The most comprehensive AI industry application white paper! Complete dismantling of eight application fields
Old friends in the old lane 2021-02-23 03:23:01

Chile stuff ( official account : zhidxcom) Ed | Wise things, internal reference

Artificial intelligence is a kind of leading-edge technology that causes disruptive changes in many fields . Countries all over the world attach great importance to the development of artificial intelligence , The White House of the United States has successively released several government reports on artificial intelligence , It is the first country to raise the development of artificial intelligence to the level of national strategy , in addition to , The British 、 The European Union 、 Japan and others have released AI related strategies 、 A plan of action , Strive to build the first mover advantage of artificial intelligence . China attaches great importance to the development of artificial intelligence industry , General secretary Xi Jinping pointed out in the nineteen major reports , want “ Promoting the Internet 、 big data 、 Deep integration of artificial intelligence and real economy ”, from 2016 It's been a long time since 《“ Internet + Three year action plan of artificial intelligence 》、《 Development plan of new generation artificial intelligence 》、《 Three year action plan to promote the development of new generation artificial intelligence industry (2018-2020 year )》 And other national policies , Positive results have also been achieved , China has gradually formed a computing chip 、 Open source platform 、 Basic applications 、 Industrial chain of artificial intelligence with perfect industrial application and products .

Intelligent internal reference in this issue , We recommend the report from China information and Communication Research Institute 《 Artificial intelligence development white paper industrial application 》, Focus on the analysis of the current AI in the software and hardware support platform 、 Basic products 、 Composite products 、 The current situation in the field of application, etc 、 Problems and Trends . If you want to collect the full report of this article (2018 Application of artificial intelligence industry ), You can reply to the key words in Zhidong's headline number “nc312” obtain .

 The most complete AI White paper on industrial applications ! Eight application areas are completely dismantled

The following is the dry goods presented by intelligent internal parameter sorting :

 The most complete AI White paper on industrial applications ! Eight application areas are completely dismantled

AI Industrial application view

At present, the theory and technology of artificial intelligence are becoming more and more mature , The scope of application is expanding , Industry is taking shape 、 Keep enriching , The corresponding business models are also evolving and diversifying . The application of artificial intelligence industry is from bottom to top , Divided into hardware and software support layer 、 Product layer and application layer .

 The most complete AI White paper on industrial applications ! Eight application areas are completely dismantled

▲ AI industry application view

1、 Software and hardware support layer

This layer includes hardware and software platform . The hardware mainly includes CPU、 GPU And so on , Deep learning 、 Brain like, etc AI Chips and sensors 、 Memory and other perceptual storage hardware , The leading manufacturers are mainly cloud computing service providers 、 Traditional chip manufacturers and emerging AI Chip manufacturer . Software platforms can be subdivided into open platforms 、 Application software, etc , The open platform layer mainly refers to the machine learning development and basic function framework for developers ; Application software mainly includes computer vision 、 natural language processing 、 Human computer interaction and other software tools, as well as the application software developed with these tools .

Diversified innovation of core devices , drive AI Computing industry development .GPU、DSP、FPGA、ASIC And brain like AI Chip innovation is frequent , Support the cloud side 、 End side AI Computational demand . AI The computing industry is developing rapidly , In particular, the demand for cloud deep learning computing platform is rapidly releasing . With NVIDIA 、 Google 、 Foreign enterprises led by Intel speed up all kinds of AI Technological innovation , Cambrian in China 、 Shenjian technology and other enterprises are also following up .

2、 Product layer

The product layer includes basic products and composite products . Among them, basic products include basic language processing products 、 Knowledge map products 、 Computer vision products 、 There are four types of human-computer interaction products , It's a technology product at the bottom of artificial intelligence , It is the foundation of AI terminal products and industry solutions . Composite products can be regarded as AI terminal products , yes AI The carrier of technology , At present, it mainly includes wearable products 、 robot 、 Unmanned vehicle 、 Intelligent speakers 、 Smart camera 、 Feature recognition equipment and other terminals and supporting software .

AI Products come in many forms , It covers hearing 、 Vision 、 Tactile sensation 、 Cognition and so on . Whether it's basic products or composite products , Can support processing text 、 voice 、 Images 、 Perception and other forms of input or output , Products come in many forms , Such as speech recognition 、 Machine translation 、 Face recognition 、 Somatosensory interaction, etc . Global Internet companies are actively laying out various product areas , Strengthen all kinds of products AI Technological innovation , Effectively support various application scenarios .

3、 application layer

The application layer refers to AI The penetration of technology into various fields forms “AI+” Industry application terminal 、 System and supporting software , And then cut into all kinds of scenes , Personalization for users 、 Precision 、 Intelligent services , Deep enabling medicine 、 traffic 、 Finance 、 retail 、 education 、 Home Furnishing 、 Agriculture 、 manufacture 、 Network security 、 human resources 、 Security and other fields .

There are no professional restrictions on AI applications . adopt AI Products are integrated with all fields of production and life , For improving the traditional process 、 Increase of efficiency 、 Improve efficiency 、 Cost reduction and other aspects provide a huge impetus , Dramatically enhance the business experience , Effectively improve the level of intelligence in various fields , Bring change to traditional fields .

 The most complete AI White paper on industrial applications ! Eight application areas are completely dismantled

AI Industry and application development status and trend

The rapid development of artificial intelligence technology , Some technologies have entered the stage of industrialization , Bring the rise of new industries . In terms of industrial scale , 2017 In, the scale of domestic AI market reached 237.4 One hundred million yuan , Compare with 2016 Annual growth 67%. Among them, biometrics 、 Image recognition 、 Video recognition and other technology as the core of the computer vision market is the largest , Proportion 34.9%, achieve 82.8 One hundred million yuan .

Look from industrial structure , AI industry can be divided into basic computing and software platform 、 Core software and devices 、 There are three major applications in the industry , The core software and equipment 、 Industry applications are the fastest growing part .

From the perspective of enterprises , Google 、 Apple 、 Facebook、 Microsoft 、 Baidu and other Internet 、 Mobile Internet companies will AI As the focus of strategic development in the next stage , Speed up basic algorithms 、 Platform and smart device development , Together with universities and scientific research institutes, it has become the main driving force for industrial development ; The upsurge of entrepreneurship and the enthusiasm of investment and financing are 2017 Back to reason , But on the whole AI Innovative enterprises and unicorn enterprises have reached a certain scale ,2016 New start-ups in the world 738 home ,2017 The number of new start-ups dropped to 324 home .

From the perspective of industrial ecology , At present, the ecological model of artificial intelligence industry has not been locked , Various industrial models are being explored . With Google 、 Leading foreign companies, led by Amazon and other companies, focus on chips 、 From operating system to running framework, creating vertical Ecology , And quickly open their own architecture through open source 、 Opening up and other ways of industrial promotion , Strive to form the industry fact standard . Domestic industrial ecology focuses on the framework layer and application layer , Especially the application layer software technology and platform develop rapidly .

1、 Software and hardware support platform

(1) A variety of artificial intelligence chips are innovating rapidly

The wave of artificial intelligence has become a new driving force for the growth of chip market . Based on Prediction , The global AI chip market is 2016 In about 24 Billion dollars , To 2020 The annual scale will be close to 150 Billion dollars , The compound annual growth rate has been maintained at more than 40% High rate ; meanwhile , The proportion of AI chips in the overall AI market will also increase year by year , It is expected to come from 2016 Year of 8% Growth to 2020 Year of 12%.

The industrial system of artificial intelligence chips has taken shape . Artificial intelligence chip refers to the computing chip that can accelerate all kinds of deep learning algorithms . The effect of deep learning algorithm on convolution 、 Matrix multiplication task and memory access operation are more frequent , For those who are better at serial logic operations CPU It's less efficient in terms of computation , It's hard to meet demand . At present, AI chip types mainly include GPU、 FPGA、 ASIC、 Brain like chip, etc . among , GPU The chip is versatile and suitable for large-scale parallel computing , But it's expensive 、 High energy consumption ; FPGA The chip architecture can be flexibly configured by programming to adapt to algorithm iteration, and the energy efficiency is better than GPU chip , But the technical threshold of product development is high , The development ecology is not perfect ; ASIC The chip achieves extreme performance and energy efficiency by solidifying the algorithm , And after large-scale mass production, the cost advantage is highlighted , But the early development cycle is long and easy to face the risk of algorithm iteration . Brain like chip is still in the stage of laboratory research and development .

