, he took root in this vertical field.

When the technology is proved to be mature with one result after another, the industry really realizes that the era of AI has really come.

The experience reflects the most real appearance of a startups from infancy to growing up, and when he is born, the business picture gradually unfolds.

Today, the product matrix developed by discerning security involves many fields, such as human, vehicle and object feature recognition, and the business line involves cloud, edge and end.

It can be said that in the current era of traditional security embracing artificial intelligence, Xianshi is building a full value chain covering "algorithms, technology, hardware products, solutions, data" in a soft-and-hard way.

Co-founder of Shangtang Technology: Yang Fan

Yang Fan, who graduated from the Department of Electronic Engineering at Tsinghua University, is a versatile person with both academic and engineering abilities.

After graduating with a master's degree, he directly joined Microsoft Research Asia, where he was responsible for the engineering of graphics-related technologies such as AI and AR for eight years.

Then, together with Professor Tang Xiaoou and others, he founded Shangtang Technology.

As a co-founder, Yang Fan's main responsibility is to help Shangtang build a complete engineering system, use state-of-the-art technology to become reliable products, and create enough value.

When it comes to commercial soup, CV practitioners should be no stranger. As the largest and most valued AI unicorn in the market, it relies on a large number of professionals and promotes many original AI technologies to blossom in security and other industries in just a few years.

Yang Fan said that the working process of AI security can be divided into five steps: collection, transmission, storage, analysis and feedback.

Before AI was integrated into the security market, the collection, transmission and storage processes of the industry had evolved excellent enough, including the continuous iteration of various high-definition cameras, the overall popularity of 4G, the omni-directional deployment of cloud computing and data lakes, and so on.

The evolution and application of these technologies have led to the formation of product solutions in the first three processes, which has led to the rapid progress of the industry, and made China's security stand on top of the world in a short period of more than ten years.

Unfortunately, the two major links of "analysis" and "feedback" have not been substantially developed in the past decade, resulting in the provision of closed-loop value of industrial informatization in this field. "

Analysis is that the data collected by the front-end IPC really have the ability of automatic value extraction; feedback means that after having the basic extraction ability, combined with the needs of users, the application ecology can be formed, which can really bring value to customers.

Yang Fan firmly believes that the reinforcement of the latter two links is the opportunity for Shangtang to make security efforts, and Shangtang will also try its best to commercialize and reconstruct the immature links of a mature industry.

With the help of top talents and many original technologies, today's Shangtang security has been fruitful.

At present, Shangtang's security business has covered more than 30 provinces across the country and has participated in the construction of a large-scale city-level video analysis system with a total of more than 100,000 roads.

In addition, Shangtang's AI security product matrix has also been formed: including the city-level visual open platform of SenseFoundry Ark released in 2018, the actual combat platform of SenseFace face control, and so on.

Through AI Nuggets, Yang Fan also shared his three views on how AI technology companies can better survive in the security industry.

First, have a thorough understanding of the mode of operation of the industry itself. for technology-based companies, they should not only go to the technology end, not to the client.

This may seem like a clich é, but there are very few companies that can actually implement it. Each industry has its own operating rules and inherent laws. How to better understand the needs of users is the top priority.

Second, technological strength is an excellent weapon to bring essential innovation to various industries.

If you only go to the client, or despise the technical side, then it is only a matter of time before you are caught up and overtaken by the opponent.

This tests the production ability of a company, and how to combine technological innovation with user needs faster and better is a required course for every technology-based start-up company.

Third, we need to learn to make full use of our strengths and avoid weaknesses.

Each player in every industry has its own advantages and strengths; but the market competition is more comprehensive, in this process, we need to fight while adjusting, as much as possible to play their own maximum value, rather than follow the trend of adjustment.

For example, just like the 'software definition' proposition that has been hotly discussed in the security industry recently, which is right or wrong?

In fact, through the very mature mobile phone, IT and other industries, we can find that a large number of companies make money through software, hardware and services.

In other words, whether it is to do software or hardware, each company will have its own living space, and ultimately depends on how much value and services it can provide to users.

In the high-speed AI security track, traditional security companies will constantly emphasize customer service and develop technical capabilities, while technology companies should focus on technology on the one hand and stay close to users on the other.

Whoever first reaches the midpoint of the intersection of technology and users will win the battle.

CEO of Toujing Infinity: Xiao Hongbo

Xiao Hongbo, who graduated from the Department of Engineering Mechanics of Tsinghua University, has served as a technical consultant for Hewlett-Packard and CA China, a senior information architect for IBM, and a project manager for the Image Group of UIUC Advanced Media Research Center.

In 2010, Xiao Hongbo, together with Lu Fan (partner), established Toujing Infinity in Beijing.

Over the past decade, the unlimited growth of Toujing has not been plain sailing, even a bit bumpy.

The path chosen at the beginning of Toujing Infinite Entrepreneurship is based on the "front-end perception" of artificial intelligence.

At that time, there were not enough successful precedents at home or even internationally, and touching the scene could be said to cross the river by feeling the stones.

But with the keen awareness and self-confidence of the market, Toujing is infinitely ahead of the industry to figure out the "front-end awareness": the necessary supplement to cloud AI and the deep application of edge intelligence.

The breakthrough growth of marginal intelligence in recent years has confirmed the correctness of the original decision.

This persistence also makes Toujing infinitely leading the market into this industry, providing users with more mature and stable front-end perception products, and has become the leading enterprise in the marginal intelligence industry.

Over the past ten years, Toujing has gradually become a national high-tech enterprise with top software and hardware R & D team, five subsidiaries at home and abroad, seven major industries and dozens of industry customers.

Compared with the back-end intelligence, the front-end has many advantages, but it also faces some complex problems, such as large model parameters, high real-time requirements, weak computing ability and outdoor complex scene operation.

In Xiao Hongbo's words, being intelligent at the front end is tantamount to "doing an ashram in a screw."

Fortunately, after ten years of clenching my teeth, fate finally gave Toujing a good explanation.

In the continuous polishing of technology and scene applications, Toujing Infinity has also launched a number of differentiated products and solutions, including Shield smart box / analysis host, artificial intelligence module "instant vision", horned hummingbird Horned Sungem AI vision kit, etc., which are also used in dozens of scenarios, including security, finance, retail, community and so on.

For the development of the AI security market, Xiao Hongbo believes that there is no end point. He put forward the concept of "three millimeters": millimeters, milliseconds and milliwatts.

This also reflects Toujing Infinitely's technological vision of perception, that is, to create a product with "small size, fast response speed and low power consumption". The end point of this concept is perceptual "chipalization".

Replace the original complex design with a single chip and put in a variety of devices to create a 'visible' world. "

This is Xiao Hongbo's dream, and it is also a dream with boundless touch.

Cyrus AI Gemini: Yao Song, Shan Yi

From left to right: Shan Yi, Wang Yu, Yao Song, Han Song

Yao Song was born in 1992 and entered the electronics department of Tsinghua University in 2011.

At the beginning of 2012, he entered Dr. Wang Yu's laboratory, starting with the research on CAD such as 3D integrated circuit design and hardware security, and then delved into computer architecture.

In July 2014, I went to the Electronics Department of Stanford University for a 2-month internship in LLVM compiler development.

After September 2014, he worked as an intern at Microsoft Research Asia, where he designed and developed a special compiler for Bing search engine acceleration processors, which laid the foundation and direction of Yao Song's entrepreneurship: deep learning hardware.

When he graduated in 2015, Yao Song received a doctoral Offer from Carnegie Mellon University (and, of course, a large academic bonus), then chose to give up his own business and co-founded Deep Science and Technology with Shan Yi, Wang Yu and Han Song as CEO.

Shan Yi graduated from Tsinghua University with a doctorate in electronic science and technology. Imperial University of London jointly trained doctoral students.

The former technology partner, CTO, was fully responsible for the management of the company's technology direction and R & D and development team.

He led the R & D team to develop DPU (Deep Processing Unit, Deep Learning processing Unit) and its software and hardware systems for artificial intelligence applications, and on this basis formed two key solutions of deep learning accelerated FPGA card module and special chip, which are applied in intelligent security, data center and other fields.