Leading enterprises speed up the layout of AI chips . NVIDIA relies on high-performance GPU Chip takes advantage of application scale , AMD、 Intel 、 Companies like Google are speeding up their catch-up . NVIDIA's fast launch of AI optimization Tesla GPU Series of products , The strongest of them V100 GPU The chip provides 120 Trillions of tensor computing power , And expand CUDA Ecological development deep learning Accelerator cuDNN, promote GPU Running efficiency of deep learning algorithm and mainstream development framework , Strong hardware performance and perfect and easy-to-use developer ecology help NVIDIA quickly form a huge market advantage , Existing customers cover Google 、 facebook 、 Giant enterprises like Microsoft and a large number of start-ups 、 Scientific research institutes, etc . AMD And speed up the chase , The latest release of the world's first 7nm Process 、 Designed for AI tasks GPU Chip products , Trying to capture the server and workstation market .

meanwhile , Intel 、 Companies like Google Develop energy efficient and cost-effective ASIC Chips build competitiveness . Google has developed its own artificial intelligence to meet the needs of Google's cloud business ASIC Series of chips TPU, among , The training chip has the ability to achieve the industry's highest per second 180 Trillions of peak floating point computing power ,TPU Chips are also associated with TensorFlow Development framework 、 The algorithm is deeply coupled with Google cloud platform to build a vertical and complete industrial ecology ; Intel buys chip startups Nervana master ASIC Training chip technology , The second generation of products will be 2019 It was officially launched in the second half of the year , Performance benchmarking Google TPU product .

(2) Multi party layout of AI computing framework

Basic development framework plays a key role in the artificial intelligence industry chain . In the era of mobile Internet , Android The system goes through GMS Loose coupling with downstream cloud services , Through version control and upstream chip 、 Complete machine manufacturer tight coupling , To achieve Android Closed loop ecology of mobile Internet based on operating system . In the age of artificial intelligence , Development framework is comparable Android Core position of operating system , With the pace of leading industrial progress 、 Drive hardware configuration 、 The core role of collaborative development of terminal scenarios and cloud services , Occupy the key position connecting the preceding and the following . With Google Deep learning development framework TensorFlow For example , TensorFlow Closely bound with Google cloud , Provide cloud machine learning service in cloud platform mode , Downward close coupling with chip and hardware manufacturers for customization and optimization , Google TPU Dedicated to TensorFlow .

Leading enterprises present diversified development mode around the development framework platform . One is the vertical access mode , From hardware to open source platform to cloud platform to application service , Through the upstream and downstream of the industrial chain , Build the whole industry ecology , Google is a typical example ; The second is to lay out the industry application service mode upward , Business oriented , Through the core platform to lay out key industry applications , Amazon 、 Ali et al ; Third, the algorithm sinks into the hardware mode , The core algorithm is fixed in hardware , Provide industry general or special computing power in the form of hardware , Like Cambrian ; The fourth is to open the basic ability with the core platform , Provide basic capabilities for the industry , For example, iFLYTEK provides basic voice recognition technology for the industry , Shangtang provides basic technology of face recognition for the industry . Among the four development models , All data generated by cloud platform and application services are returned to the training platform for data feedback , It can effectively improve the comprehensive ability of the platform .

The international giant open source AI development framework aims to speed up the initiative of technology industry organization . International giants are laying out development frameworks one after another , The intention is to speed up the control of the initiative of technological industry organization , Capture customers 、 Applications and data resources , Gradually establish new industrial structure and technical standards .2013 year , Berkeley University Dr. Jia Qingyang announced the open source deep learning framework Caffe, Become the first mainstream industrial deep learning tool .

2015 year 11 month ,Google Open source deep learning framework TensorFlow, Have the basic algorithm of deep learning , It can meet the requirements of graphic classification 、 Audio processing 、 Recommendation system and natural language processing , Become GitHub The most popular machine learning open source project , At present, it attracts ARM、 Jingdong and many other partners .2016 year , Amazon announces MXNet As its official support framework , It has excellent distributed computing performance , Own Carnegie Mellon 、 Intel 、 NVIDIA and many other partners , Domestic companies such as Tucson Internet and horizon also use . 2015 year 11 month , IBM Announce open source machine learning platform SystemML, According to the characteristics of data and cluster, rules based and cost based optimization techniques can be used to dynamically compile and optimize , Applied in different industries . 2016 year 9 month , Baidu open source its deep learning platform PaddlePaddle, Can provide machine vision 、 natural language understanding 、 Search engine ranking 、 Recommendation system and other functions . 2017 year 6 month , Tencent and Peking University 、 High performance distributed computing platform jointly developed by Hong Kong University of science and technology Angel Official open source , It has strong fault-tolerant design and stability . Many open source learning frameworks promote the development of AI applications . According to the IDC forecast , To 2020 year , 60% 's AI applications will run on an open source platform .

2、 Basic products of artificial intelligence

(1) Natural language processing products present the trend of practical development

natural language processing (NLP) It means that machines understand and interpret human writing 、 The ability to talk , It's part of artificial intelligence and Linguistics , It is dedicated to using computers to understand or produce words or sentences in human language . Natural language processing mainly involves speech recognition 、 speech synthesis 、 Semantic understanding 、 Machine translation , Natural language products present the trend of practical development , But there is still room for improvement in product maturity .

Speech recognition is widely concerned by business and academic circles at home and abroad , It can approach the human level without noise and accent interference . At present, the technology maturity of speech recognition is relatively high , Has reached the 95% The accuracy of , But background noise is still difficult to solve , The practical application is limited to close range use . The research level of speech recognition technology in China is good , Basically synchronized with foreign countries , IFLYTEK speech recognition success rate reached 97%, Offline recognition rate is also up to 95%. Besides , China has its own characteristics and advantages in Chinese speech recognition technology , It has reached the international advanced level . Voice recognition products , Microsoft 、 Google 、 Amazon , And Baidu in China 、 IFLYTEK 、 Companies such as Spitzer have launched their own products based on voice interaction , Among them, the input method 、 Car voice 、 Smart home 、 Educational assessment is the most common .

Machine translation is the most popular application direction , Because of the complexity of natural language semantic analysis , The level of translation is far from human . In recent years, machine translation technology has become more and more mature , All the major manufacturers are actively engaged in the field of machine translation , Google uses deep learning technology , It significantly improves the performance and quality of translation . Major Internet companies have launched their own translation systems , Google 、 Microsoft 、 youdao 、 Hkust xunfei 、 Baidu 、 Sogou and others have launched or updated translation products . For example, Alibaba machine translation is based on Alibaba's massive e-commerce data , And combined with machine learning 、 Natural language processing technology , Realize the function of multi language recognition and automatic translation , Provide accurate information for cross-border e-commerce information localization and cross language communication 、 quick 、 Reliable online translation service .

(2) Knowledge mapping presents multi-dimensional application from practical problems

The concept of knowledge map is developed by Google 2012 Formally put forward in , Its original intention is to improve the ability of search engine , Improve users' search quality and search experience . Knowledge map is a knowledge base with digraph structure , The nodes of the graph represent entities or concepts , And the edges of a graph represent entities / The various semantic relations between concepts , Its origin can be traced back to 20 century 50 The semantic web of the age , The essence is to make the machine store information in a way close to natural language semantics , So as to enhance the ability of intelligent information retrieval , It has been widely used in intelligent search 、 Intelligent q&a 、 Personalized recommendation, etc .

Knowledge mapping has experienced a process from manual and group cooperation to automatic acquisition by machine learning and information extraction technology . The early knowledge map was mainly obtained by manual processing , In English WordNet and Cyc project . By manual processing , Knowledge mapping processes millions of pieces of knowledge into forms that machines can understand , Give machines the ability to judge and reason . With the establishment of the largest swarm intelligence knowledge base Wikipedia on the Internet , There is DBpedia、YAGO as well as Freebase And so on, relying on the knowledge map established by large-scale collaborative cooperation . With the coming of big data Era , The data source of knowledge mapping is no longer limited to the semi-structured data of encyclopedia and various types of network data .

The service and application based on knowledge map is the research hotspot of artificial intelligence . At present , The application of knowledge mapping can be summarized as semantic search 、 Knowledge Q & A and big data analysis and decision-making based on knowledge :

1、 In terms of semantic search , Because of the well-defined structure of knowledge map , Semantic search uses large-scale database to annotate keywords and document content , To improve search results . Foreign search engines are Google search and Microsoft Bing The most typical . One side , Search engines based on knowledge mapping have been integrated into Wikipedia 、 CIA World Survey and other public resources . On the other hand , Search engine and Facebook、 Twitter And so on , In the collection of personalized content 、 There are significant advantages in customization . In recent years, the domestic mainstream search engine companies have shifted the research of knowledge mapping from concept to specific product application . sogou “ Know Cube ” It is the first knowledge map product in the domestic search engine , It works by integrating fragmented semantic information , Logical recommendation and calculation of user's search , And feed back the core knowledge to users . Baidu named the knowledge map as “ Intimate ”, It is mainly committed to building a large general knowledge network , Show all aspects of knowledge in the form of pictures and texts .

2、 In terms of Knowledge Q & A , The question answering system based on knowledge map makes semantic analysis and grammatical analysis on the questions raised by users using natural language , Then it can be transformed into structured query statements , Then search the answers in the knowledge map . at present , Knowledge mapping has been introduced into various Q & A platforms at home and abroad , For example, Apple's intelligent voice assistant Siri Be able to provide users with answers 、 Introduction and search service ; Natural language assistant acquired by Amazon Evi, use True Knowledge Engine development , Similar Siri Service for . Small robot developed by domestic Baidu company 、 Xiaomi smart audio 、 Alibaba, tmall, genie and others have introduced knowledge mapping technology , Start offering interactive Q & A Services .

3、 In terms of analysis and decision making , Knowledge mapping can be used to assist big data analysis and decision-making in industries and fields . For example, in the aspect of stock investment and research intelligence analysis , Through knowledge mapping technology from the prospectus 、 Annual report of the company / Notice 、 Research Report of securities companies 、 Automatic extraction of company related information from news and other semi-structured text data , This map can be used to make deeper analysis and better investment decisions in a macroeconomic event or enterprise emergency . at present , Goldman sachs 、 JP morgan 、 Citibank and other international famous investment banks have carried out relevant exploration and application . The United States Netflix It also uses the registration information and viewing behavior of its subscribers to build a knowledge map , Analyze user preferences to launch new online dramas .

(3) The coordinated development of technology industry promotes computer vision to realize commercial value

Computer vision refers to the perception and cognition of images through electronic means , To achieve or even surpass the effect of human visual intelligence , It's one of the most interesting directions in the field of artificial intelligence . Although there are still a lot of unsolved problems in computer vision at the present stage , But thanks to the maturity and application of deep learning algorithm , Image classification and recognition as the representative of computer vision products focusing on perceptual intelligence has been widely used in security 、 Finance 、 Retail and other industries , Help related industries upgrade to the direction of intelligence .

The rapid development of neural network and deep learning has greatly promoted the development of computer vision , Large scale neural networks have achieved excellent results in some subdivision fields of computer vision .2017 year ImageNet In the last image classification competition , Classification algorithm based on large neural network in image classification (1000 class ) Tasks , take TOP5 The error rate of classification is reduced to 2.25%, It's far ahead of the human eye's ability to classify . 2018 In ActivityNet In the video understanding contest , Baidu team is Kinetics The average error rate is reduced to 10.9%, The related technology has been applied to the real online video classification system , Tag the video 、 Business scenarios such as video comparison and video push provide semantic parsing functions .

Computer vision products have been used in security 、 Finance 、 Internet 、 retail 、 Medical care 、 Mobile and entertainment industries are gradually exporting business value . In Finance 、 Move 、 Security and other industries , Face recognition is a computer vision product with high commercial maturity , It is widely used in account authentication 、 Cell phone brush face unlock 、 There are many scenes such as automatic statistics of people flow and screening of specific people . In the Internet 、 retail 、 The mobile industry , Image search products can provide users with more convenient visual search ability .

for example : eBay On 2017 year 10 A new reverse image search tool was added to its shopping platform in May , To help users find product items using existing photos ; Google company 2018 year 3 In May, it announced that Google Lens Image search service can be applied to android and IOS( adopt GooglePhoto project ) A smart phone , The service can view the surrounding environment through the mobile phone camera and provide users with relevant situational information . In the medical industry , Computer vision can provide the ability of early pathological screening in clinical treatment . The Department of Radiology and biomedical imaging at the University of California, Berkeley and the radiology big data team are working on the diagnosis of early Alzheimer's disease , Through computer vision technology in small-scale testing ( from 40 Of patients 40 A separate test of imaging ) in , The average incidence of the disease is more than 6 The detection rate of Alzheimer's disease cases reached 100%.

The global computer vision industry is developing rapidly , Computer vision companies are springing up . according to MarketsandMarkets According to the report , 2017 The global market scale of computer vision based on artificial intelligence was 23.7 Billion dollars , expect 2023 Annual meeting 253.2 Billion dollars . Forecast period (2018-2023) The compound annual growth rate within the year 47.54%10. A large number of computer vision companies have sprung up in the market , Among them, Google 、 Microsoft 、 Amazon as the representative of large multinational technology enterprises, in addition to the field of computer vision , We are also actively engaged in all fields of the AI industry .

Although Chinese enterprises started late in the field of computer vision , But it's growing fast , A number of Unicorn enterprises with a market valuation of 10 billion yuan have emerged . for example : Founded on 2014 Shangtang technology in 2000 , Widely serve security 、 Finance 、 Mobile and other industries , Customers include Qualcomm、 Ying Wei Da 、 UnionPay 、 Huawei and other well-known enterprises and government agencies .2017 year 7 month , Shangtang technology announced the completion of 4.1 Billion dollars B Round of funding , It set the highest single round financing record in the field of artificial intelligence at that time . 2018 year , Shangtang technology is in 4 The month and 5 It has been announced in consecutive months that 6 Billion dollars C Round financing and 6.2 Billion dollars C+ Round of funding . Founded on 2015 The cloud of 2000 starts with technology , Dig deep into security 、 Bank 、 Airport and other key industries , I worked with the Ministry of public security 、 Four big banks 、 CAAC and other industries set up joint laboratories . 2017 year 11 Yueyun is officially completed from science and technology B Round of funding , Total gain 25 Billion yuan of development funds . Founded on 2014 Year of Malone Technology , For Jingdong 、 Vipshop 、 Coca Cola 、 Mengniu and other retail enterprises provide commodity attribute identification 、 Commodity image retrieval service . 2017 year 11 Yuemalong technology completed the investment led by Softbank China 2.2 A hundred million yuan B Round of funding , Become the first AI company invested by Softbank China in China .

(4) Human computer interaction products have been implemented in many fields

Human computer interaction mainly studies the information exchange between human and computer , According to the interaction mode, it can be divided into voice interaction 、 Emotional interaction 、 Body feeling interaction 、 Brain computer interaction . at present , Human computer interaction has achieved certain research results , Depending on different human-computer interaction technologies , A lot of products have come out , And covers many areas . But on the whole , Affected by voice 、 Vision 、 Semantic understanding and other technical conditions , The human-computer interaction industry is still in its infancy . Facial expression interaction has been preliminarily applied in mobile application product design , For example, by Takuto Onishi Developed iOS Applications “twika^o^”, It can help users transform the real facial expression of characters into literal symbol expression . Somatosensory interaction is in its infancy , It is mainly used in smart home 、 Somatosensory games, etc , Users can use their own body movement to control smart home devices , Kinect I've been working on somatosensory games , There are also related products in China , For example, Su Meng Xiang Dong 、 Green move 、 Sports and so on , But there are different levels of development in the aspects of effect and experience .

The development of human-computer interaction , Experienced PC Time 、 Mobile Internet Era , Now we have entered the era of intelligent life . PC The main interaction mode of the times is keyboard + mouse , The main way of interaction in the mobile Internet era is touch 、 Handwriting and gestures , In the age of intelligent life, the way of interaction begins to move towards voice and vision . The history of human-computer interaction , It's a process of natural interaction —— From machine centered human-computer interaction , Towards human centered natural interaction .

The development of voice assistant in the field of artificial intelligence has been relatively perfect . According to market research institutions Strategy Analytics Data show that , 2017 year , GoogleAssistant Dominates the smartphone voice assistant market , by 46%, Apple Siri The second , Occupy 40.1%, Baidu DuerOS And Samsung Bixby Accounting for 13%. 2019 More than half of the world's smartphones will have voice assistants in 2010 , Even to 2023 year , This share will grow to 90%.

at present , Intelligent voice assistant is still in the early stage of intelligent application , Just as a built-in or user downloaded APP For user's use , It doesn't have a killer effect in practice . Smart voice assistant usage 、 Activity rate 、 Retention rates are low , Even if Siri No exception . The voice interaction output of intelligent voice assistant is unable to show the rich information of pictures in many scenes , The input feedback output of a voice is less , High frequency interaction is necessary to improve the recognition rate . From the perspective of application direction and scene , Voice assistant is mainly used in consumer products and professional industry applications , Consumer market is mainly used in daily life such as clothing, food, housing and transportation , Such as mobile phone 、 Smart car 、 Smart home 、 Wearable devices and so on , Professional applications are mainly used in medical treatment 、 education 、 call center 、 Court trials and other specific scenes .

Brain computer interaction will help artificial intelligence move towards human intelligence . In the research of brain computer interaction abroad ,“ Implantable ” The beauty of technology 、 Holland leads , The United States has made a breakthrough in human-computer application research .“ Non implantable ” Technology explores the market , Products come out one after another , For example, the Japanese Honda company produced the idea control robot , The operator can control the robot around him by imagining the movement of his own limbs . A study by the University of Rochester in the United States , Subjects can pass P300 The signal controls some objects in the virtual reality scene , For example, switch lights or operate virtual cars .

3、 AI composite products

(1) Biometric technology continues to integrate into various fields

Biometric products mainly refer to a kind of products that carry out identity authentication through human biometrics . Human biological characteristics are usually unique 、 Measurable or automatic identification and verification 、 Hereditary or unchangeable for life , Therefore, biometric authentication technology has more advantages than traditional authentication technology . By sampling biological characteristics , Extract its unique features and convert them into digital code , And further these codes form feature templates . Biometric products include things like fingerprint identification 、 Face recognition 、 Iris recognition 、 Finger vein recognition 、 Voiceprint recognition and eye print recognition, etc .