In August 2018, Shen Jian Technology was wholly acquired by Cylins, and now the team is mainly responsible for the AI R & D and project landing contact.

According to AI Zhidao reported that the initial product positioning is drones, later considering a variety of factors, the final choice to transform the market such as security (of course, the accumulation of drones has not given up).

Based on the accumulation of UAV technology, they began to develop a trinity automation process of compression, compilation and acceleration, trying to achieve collaborative optimization of algorithms, software and hardware, so as to create an efficient deep learning DPU platform (Deep Learning Processing Unit, deep learning processing unit) for developers.

In 2018, Shenjian Technology already has six products: deep learning development SDK DNNDK, face detection and recognition module (front-end), face analysis solution (back-end), video structured solution (human / vehicle analysis), deep learning ARISTOTLE architecture platform, speech recognition acceleration scheme.

On the other hand, DNNDK provides a complete set of full-process software tool chain for Standard Nvidia's TensorRT,DNNDK, which covers the whole process support of neural network reasoning (Inference) phase from model compression, heterogeneous programming, compilation to deployment.

It reduces the threshold and deployment difficulty of deep learning application development on DPU platform for deep learning algorithm engineers and software development engineers, shortens the cycle of algorithm deployment to hardware, and speeds up the process of AI product development.

Chairman of Wen'an Intelligence: Tao Hai

Tao Hai received a master's degree in automation from Tsinghua University in 1993 and a doctorate in Electrical and computer Engineering from the University of Illinois at Urbana-Champaign in 1999 under Professor Thomas S. Huang, the father of computer vision.

In 2001, Tao Hai taught at the University of California, USA, and later became a tenured professor of the university (until 2010). He has published dozens of influential academic papers and won more than ten invention patents in the field of computer vision.

In 2005, Tao Hai led the team to establish Wen'an Intelligence, focusing on artificial intelligence vision technology and product research and development, mainly used in the field of intelligent transportation.

"different from others, we only do vertical, focus on a single area of deep ploughing, it is not only to land, but also to give users a value embodiment."

At present, Wenan Intelligent's product line has covered front-end cameras, edge end, cluster servers, back-end application platforms.

In Tao Hai's view, the technology trend of intelligent transportation, cloud collaboration is the future, and the front-end AI is essential.

It is not advisable to use back-end AI to solve all problems. Electronic police is completely different from ordinary face recognition. In order to accurately and quickly recognize license plates in the case of high-speed movement and multi-lanes, some special applications need to be carried out in the front end to achieve front-end cooperation.

The network transmission is established in the front end, and the coordination is compatible; the old electric police is reformed and the function is added in the back end. At the time of construction, we should plan on schedule and gradually expand the network to solve the differences in different actual needs of different places.

The industry trend of intelligent transportation is mainly divided into three levels: intelligent vehicles, intelligent roads and intelligent management.

The intelligent vehicle is similar to the to C terminal, and the current vehicle networking, autopilot, auxiliary driving and so on are actually done by the enterprise to C side.

Road intelligence, vehicle-road coordination needs construction, road side construction brings road management, including congestion, traffic lights, traffic organization, information service, information release and so on.

The other part is the management intelligence of the government management side, in fact, it is analysis, decision-making, law enforcement, optimization to improve the service ability, in the end, it is not management but management, is the integration of services into daily life.

Wen'an Intelligence is mainly to build the corresponding system with the main goal of intelligent road and intelligent management.

The road is intelligent, that is, the urban road is refined, the highway is strengthened, and the rural road is mended.

Intelligent management, the goal of the next stage, in fact, is to treat congestion.

Chairman of Liwei Zhaopin: Xu Ming

Xu Ming, PhD from Tsinghua University, has worked in ZTE for more than 15 years and served as chairman and president of Liwei Zhaopin in August 2017.

He has made great achievements in the field of smart city, has more than a dozen patents for personal inventions, and is the main practitioner and promoter of the "one map, one network, one cloud" structure of the smart city.

He believes that under the trillion market of the Internet of things, the combination of AI and Internet of things technology has been very necessary and inevitable, the integration and operation of the two will bring a new round of changes to the security industry and even the world.