Fingerprint identification technology is the most mature and the lowest cost biometric technology . It has the highest proportion in the biometric technology industry , But with the development of other recognition technologies , The proportion is decreasing year by year . Fingerprint recognition is based on the analysis of global and local characteristics of fingerprint , Like a ridge 、 valley 、 End 、 A point of bifurcation or divergence , And then compare them to confirm a person's identity . Capacitance technology is the most commonly used fingerprint acquisition technology . By pressing the ridge and valley of the finger on the acquisition head, different capacitance is generated between the finger skin and the chip , The chip gets complete fingerprint information through testing . Germany IT Werke The company 2011 It released a “ Fingerprint payment ” Software , It's a set of just “ Fingerprinting ” Then you can complete the payment of new software . This convenient “ Fingerprinting ” The payment service is already in some supermarkets in southwest Germany 、 Bars and even school canteens . Famous supermarket chain in Germany Edeka Supermarket survey data shows that , There are about 1/4 Our customers are willing to choose “ Fingerprint payment ”.

Face recognition is based on the distance between facial features and facial organs 、 angle 、 A series of parameters are quantized to identify the size and shape . Because face recognition is easy to use and suitable for public safety and other multi crowd areas , It is widely used in smart home 、 Mobile phone recognition and online face verification , Its proportion is gradually rising . 2010 year 5 month , Shanghai World Expo used “E Miantong ” Face recognition system , About 50 Ten thousand certificate holders and 7000 Ten thousand visitors have used the “ Face pass ” . But face recognition involves many organs 、 Large area , So its recognition is very complicated , The accuracy of face recognition is relatively high , But it costs a little more than other recognition technologies .

Iris recognition technology is to use the characteristics of iris lifetime invariance and difference to identify identity . Because each iris contains a unique image-based crown 、 Crystalloid 、 Filaments 、 speckle 、 Pit 、 The structure of wrinkles and stripes . Theoretically , Iris's permanent , The false recognition rate of iris recognition is 1/1500000, Higher than fingerprint recognition 1/50000, High security , More suitable for “ password ”. For example, United Bank of Texas has applied iris recognition system to depositor identification , Depositors do not need a bank card to handle banking business , No need to recall the password —— adopt ATM A camera on the computer first scans the user's iris , Then the scanned image is converted into digital information and checked with the data in the database , Can realize the user's identity authentication . But due to the high security but high cost of iris recognition , Popularization will take time , At present, it is mainly used in bank vault encryption 、 Military, national defense and other fields .

Voiceprint recognition passed the test 、 Waveform and change of collected sound , Match with registered sound template . This is a non-contact recognition technology , It's very natural . however , The range of sound changes is very large , The volume 、 Speed 、 The change of sound quality will affect the results of acquisition and comparison . But by recording or synthesizing , It's easy to fake a voice , Poor safety , Now it's used in social security 、 Public security, criminal investigation, mobile phone lock screen and other fields .

In recent years , As countries around the world pay more attention to the security field , Identification technology and products are gradually becoming mature and perfect , Biometrics has ushered in a period of rapid development , Face recognition 、 Iris recognition 、 Biometric technologies such as vein recognition are developing rapidly , The market application scenarios are broad , The proportion of products is increasing . at present , The proportion of fingerprint identification products has increased from 90% Down to less than 60%, Biometric industry is developing in the direction of diversification .

Biometric industry chain tends to be perfect , The size of the market is growing rapidly . In our country , The number of biometric enterprises is growing rapidly , The scale of enterprises is increasing , The biometric market is growing explosively . At present , Enterprises in the field of biometrics have developed from 20 Yu family has developed to 200 More than , The market scale has also reached billions of yuan 12. Face recognition as an example , At present, it has formed a complete industrial chain of various industrial roles, including face recognition algorithm research enterprises and so on . At present, with the gradual introduction of e-passport , Security issues have received further attention , There is still much room for the development of biometric industry in China , The scale of the industry is expected to further increase in the future .

The biometric industry is diversified , The field of security has become a hot spot . at present , In China's biometric industry , Fingerprint technology and products still dominate , But with face recognition 、 Iris recognition 、 Vein recognition 、 Voiceprint recognition and other technologies are developing rapidly , All kinds of biometric products and market potential should not be underestimated . At present, people pay more attention to security , There are a number of security products, such as biometric access control , In the future, biometric technology will be gradually used in the field of security to improve security performance .

(2) Intelligent transportation products represented by autonomous driving are developing rapidly

The main applications of intelligent carrier products are automatic driving 、 Unmanned aerial vehicle (uav) 、 Unmanned ships, etc , At present, the application of intelligent carrier products is in the stage of rapid development , The development of UAVs and unmanned ships is relatively mature , It has been preliminarily applied , And autonomous driving is still in the development and experimental stage .

According to the U.S. Highway Safety Administration (NTHSA) The definition of , Auto driving can be divided into four stages . At present, the high-level autonomous vehicle is still in the research and experimental stage , No industrialization . In the past two years , The major automatic driving enterprises have announced the schedule to realize the mass production of automatic driving , Most of them focus on 2020-2025 In between . Level-2 Class of autonomous vehicles , Advanced assisted driving (ADAS) The vehicles have been put into mass production . 2017 In the world ADAS The market size is 300 Around $100 million , And showing a steady growth trend . With the acceleration of the trend of intelligent vehicles and the improvement of safety requirements , Global future ADAS Market penetration will increase significantly . To 2020 year , The global ADAS Permeability is expected to reach 25%, Global new cars ADAS The carrying rate is expected to reach 50%.

Automatic driving can be divided into “ Gradualness ”、“ revolutionary character ” Two technological routes . At present, the field of automatic driving can be roughly divided into two routes according to the technologies adopted by the entry enterprises , One is Ford 、 BMW 、 Audi and other traditional car companies adopt “ Gradualness ” Route , That is to gradually add some automatic driving functions to the car , With the camera 、 Navigation maps and all kinds of sensors , Provide automatic emergency braking for the driver 、 Panoramic parking 、 Adaptive cruise and other auxiliary driving functions .

The second is Google 、 Baidu and other Internet technology giants use “ revolutionary character ” Route , By using lidar 、 High definition map and artificial intelligence technology directly realize the purpose of unmanned driving , Emphasis on product innovation and convenience . Google predates 2009 We've been laying out autonomous driving since , Become the first company to get the U.S. government road test license , Its self driving vehicles Waymo Completed 800 Ten thousand kilometers of self driving road test mileage , The technical level keeps the leading position in the world . Tesla in 2015 The first generation of Autopilot automobile , For the world's first mass production self driving vehicle .

Baidu in 2013 The driverless vehicle project started in , Its driverless car has obtained the first batch of automatic driving license plates in China , 2018 Baidu Apollo The world's first model produced in cooperation with Jinlong bus L4 Class a driverless bus “ Apollon ” It has been officially put into mass production . Except Google 、 Baidu 、 Outside Tesla , Intel 、 Apple 、 Uber And other technology giants also carry out layout in the field of unmanned driving . Intel buys Altera as well as Mobileye after , To start L4 Level autonomous driving technology research and development . 2018 NVIDIA announced its Drive PX Its latest products Xavier And the next generation of products in the future Pegasus, And based on Xavier Together with Bosch and ZF, respectively, they launched the vehicle AI Supercomputers . Chip giant Qualcomm , After the acquisition of NXP , On 2017 year 12 At the beginning of this month, he obtained the license of automatic driving test in California .

UAVs are mainly military UAVs , Demand is on the rise . As UAV R & D technology matures , The cost of manufacturing has been greatly reduced , UAVs are widely used in various fields . UAVs are mainly divided into military UAVs according to their application fields 、 Industrial drones 、 Consumer drones . The main applications of military UAV are investigation 、 Electronic warfare 、 Drones, etc , Industrial UAVs are mainly used in agricultural plant protection 、 Power inspection 、 Police law enforcement 、 geological prospecting 、 The environmental monitoring 、 Forest fire prevention and other fields , Consumer UAVs are mainly used in personal aerial photography 、 Film and television aerial photography and remote control toys, etc . 142017 The scale of UAV market will reach 60 Billion dollars , and 2020 It will further increase to 112 Billion dollars . Global UAV production will reach 300 Ten thousand frames , Year on year growth of up to 39%, Among them, the sales volume of consumer UAVs will account for 20% 94%, But it only accounts for the sales of UAVs 40% about .

Consumer UAVs are still in their infancy , Autonomy still needs to be improved . At present, some consumer UAVs can pass through sensors 、 Camera, etc. for automatic obstacle avoidance , At the same time, it can also rely on machine vision to detect the flight environment , Analysis of the characteristics of the environment to achieve self planning path . 2016 year , Intel It's successfully implemented through intelligent algorithms 500 A multi rotor UAV staged a light show of air formation , Consumer UAVs are moving towards a higher level of UAV intelligence . China is the largest UAV manufacturer in the world , Dajiang accounts for the global consumption of UAVs 70% Consumer UAV market share , However, according to the UAV system roadmap standard , The technology level of Dajiang consumer UAV is still in the primary stage .

(3) Intelligent robot technology and product innovation are active

From the perspective of application , Intelligent robots can be divided into industrial robots 、 personal / Home service robots 、 Public service robots and special robots . among , Industrial robots include welding robots 、 Spraying robot 、 Carrying robots 、 Processing robots 、 Assembly robots 、 Cleaning robots and other industrial robots .

personal / Home service robots include home service robots 、 Education and entertainment service robot 、 Robots for the elderly and the Disabled 、 Personal transport service robots and security monitoring robots, etc .

Public service robots include hotel service robots 、 Bank service robot 、 Venue service robots and catering service robots, etc . personal / Home service robots and public service robots can also be referred to as service robots .

Special robots include special extreme robots 、 Rehabilitation robot 、 Agricultural robots 、 Underwater robots 、 Military and police robots 、 Electric robots 、 Petrochemical robots 、 Mining robots 、 Construction robots 、 Logistics robots 、 Security robots 、 Cleaning robots and medical service robots, etc .

The industrial robot market is highly concentrated , Robot is the most widely used industry . according to IFR( International Federation of Robotics ) Published data , 2017 year , Industrial robots occupy as much as 63.4% Market share , The most vigorous development . in 、 Han 、 Japan 、 beautiful 、 The five German countries 2017 The sales of industrial robots accounted for... Of the total sales in the world 71%. Among them, China's industrial robot sales reached 13.8 Ten thousand units , The second is South Korea 4 Ten thousand units , Japan about 3.8 Ten thousand units , About the United States 3.3 Ten thousand units , Germany is about 2.2 Ten thousand units . New industrial robots can replace manual labor in the heavy manufacturing process , In professional metal processing automation, it can be used for the production of metal devices , Carry 、 Palletize , It also has an intelligent service kernel 、 Learning “ The brain ”, In the process of training and practice, the machining accuracy of metal products can be continuously improved .

The rise of artificial intelligence has promoted the intellectualization of domestic industry , personal / Home robots are more widely used . A leading enterprise in the domestic industry “ Housekeeper help ” Launch intelligent housekeeper robot for home service , Can achieve voice interactive control, complete housekeeping service online order 、 Make a phone call 、 Home defense 、 Family care 、 Health monitoring 、 Remote monitoring 、 Take the initiative to remind 、 Home entertainment 、 Enlightenment early education 、 Emergency alarm 、 Language learning and many other services , It's a playmate for children and a close guardian for the elderly . Softbank of Japan sells humanoid robots , Have the ability to learn , Can express emotions , Will talk , Can take care of infants and patients , Even company at parties . They can use cloud computing to share data , So as to develop their emotional ability , But they don't share the owner's personal information . Intel's 3D Printing robots , Besides walking 、 speak , It can also help the host to tweet 、 Translation language , Or open the fridge for drinks . The sweeping robot developed by Xiaomi company in China can detect obstacles and indoor terrain independently , Realize the automation of indoor cleaning .

The public service robot is in the hotel 、 Finance 、 telecom 、 Electric power 、 Logistics and other industries with large-scale intelligent service demand are widely used , On the basis of low investment to provide high quality and efficient services for enterprises .

Miklimi's intelligent hotel service robot can automatically learn the Hotel Channel 、 Elevator and room location , Automatic construction of virtual electronic map for navigation , Determine the way to walk , It can avoid people and obstacles automatically , And can take the elevator automatically . To complete all kinds of services without company , Reduce the hotel labor costs while improving operational efficiency .

i Intelligent customer service robot is a new intelligent tool , Sure 24 Reply to user's questions online in real time , As an effective complement to manual customer service . At present, China Merchants Bank 、 Ping An Bank 、 China Construction Bank and other banks and China Unicom 、 China Mobile and nearly a thousand other companies have reached cooperation . In the field of warehousing and logistics , Have the ability to carry 、 Palletize 、 Intelligent robots with sorting and other functions , It has become a hot spot in the logistics industry .

2012 Amazon in 6.78 Billion dollars to buy automated logistics providers Kiva After our robot warehousing business , Use the robot to handle the inventory and distribution of goods in the warehouse . All employees only need to count or distribute goods in a fixed location , and Kiva The robot is responsible for transporting the goods ( Along with the shelves ) Move together in front of the staff .

Starship The company has launched a special delivery system for small goods “ Box robot ”, Its hardware is equipped with a series of cameras and sensors , It can ensure the safety of walking on the sidewalk , Depart from the logistics center at the appointed time , Through the streets , Come to the customer's door to complete the express delivery task . In the delivery process , All the packages are sealed off , The recipient can only open it through his smartphone .

Ali's self-developed robot “ Cao Cao ” After receiving the order , It can quickly locate the location of goods in the warehouse , And planning the optimal picking path , After picking the goods, it will automatically send the goods to the packing table . stay 2018 year 618 During the Shopping Festival , JD.COM 、 Ali rookie 、 Shunfeng and other logistics enterprises actively apply in warehouse robots 、 Sorting robots and other intelligent devices , Improve the intelligent level of warehouse automation .

The intelligent level of special robots is constantly improving , Replace human beings to complete the difficult work under the special environment . In the medical field , Domestic surgical robot “ Tianji ”, In orthopedic surgery has entered clinical practice , It can effectively reduce the spinal cord injury caused by manual operation in orthopedic surgery 、 Risk of vascular injury . In the post diagnosis rehabilitation process , With lightweight 、 Rehabilitation robots with high flexibility are gradually applied and promoted .

Shanghai Jinghe robot company launched a multi position intelligent rehabilitation robot system Flexbot, It is applicable to the rehabilitation department of medical institutions at all levels 、 The orthopaedic 、 Internal Medicine-Neurology 、 Brain surgery and other related clinical departments , For clinical gait analysis , With robot gait training 、 Virtual walking interactive training 、 Gait analysis and rehabilitation assessment .

In the field of agricultural special robots , American investment company Khosla Ventures According to the report of , Special agricultural robots can distinguish crops and weeds by themselves , Select and spray weeds with special herbicides , It can reduce pesticide pollution 20%, At the same time, reduce the cost of Planting .

The technology level of China's intelligent robot industry continues to improve . Industrial Robotics , Xinsong 、 New star 、 Yunnan Kunchuan 、 Beijing Machinery Technology Co., Ltd. leads the first echelon of local industrial robots , Related products are gradually recognized by the market . Xinsong group applied artificial intelligence and virtual reality technology to the first machine in China 7 Cooperative robots with degrees of freedom , Fast configuration 、 Traction teaching 、 Visual guidance 、 Collision detection and other functions . Service Robotics , The intelligence level of service robots in China is comparable to the international advanced level , Shenzhen qihan technology has emerged 、 Shenzhen Yuejiang and other competitive innovation and entrepreneurship enterprises . In the field of special robots , Open mindedness 、GQY video 、 Enterprises such as Heinze are active in innovation , The level of technology is constantly improving , Positioning indoors 、 High precision positioning, navigation and obstacle avoidance 、 A breakthrough has been made in the field of rapid identification of dangerous goods on automobile chassis .

(4) Smart devices have a vast market in the future

The integration of artificial intelligence and wearable intelligent devices brings a new technological experience . Wearable devices include smart watches 、 Smart glasses 、 Smart clothes 、 Pedometer and other product forms , By using perception 、 distinguish 、 wireless communication 、 Big data and other technologies to achieve user interaction 、 Life entertainment 、 Health care and other functions , Provide a perfect technology experience for the wearer . Wearable smart devices will become a part of people , As the carrier of the sensor , Further supplement and extend the human perception ability , Realize people 、 machine 、 The cloud is more advanced 、 Seamless interaction , Realize situational awareness .

The wearable market is in its infancy , Product homogeneity is serious . Global wearable devices will continue to grow rapidly , According to market research ABI Research data display , 2018 The global wearable device market will reach 4.85 One hundred million units , Market Research Organization IHS expect , 2018 Annual sales will reach 336 Billion dollars , The compound annual growth rate is as high as 22.9%. Wearable intelligent devices are widely used in many social fields , In medical treatment 、 Financial payments 、 Identity authentication even plays an important role in industry .

For now , The wearable market is still in its infancy , The apple 、 samsung 、 After Huawei and other enterprises enter the field of intelligent wear , konka 、 More and more enterprises, such as Lenovo, have begun to focus on segmentation , And launched related products one after another , Like Samsung Galaxy Gear Intelligent Watch 、 Epson smart watch PS-500 etc. . Domestic manufacturers are also actively planning , If the smart watch of shell Electronics Geak Watch、 Baidu United TCL released Boom Band Bracelet 、 Huawei TalkBand B1 etc. . However , At present, the homogeneity of smart wearable market is serious , Many products are painless and not just needed , Practicality is hard to satisfy , Consumers are not very dependent on wearable devices . For example, there are many kinds of health bracelets , The core function is step measurement 、 Monitoring sleep, etc .

The smart speaker market has entered the fast lane of development . As a part of smart home , The unique human-computer interaction function of smart speaker can become the entrance terminal in the field of smart home , The popularity of smart home promotes the rapid development of smart speaker industry .

from 2014 Amazon 2008 Echo Released so far ,2017 The scale of global smart speaker market has broken through 120 One hundred million yuan . according to StrategyAnalytics According to the published research report , 2017 The annual shipment of smart speakers reached 3200 Thousands of , Year-on-year growth exceeding 300%. according to an uncompleted statistic , In recent years, more than 500 Three companies start to lay out the smart speaker market . The whole intelligent speaker industry chain covers hardware manufacturers such as chips and microphones 、 Voice technology service providers 、 Content provider 、 OEM/ODM Suppliers and Internet companies . With the development of smart speakers , The industrial chain will realize “ Hardware + Software + Content + service ” The integration of resources , Gradually form an ecological closed loop . Smart speaker manufacturers open up software and hardware technologies such as voice recognition and microphone 、 Rich voice service skills 、 Expand smart device connectivity , Continuously improve the intelligent voice ecology , It also provides conditions for enterprises to make profits by bundling content and services , Drive smart speaker sales growth .