In his view, there are still two inadequacies in the landing of relevant technologies in the industry:

1. The depth of the combination of security industry and public security actual combat is not enough.

Artificial intelligence, Internet of things, big data and other new technologies must serve the problems and challenges in the actual combat of the industry and public security, and be able to solve the problems in order to truly achieve the landing in the actual combat of the industry and public security.

2. Relying on a single video data is not enough to solve all the complex security problems.

The solution under a single video data is too luxurious and will increasingly involve the sensitive issue of public privacy.

The existence of the above two major problems also means that the three major technological areas of artificial intelligence, the Internet of things and big data themselves need to be deeply integrated in order to be deeply integrated with the actual combat of the public security.

In other words, AIoT+ public safety strategy may become the main theme of the industry in the future.

In the later development and landing of AIoT technology, Xu Ming said frankly that Liwei will continue to adhere to three things:

First, to open up the governor's second pulse from perception to cognition, and form a closed loop of perception and cognition.

Second, establish the automatic evolution and progress of data, information, knowledge and wisdom.

Third, the IoT node router with brain-like neural function AI chip is widely used, which distributes intelligence to any node of the Internet of things.

Chairman of Shen Xing: yuan Peijiang

Yuan Peijiang graduated from Tsinghua University and received his doctorate from the University of Western Ontario in Canada. He studied under academician Zhang cymbal of Chinese artificial intelligence and Professor Patel, academician of both Canadian academies.

In 2015, Yuan Peijiang wrote a more than 20 trillion face recognition Mini Program. The Mini Program attracted the attention of the police and was taken to Xinjiang for testing.

As a result, less than a week later, the crude Mini Program identified an anti-terrorist, and this attempt made Yuan Peijiang see the demand in the field of public security.

In 2016, Yuan Peijiang participated in the founding of Shen Xing Technology, focusing on the independent and controllable innovation and development of artificial intelligence products and application services in the field of security.

Compared with a large number of previously established, high-financing, high-valuation computer vision companies, deep awakening seems to be a little late. But it doesn't seem to affect their grades at all.

At present, more than a dozen provinces and dozens of cities, including Xinjiang, Guizhou and Qinghai, have used deep awakening products to achieve AI landing.

Over the past few years, Yuan Peijiang has gradually deepened his understanding of AI security.

On the one hand, although many enterprises are talking about technical tactics, in fact, only a few manufacturers will really spend time on it.

Second, there are criminal police, traffic police, public security, household registration, economic investigation, patrol police and other types of police in public security, and the demand for the application of video resources is also different.

Under the scene segmentation, there are many difficulties to be overcome, and how to balance the actual scene demand and the cost of production is also an important problem.

In contrast to the industry's pursuit of demand for segmentation, Yuan Peijiang believes that customer demand is multi-scene, but scene segmentation does not mean that all products have to be formed. if they cannot be copied, it will be difficult to balance the problem between demand and cost. For enterprises, the burden will be heavy.

"products are extracted from projects, but not all projects can extract products. Products must be highly consistent in some scenarios, and can only be formed when the product is highly programmable."

In the attention of subdividing the scene, Shen Xing also refines some more detailed applications. Shen Xing declared a key topic early last year. Through a set of footprints, or shoe prints, one can invert a person's height, age, posture and behavior.

For the role of AI start-ups in the market, Yuan Peijiang believes that although there are many players, the penetration of AI in the security market is still very low, which is not a "Red Sea" market, and AI also has a lot to do in it.

Founder of Zero Intelligence Control: Yang Jianjun

Yang Jianjun, who graduated from the Department of Electronic Engineering of Tsinghua University in 2002, joined a research institute of the second Artillery Corps to engage in the research of aerospace vehicles, and founded Zero Intelligence Control in 2007.

In May 2016, Zero concentrated all its resources to launch a pocket drone, DOBBY, in an attempt to capture the consumer drone market.

However, six months later, the rapid expansion of zero degrees due to product sales did not meet expectations of a tight cash flow situation, and finally had to lay off 134 people at once.