The intelligent level of smart camera has been improved rapidly , Broad market prospect . Smart cameras are the largest blue ocean in the civil security market , In addition to traditional security enterprises , Include 360、 millet 、 The Internet, including Konka 、 Home appliance companies have released smart camera products . With Google to 5.55 A $100 million acquisition of the American home surveillance camera startup Dropcam, The concept of home monitoring products has been detonated , Start to set off a wave of smart cameras in China . In terms of market share , 360、 ZTE smart camera 、 Little ants 、 Fluorite 、 Le orange 、 Lenovo looks at the family treasure 、 Joanne 、 Fushikon and others occupy most of the domestic market . By embedding intelligence SOC chip 、 GPU And other hardware and structural analysis 、 Deep learning and other machine vision algorithms , The intelligent level of smart camera is constantly improving . At present, the mainstream smart cameras generally have the ability of behavior analysis 、 Anomaly detection 、 Identification detection 、 Statistics and other functions , Take Haikang as an example “ Deep eyes ” For deep learning represented by camera built in GPU processor , Using deep learning algorithm in the front of the camera can extract target features , Form deep image data for learning , Greatly improve the detection rate of the target .

4、 Application of artificial intelligence in various fields

(1) The energy-saving effect of each link of artificial intelligence enabled medical treatment is beginning to show

In recent years, with the deepening of medical data digitization , Breakthrough of deep neural network learning algorithm and improvement of chip computing ability , The second wave of AI application in medical field , Has penetrated into disease risk prediction 、 Medical imaging 、 Auxiliary diagnosis and treatment 、 Virtual Assistant 、 Health management 、 Pharmaceutical research and development 、 hospital management 、 Medical insurance cost control and other links , And achieved initial results .

beautiful 、 Britain 、 The governments of Japan and other countries attach great importance to the application of artificial intelligence in the medical field . The United States 《 Health Insurance Portability and Liability Act 》 Cleared the way for AI applications , FDA( Food and drug administration ) The implementation of “ Digital health innovation action plan ”, Reconstruct the digital health product supervision system , And set up separately AI With the digital medical review department , Speed up AI Medical development ; The British NHS( The NHS ) Plans are under way to expand AI across the health services sector , For daily operation and treatment .

2016 Japan's Ministry of health, labor and welfare began planning AI Medical policies , Including the revision of medical expenses 、 Using artificial intelligence, medical incentives, etc , And is expected to 2020 We will fully implement and promote the artificial intelligence medical system

From the perspective of application effect , The application of artificial intelligence technology in patient-centered medical treatment is still in its infancy , The products are mainly for trial use , There is a high degree of homogenization 、 High concentration 、 The practical effect does not conform to the expectation of doctors and patients . In medicine 、 Health care 、 Hospitals are more enterprise oriented 、 Medical users , The business model is relatively mature , The main test is the technical ability of the supply side .2018 Since then, the development of AI medical application has become more rational , Some companies continue to make bold attempts , Gradually explore different models on the road of commercialization .

Uniform standard 、 Open platform , Promote the deep integration of artificial intelligence and medical care . Microsoft 、 Amazon 、 Google 、 IBM、 Oracle and Salesforce stay 2018 year 8 In the middle of this month, it was jointly announced that the standards would be gradually opened , And remove the technical barriers of medical interoperability through cloud and artificial intelligence technology , Tap the potential of medical data , Better results at lower cost ; Google is in 2018 year 7 month Google Cloud Next It was revealed at the meeting that , AI products AutoML The number of registered users has exceeded 1.8 Thousands of families , With more than 10% Our customers come from the medical and life medicine industries , Effectively promote the user in the medical image auxiliary detection , And timely detection and early warning of stroke 、 asthma 、 Innovations in sudden infant death syndrome . China BAT The three major Internet enterprises make use of their own platform characteristics and advantages , If any AI Medical image analysis and AI Tencent Miying, which has two core capabilities of auxiliary diagnosis and treatment, has been selected as one of the first batch of national AI open innovation platforms by the Ministry of science and technology , 2018 year 6 month AI The interface of auxiliary diagnosis engine is open , Speed up the communication with hospitals HIS System integration .

Focus on partners , To achieve breakthroughs in the application of medical imaging . Medical Science AI The technology R & D company, his company, moved from Beijing to Chengdu , Focus on deep cooperation with a hospital in West China , Joint establishment “ Huaxi - His medical artificial intelligence research and Development Center ”, Establish a mechanism for sharing achievements , Fully mobilize the enthusiasm of doctors , And get stability 、 Safety data . Through to 20 Ten thousand case data learning , The two sides jointly developed the world's first AI Digestive endoscopy prototype , Its effect on polyps 、 Tumor 、 The accuracy rates of initial diagnosis of varicose veins were 92.7%、93.9% and 96.8%, And continuous iterative optimization , Digestive endoscopy AI Technology localization 、 The key step in equipping .

adopt AI Empower , Improve the service level of traditional medical devices . General Electric 、 Siemens 、 Philips and United pictures of China 、 Mindray 、 Yuyue and other medical device manufacturing companies rely on clinical experience and digitalization 、 AI Technology , Continue to launch integrated solutions on the basis of existing medical equipment products , To provide people with better health care and medical care at lower cost . For example, more than half of Philips R & D personnel in the world focus on software development , Most of them are also engaged in artificial intelligence research , In the future, most of Philips products will be based on artificial intelligence technology , We have successively released the overall solutions for cancer diseases 、 Chest pain center / The overall solution of Stroke Center 、 The whole solution of sleep respiratory disease 、 Monitoring system and other solutions .

Interdisciplinary technology requires high , European and American companies lead drug research and development . Drug development is inefficient and time-consuming , A new drug costs more than 1 Billion dollars , The cycle is as long as 8-12 year , It also needs Medicinal Chemistry 、 Computer Chemistry 、 Molecular modeling and molecular graphing, etc , Therefore, it is the most challenging in the application of artificial intelligence medicine . At present, some technology companies use artificial intelligence technology to train a large number of molecular data to predict drug candidates , And analyze the data of healthy people and patients to find new biomarkers and therapeutic targets , Building molecular models , Predict binding affinity and screen drug properties , Effectively reduce the cost of drug development , Shorten time to market and increase the likelihood of new drug success . Such as BergHealth Using artificial intelligence technology, the company successfully found the key molecules of cancer metabolism , Improve the efficiency of new cancer drug research and development , Its main anticancer drugs —BPM31510, At present, it is in the stage of treatment for patients with advanced pancreatic cancer II In the course of phase II clinical trials .

Intelligent regulation , The inevitable choice of national health insurance regulatory agencies . Intelligent regulation combines time and space , From the patient 、 disease 、 Diagnosis and treatment 、 Doctor 、 Hospital and other dimensions to establish a medical relationship network , Using machine learning and other related algorithms , Identify fraud and groups . At present, more than half of the controlled medical organizations in the United States use professional anti fraud information system in the implementation of medical anti fraud actions , To help auditors analyze large amounts of data and conduct prospective fraud investigations , To detect and identify inconsistent data or patterns, etc , With the continuous development of information technology, especially artificial intelligence technology , Medical insurance monitoring is gradually moving towards the era of intelligence .

Our government strongly supports the promotion of intelligent supervision mode of medical insurance , Integrate AI technology with “ Three medical linkage reform ” Combination , In the field of health care regulation , Promote the intelligent supervision mode of medical insurance nationwide , All designated medical institutions with medical insurance will be included in the scope , Realize hospitalization and outpatient medical expenses 100% Intelligent auditing .

(2) Intelligent education accelerates the innovation of education and teaching

Current AI 、 Rapid development of big data and other technologies , Education intelligence becomes the direction of education development . Intelligent education is changing the existing teaching methods , Liberating teacher resources , Profound changes in educational concept and educational ecology . At present, major developed countries in the world accelerate the innovation of education and teaching , Actively explore the new mode of Education , Developing new educational products .

In terms of changing the existing teaching methods , One is to realize the intelligent evaluation of teaching achievements , Improve the quality of teaching . Using artificial intelligence technology to digitize 、 Standardized teachers' teaching behavior and students' learning situation are tested 、 Analysis and evaluation , Help teachers and students quickly and accurately locate teaching problems , Achieve targeted 、 Scientific teaching , Improve teaching effect . The second is to build a personalized learning system , Stimulate students' motivation for autonomous learning . Educational enterprises explore the establishment of knowledge portraits through students' learning characteristics , Push targeted teaching content , Further stimulate students' willingness to learn independently . 2017 year 4 month , Australian independent teaching platform Smart Sparrow get 400 Ten thousand dollar financing , Its educational model has been preliminarily recognized . 2014 year , U.S. adaptive education man-machine war data show that , Independent teaching platform can effectively improve students' learning effect , The average passing rate of students has increased 10%, The average time of acquiring new knowledge is shortened 44%, Domestic ape question bank 、 Crazy teacher 、 Homework box and other Internet education enterprises are gradually introducing similar functions .