After learning from the bitter experience, Yang Jianjun greatly adjusted his strategy, made a big push into the To B market from the To C field, and chose security as the technical commanding point.

Through a lot of research, they found that there is no enterprise in the security industry, which has both flight control and aircraft, as well as pods and data links, and understands the industry and has Internet thinking.

Thanks to the previous team's technical accumulation, Zero Intelligence Control has launched flight control, pods, data links, vision modules, and then do the whole machine.

The main scenes are police arrests and daily inspections. According to their product definition, future operators only need to operate indoors, and leaders can see any location within their jurisdiction at any time, and can also broadcast live to ground personnel for navigation and tracking.

Yang Jianjun inferred that the demand for aircraft in the security market will be very large, and it is relatively easy to do 500 million a year.

"Today's mature technologies will support the vigorous emergence of an era of police security, and police security, border defense, and inspection will be the next key direction."

For zero degrees, there are two key things to do in the future:

First, to find a long-term direction, the market needs to be big enough, deep enough and technical enough to keep going.

Second, we should make a long-term direction and make a profit in the short term, control cash flow and take every step.

General Manager of Deep net Vision: Wan Dingrui

Wan Dingrui received his bachelor's degree from the Department of Automation of Tsinghua University in 2004 and his doctorate in Control Science and Engineering from the Institute of Information processing, Department of Automation of Tsinghua University in 2009.

Since 2015, he has been appointed Vice President of Dongfang Wangli Technology Co., Ltd and General Manager of Shenzhen Shenwang Vision Technology Co., Ltd.

When it comes to artificial intelligence, this Tsinghua native's chatterbox has been completely opened.

"Security eventually needs to complete the transition from perception to cognition, but there is no real DI (data intelligence) in the security industry yet."

At this stage, the technological trend of the security industry in the future is cloud big data, deep artificial intelligence and the integration of software and hardware.

With the continuous evolution of technology, it is very difficult to continuously and accurately judge the future technology trend, which requires continuous tracking and judgment to hone one's own technical sense of smell.

While ensuring a sensitive sense of technical smell, the landing of technology is a top priority. Wan Dingrui believes that there are still many deficiencies in the application of intelligent technology to related systems.

Nowadays, the AI monitoring business of mature players in the track mainly focuses on the above two points:

AI unicorns open the gap in algorithm accuracy at all costs, while traditional manufacturers systematically improve their stability, cost and service on this basis.

"Security systems place special emphasis on accuracy and real-time performance. Traditional security systems do precise layout and control, in addition to high accuracy, timeliness can not be ignored, at least seconds of response is required, but for now, the challenge is enormous."

The main reason for not being "fast" is the deficiency of the algorithm and the lack of computing power: the city-level security system has a large number of points; in addition, as a kind of unstructured data, video pictures are processed by deep learning, resulting in a huge amount of computation and high cost. In order to achieve real-time response, it requires a lot of computing power and high cost.

"all these problems can ultimately be summed up at the talent level, which is one of the important factors hindering the landing of AI. Technological innovation is accelerating, and the landing of a systematic project requires cross-integration and comprehensive utilization of a variety of technologies. At present, there is a shortage of comprehensive talents in our country."

From the perspective of Dongfang Power itself, over the past years, they have invested a lot of energy in intelligent algorithms, intelligent hardware, human-computer interaction, self-driving and other emerging fields, and successively invested in world-class robot and artificial intelligence projects such as Jibo and Exyn in the United States.

A hero is known in the time of misfortune

A group of people from Tsinghua University, all from technical background, deeply understand and understand the subversive effect of AI on the world.

Forty years of Fang Hua, today's Tsinghua AI has been fruitful, in addition to the above heroes, there are too many Tsinghua AI people are making efforts.

From June 20 to June 21, 2020, the third China artificial Intelligence Security Summit will officially kick off in Hangzhou, where many AI people from Tsinghua University will bring wonderful thinking and theme sharing.

With the integration of AI and security, through the deep flow of still water in 2018 and the chaotic fight in 2019, the technical research and program in 2020 will be more clear and mature.

Https://www.toutiao.com/i6825325569647837710/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.