In terms of liberating teacher resources , One is to achieve intelligent homework correction , Reduce the teaching burden of teachers . With the continuous innovation of image recognition and semantic analysis technology , The ability of automatic marking of students' homework has been preliminarily realized , 2018 year 4 month , Anhui Provincial Department of Education issued 《 Guidance on the construction of smart campus in primary and secondary schools in Anhui Province 》, clear 2020 A job evaluation system will be built in , Realize the automatic correction of students' homework . According to the data of mobile Internet application service monitoring platform of China Academy of information technology , end 2018 year 4 month , There are many mobile applications that provide automatic homework correction 95 home , Mainly concentrated in the field of primary school speed calculation , Among them, the number of daily active users of love job app exceeds 20 ten thousand , Daily average processing work 50 copies . The second is to expand students' after-school learning ways , Share the pressure of teaching . Education enterprises through the construction of after-school exercise library and combined with image recognition technology , Realize the rapid identification of the students' uploaded topics , Immediate feedback of answers and solutions . London educational institutions Whizz Education, Explore the construction of after class learning system which is highly consistent with the progress of classroom teaching , Through online voice interaction , After class guidance and question answering can be realized .

(3) Intelligent transportation improves urban management level

With the rapid development of global economy , The process of urbanization is speeding up , Increase in vehicle ownership , Increasing road traffic volume , Various traffic problems are highlighted , Developing intelligent transportation can improve government management , Improve user experience , Promote urban development .

Traffic management , One is real-time analysis of urban traffic flow , Reduce vehicle waiting time . The intelligent traffic signal system driven by artificial intelligence uses radar sensors and cameras to monitor traffic conditions , Using artificial intelligence algorithm to determine the light color conversion time , Through the integration of artificial intelligence and traffic control theory , Optimize traffic flow in urban road network . Second, big data analysis of public resource data , Reasonable construction of transportation facilities . The AI algorithm is based on the travel preference of urban people 、 life 、 Consumption and other habits , Analyze the flow of people in the city 、 Traffic movement and urban public resources , Based on the results of big data analysis , Making urban planning for the government , In particular, it provides guidance and reference for the infrastructure construction of public transport facilities . Third, real-time detection of vehicles , Improve the efficiency of law enforcement . By integrating image processing 、 Pattern recognition and other technologies , Realize the motor vehicle lane of the monitoring section 、 Non motorized lanes are monitored in real time all day . The front-end bayonet processing system analyzes the captured image to obtain the number plate number 、 Number plate color 、 The body color 、 Auto Logos 、 Vehicle sub brands and other data , And along with the passing time of the vehicle 、 place 、 The driving direction and other information are transmitted to the database of the control center of the bayonet system through the computer network for data storage 、 Inquire about 、 It's better than peer-to-peer processing , When a hit and run is found 、 Illegal or suspicious vehicles , The system automatically sends alarm signals to the interception system and related personnel .

Car owner experience , One is car assisted driving and driverless driving . The vehicle assisted safe driving system includes on-board sensors 、 On board computer and control execution, etc , The distance between vehicles and surrounding vehicles, road facilities and surrounding environment is measured by on-board sensors , In an emergency , Make all kinds of security measures . Vehicle automatic driving system , Realize automatic guidance during driving 、 Automatic detection and avoidance of obstacles . Second, smart parking . Domestic zebra smart parking and SAIC cooperate to develop China's first Internet car Roewe RX5, Realize intelligent parking 、 Parking status acquisition 、 Safe driving and other functions .

Urban development , First, energy conservation and environmental protection . Intelligent transportation system realizes the effect of energy conservation and emission reduction , By building an intelligent transportation system , Effectively improve the operation efficiency of the existing road traffic network , To ease congestion 、 energy conservation 、 The purpose of reducing pollution , Through intelligent traffic control , Finally, it can reduce the amount of exhaust gas and make a significant contribution to energy conservation and environmental protection . Second, reduce accidents . Take intelligent transportation technology , Improve road management capabilities , Reduce the number of deaths in traffic accidents every year . At present , Developed countries in the world have invested a lot of money and manpower , Large scale intelligent transportation technology research and industrial application , Many developed countries have entered the stage of full deployment .

(4) Artificial intelligence improves the ability of public security

Artificial intelligence has been applied in social security 、 Anti violence and anti-terrorism 、 Disaster early warning 、 Post disaster search and rescue 、 Food safety and other public services , Through artificial intelligence, social security can be accurately perceived and predicted The major situation of the whole operation , Improve the accuracy level of public services , Guarantee the safety of people's life and property . From the depth and breadth of application , Global artificial intelligence is still in the exploratory stage in the field of public services .

In the field of social security , Artificial intelligence has been applied to the police investigation process , Provide important clues for the police to solve the case . Relying on the foundation of the security industry , Criminal investigation has become the first scene of artificial intelligence in the field of public security . Based on computer vision technology in public security control , It can detect the abnormal situation in time , For the police 、 Criminal investigation by procuratorial organs and other judicial organs 、 Public security management and other behaviors provide strong support . Police in many parts of the United States have deployed AI police risk assessment software , Keep crime in the bud . Intelligent software predicts which areas with high crime incidence may have new problems according to the saved crime data .

The application of artificial intelligence in our country can effectively meet the actual combat requirements of public security , Problem oriented problem solving . 2017 During the National Day , The public security department has adopted dynamic portrait control technology in Beijing Tiananmen Square , Total number of alarms 90 many times , Effective interrogation 60 many times , Hit all kinds of objects accurately 50 Many people . And in the BRICs (BRICS) During the ninth summit in Xiamen , The intelligent security system helps the public security department to capture the national fugitives 20 More than . But at present, the world's social security field AI Application development is not balanced . Take Britain for example , Although England AI Technological innovation is relatively active , But the error rate of face recognition is high , The application effect is not satisfactory . according to 《 The independent 》 Published data , The face recognition software used by the Metropolitan Police in the UK produced 104 In the second alarm , There are only two accurate matches .

In the field of counter-terrorism and counter violence , AI is fighting terrorists 、 Areas such as bomb elimination can play an important role . The no fly system established by the United States can predict the possibility of terrorist attacks , Big data systems transmit crime prediction data to the police officers' on duty electronic devices every day , Predictive investigation has been widely carried out . In addition, anti terrorist robots can automatically detect and track suspicious targets , And it has the ability to strike the target in a long-range and accurate way , In the fight against terrorists 、 Assisting the military in areas such as counter-terrorism can play an important role . In our country , An armed strike robot developed by the robot group of Harbin Institute of technology 、 Reconnaissance robots 、 Small explosive disposal robots have been used in anti-terrorism security 、 Target detection 、 Check and strike suspicious objects 、 Roadside bomb removal 、 Hazardous material handling and other fields .

In the field of post disaster relief , Artificial intelligence is effective in dealing with disasters , Play a key role in avoiding casualties . Whether it's search and rescue after natural disasters , Or routine rescue operations , With the integration of artificial intelligence , It can quickly process aerial images of disaster areas , And to provide important assessment and planning guidance to rescue workers in real time , Not only to protect the natural environment 、 The safety of people's lives and property , At the same time, it can minimize the sacrifice of rescue workers .

For example, the Fire Department of Japan's Ministry of general affairs promotes the development of “ Robot fire brigade ”, It's a small drone that takes pictures of the scene from above 、 A reconnaissance robot that collects ground information 、 It is composed of a water gun robot that can walk automatically . NASA NASA To launch the AI System Audrey, Through the sensors that firefighters wear , Get the location of the fire 、 Ambient temperature 、 Signals of hazardous chemicals and gases, regional satellite images, etc , And the prediction based on machine learning can provide more effective information and team suggestions for firefighters , To protect the safety of firefighters to the greatest extent . In our country , Fire fighting 、 Investigation of crimes 、 Smoke exhaust and fire fighting robot technology and products have been relatively mature , And has entered the actual combat , In dealing with disasters efficiently 、 It plays an increasingly important role in avoiding casualties and reducing property losses . In addition, it was developed by the national seismological station “ Earthquake information broadcast robot ”, stay 2017 year 8 month 8 During the earthquake in Jiuzhaigou, Sichuan Province on May , Just use 25 Second wrote the world's first quick report about the earthquake , Push through the official wechat platform of China seismic network , To avoid earthquake disaster 、 Life saving and messaging buy time .

Besides , In food safety 、 Big event management 、 Environmental monitoring and other public security scenarios , Using artificial intelligence technology can reduce labor input and resource consumption , Improve the effectiveness of early warning , Provide strong support for timely and effective disposal .

(5) AI expands the breadth and depth of financial services

Intelligent finance is a comprehensive integration of artificial intelligence and finance . Intelligent finance is based on high technology such as artificial intelligence , Fully empowered financial institutions , Improve the service efficiency of financial institutions , Expand the breadth and depth of financial services , Intelligent financial services 、 Personalization and customization .

The integration of artificial intelligence and traditional financial industry chain is mainly divided into three stages . The first stage is the enabling stage of science and technology , This phase emphasizes application scenarios , Transfer the mature AI technology in other fields to the financial field in parallel , Improve the business efficiency of some links ; The second stage is the stage of technological empowerment , This phase emphasizes model application , Because the direct application of the model will bring compliance risks , Therefore, this stage will produce a lot of third-party professional services , The financial industry is aware of the characteristics and advantages of artificial intelligence , Take the initiative to apply AI in business , It has brought about profound changes in business mode and greatly improved efficiency ; The third stage is the stage of scientific and technological production capacity , It is characterized by value application , Financial core business will be artificial intelligence , Artificial intelligence has become a means of financial core value creation , At the same time, with the efficiency of supervision and the intellectualization of supervision measures .

Artificial intelligence has been widely used in banks 、 investment 、 credit 、 Many financial business scenarios such as insurance and regulation . at present , Traditional financial institutions 、 Large Internet companies and artificial intelligence companies are laying out their financial field one after another , Smart bank 、 Intelligent investment 、 Intelligent investment research 、 Smart credit 、 Intelligent insurance and intelligent supervision are the main applications of artificial intelligence in the financial field , They act on bank operation respectively 、 Investment and Financing 、 credit 、 Business scenarios such as insurance and regulation , But on the whole, the application of artificial intelligence in the financial field is not mature . Artificial intelligence related technologies applied in the financial field mainly include machine learning 、 Biometrics 、 natural language processing 、 Speech recognition and knowledge mapping . The current application scenario is still in its infancy , Most of them are man-machine , The application of artificial intelligence plays an auxiliary role in financial business . But both financial business scenarios and technology application scenarios have strong innovation potential , In the long run , In financial investment advisory 、 Intelligent customer service and other applications may have a disruptive impact on the industry .

Smart investment advisory applications . Intelligent investment advisor mainly refers to the risk preference provided by individual investors 、 Investment income requirements and investment style information , Using intelligent algorithm technology 、 The theoretical model of portfolio optimization , Provide investment decision information reference for users , And with the dynamic changes of financial market, it provides suggestions for the improvement of asset portfolio and allocation . Intelligent investment advisers can not only surpass human beings in investment allocation and transaction execution ability , It can also help investors overcome emotional weaknesses . Industrial and Commercial Bank of China 、 Bank of China and other state-owned banks have also launched intelligent investment advisory services , Citibank expects 2025 The total scale of assets managed by intelligent investment advisers will be high in this year 5 Trillions of dollars . With artificial intelligence, neural networks 、 Continuous iterative innovation and development of decision tree technology , Intelligent investment advisers will be further applied and developed in the financial industry .

Intelligent risk control applications . The application of artificial intelligence technology in intelligent risk control has developed rapidly , With the rapid development of Internet Finance , It's like ants 、 Jingdong finance and many other financial institutions and Internet finance companies vigorously develop intelligent credit services . Intelligent risk control mainly relies on high latitude big data and artificial intelligence technology to timely and effectively identify financial risks 、 Early warning and Prevention . Financial institutions use artificial intelligence and other modern scientific and technological means to analyze the network behavior data of target users 、 Authorization data 、 Behavior modeling and portrait analysis of transaction data , Carry out risk assessment, analysis and tracking , And then speculate the risk point of financing . According to the prior probability of some behavior characteristics that may affect the borrower's ability to repay the loan, the posterior probability is calculated , Financial institutions can monitor the ability of borrowers to repay their loans in real time , Help reduce bad debt losses .

Intelligent financial customer service application . For the financial industry at the high end of the value chain of the service industry , Artificial intelligence technology will be of great significance to the service channels in the financial field 、 service mode 、 Risk management 、 Facility financing 、 Investment decision-making and other aspects bring profound transformational influence , Become a communication customer in the financial industry 、 Identify key determinants of customer needs . at present , Bank of Communications 、 Ping An insurance and other financial institutions have begun to use artificial intelligence technology to carry out natural language processing 、 speech recognition 、 Voiceprint recognition , Serving remote customers 、 Provide effective technical support for business consulting and handling , It's not just an effective response to customer demands , And it greatly reduces the pressure of manual service , Effectively reduce the operating costs of various institutions engaged in financial services .

The impact of AI on financial markets 、 Financial institutions and consumers have a profound impact . For financial markets , AI reduces information asymmetry , Improve market efficiency and stability ; Improve the price discovery mechanism of the whole financial market , Reduce overall transaction costs ; Effectively improve the speed and efficiency of trading , Increase liquidity in financial markets . For financial institutions , AI promotes more financial institutions to use AI to automate their daily business processes , Effectively identify customer needs and provide customized products , Significantly improve performance ; Encourage financial institutions to detect fraud in advance 、 Suspicious transactions 、 Default and cyber attacks , Improve the level of risk management . For consumers and investors , AI reduces the cost of financial services for consumers and investors , Promote their access to a wider range of financial services ; Grasp each consumer or investor's consumption preference through intelligent data analysis , To provide more customized and personalized financial services .

(6) Smart home helps build smart home

The application scenarios of artificial intelligence in home field mainly include smart home appliances 、 Home security monitoring 、 Smart home control center, etc , By identifying biometrics 、 Automatic speech recognition 、 Image recognition and other artificial intelligence technologies are applied to traditional household products , Realize the intelligent upgrade of household products , Building a smart family in an all-round way . Smart home products are relatively mature , The future market development space is huge .

One is to build smart home appliance terminal products . Through image recognition 、 Automatic speech recognition and other artificial intelligence technology 、 Air conditioner 、 Intelligent upgrade of TV and other household appliances , Promote intelligent control of household appliances 、 Functional diversity , Enhance the experience of using household appliances . For example, Aucma and Jingdong jointly developed and launched a smart large screen Internet refrigerator , Built in camera can automatically capture images , Automatic recognition based on image recognition technology 120 A variety of ingredients , Set up food material library for users , Automatic food monitoring , And can track and learn user habits , Recommend recipes intelligently for users . Changhong launched Alpha AI voice air conditioning , Equipped with intelligent voice control module , Through automatic speech recognition technology , Realization 6 Mina voice interaction 、 Full semantic recognition manipulation , Efficient recognition and speech control, accuracy of 95% above .

Second, the realization of family security monitoring . Image recognition based 、 biometrics 、 Artificial intelligence sensors and other technologies to achieve home external environment monitoring ( Such as buildings )、 Home lock control ( Such as intelligent door lock 、 Cat's Eye )、 Home environment detection ( Such as air quality 、 Smoke detection 、 Personnel activities, etc ) And so on . Such as LifeSmart The face recognition visual door lock created by yunqi and Intel , Capture images or video streams containing human faces through cameras , Automatic face detection and tracking in images , Identity recognition based on human face feature information , Face recognition is realized 、 Remote visualization 、 Linkage defense of intelligent door lock . The comprehensive accuracy of face recognition of spotted cat's intelligent cat's eye product can reach 99.6%, After collecting family information , Smart cat's eye will quickly recognize the family , And broadcast information about family members coming home , Build a warm smart home life scene ; And if strangers visit , Smart cat's eye will alert strangers , And can recognize a variety of face attributes , Will age 、 Gender and other information is sent to the user's mobile phone , Let users respond in time , Build a safe home environment .

The third is to build a smart home control center . Based on automatic speech recognition 、 Semantic recognition 、 Question answering system 、 Intelligent sensors and other artificial intelligence technologies , Develop smart home control system ( Overall solution ), Home appliances 、 The curtain 、 Lighting and other different types of equipment interconnection , From simple device on and off , Step by step towards intelligence 、 Convenience 、 Personalization . At present, smart home control center has APP control 、 Intelligent device control ( Like a smart speaker ) And intelligent robot control . Google Assistant、 samsung Smart Things Smart home control center adopts APP Control mode . Through Google Pixel Installed in the mobile terminal Google Assistant Software , And in Google Assistant Add a new feature based on automatic speech recognition technology to “Home Control”, Users can send Pixel Give voice commands , Finish adjusting the temperature inside the house 、 Control the lighting 、 Switch TV channels 、 Play music and other operations . Amazon echo、 Google Home Adopt intelligent device control mode . haier Ubot Adopt intelligent robot control mode .

Chile stuff Think , at present , The relatively mature products of artificial intelligence are mainly concentrated in the security monitoring equipment and other local subdivisions , Intelligent sweeping robot 、 Intelligent speakers 、 Machine translation machines and other products are common, with small coverage 、 The number of users is small 、 The level of intelligence is low , Besides , There are more product gaps . Although in the foreseeable future , The kind of autonomous artificial intelligence in movies and TV plays will not appear , But simplifying software complexity through machine learning algorithms 、 Enhance the performance of the machine “ intelligence ” There is still a vast space for development . for example , Assistant driving system will be a must for cars , Although completely driverless may not appear for a long time . Home appliances will be more intelligent , At the same time, there will be new home appliances such as home service robots .

本文为[Old friends in the old lane]所创,转载请带上原文链接,感谢