[Ai RE ZERO]How to do it?Ai Yunhua, Cloud AII

  Editor’s Press: This article is a columnist author Luo Chao Channel authorized entrepreneurial statement,Copyright belongs to the original author.

  Ai is doing no hole,Even the relatively conservative power industry,It also became a popular scene for AI applications.

  As a state-owned enterprise directly under the Guangdong Power Grid Corporation of China Southern Power Grid Corporation,The Foshan Power Supply Bureau is responsible for the safety power supply, power grid construction and supply of electricity service in Foshan City.It has about 4,500 kilometers of transmission lines within its jurisdiction.More than 16,000 hoven tower units,There are more than 300 construction points that are prone to external forces near the line.Large equipment such as tower cranes or presses the electric wire will affect transmission safety.

  Foshan Power Supply Bureau outside the artificial inspectionInstall more than 100 sets of online monitoring equipment near the line,Concrete the photo and upload the system database.The initial photo will be monitored and monitored.As photos reach millions,The workload is getting bigger and bigger,There are more and more missions.

  Technicians think of AI: Based on machine visual technology intelligence, real-time identification of external security hazards such as large locomotives.Ai development threshold,It is impossible to develop from 0 to 1.Open source framework is preferred,But the difficulty of studying the open source framework in the market is still very high.Later they found a solution: Based on Baidu Intelligent Cloud,With Baidu Easydl customized image recognition for 3 month training,Form a model model of “external hidden dangers of transmission lines”,It can intelligently identify external hidden dangers such as cranes, excavators, etc. in the transmission line.Identification accuracy reaches 80%,beyond expectation.

  Foshan Power Supply Bureau technical personnel doing very simple: synchronize the picture collected by the field equipment to the cloud,Then call the Baidu EASYDL model interface to use the “power transmission line external hidden danger” model to return parameters in the cloud recognition,The analysis results are passed to the regulatory system.Safety hidden dangers will be investigated by real-time notification of the project team.Auxiliary artificial site inspection,Make sure the power transmission line is safe.


  Through the construction point near the AI monitoring power line,Large improvement of audit timeliness, reduce the audit work and human cost of hundreds of people.Now “External Hidden Digital Identification” model of power transmission line continues to be updated in a frequency of 2-3 weeks.Accuracy is still continuously improved.

  Apply AI to the construction point monitoring of the power transmission line.Just a small tapping knife.The application space in the power industry AI is everywhere.

  The State Grid released the top ten key construction tasks of “Digital New Infrastructure” a year ago.Sign a strategic cooperation agreement with 41 partners such as Baidu,One of the key tasks of “Digital New Infrastructure”,Is “building a power artificial intelligence open platform,An artificial intelligent sample library, model library and training platform is built within the year.Explore 13 typical applications.”

  Today, AI is in the application of national grid, loses, distribution, use, etc.Such as distributed energy prediction, equipment inspection, on-site operation safety control, intra-enterprise search, grid scheduling, intelligent response assistant, charged operation robot and other scenes.

  For example, in the transmission inspection scene,Key technologies such as the joint research institute, China Electric Academy and other units, etc.Intelligent analysis model for forming 8 categories 29 small defects,The artificial intelligence image identification cloud service of the company’s transmission line is achieved through the national network cloud platform.It has piloted 27 provinces, about 300 levels of units in the national network company.Get extensive recognition and praise.


  Another example is a smart customer service scene,National Net Customer Service Center knowledge maps in the field of electricity customer serviceUtilize intent identification, natural language understanding, image identification, etc.Building a multi-channel intelligent customer service system,Provide higher quality, high efficiency, convenient intelligent services for our power users.

  National Grid is applying machine learning, natural language processing, computer vision, power intelligence robots, etc. Ai technology to various scenes.National power grid who tasted sweets put forward the power artificial intelligent development road map,The first stage objective,It is a digital transformation and intelligent leap in the grid before 2035; the long-term goal is to achieve full intelligence of the overall power grid operation.Use artificial intelligence as a tool,Fully integrated into the operating process of the grid.


  Not just the power industry,Today, education training, cultural and entertainment, traffic travel, e-commerce, financial retail, etc. have a number of AI application practices.

  Beijing Xinqiao uses AI technology to improve the quality and efficiency of bridge inspection;

  China Air China APP is recognized by accessing Baidu brain OCR ID card.Implement the function of passenger online self-help correctment of the wrong ticket information;

  People’s Daily is based on Baidu Ai Technology to achieve “creation brain”,Use AI to help press policy, mining, editing, review, hair, etc.Improve the quality of news and accuracy,Reduce news production costs;

  Jinan subway brush face payment gate,Identify only 1.8 seconds,Extra double passing the passenger passing speed;


  Different companies and organization applications AI have different scenes.The result is much smaller: the result is increased.Educate this,Is AI replacement, assistant,Reduce cost; efficiency,It is to improve efficiency, improve efficiency, improve quality, promote innovation, and improve value.

  ”Artificial intelligence is the fourth industrial revolution after the steam revolution, the power revolution, the Internet revolution,Artificial intelligence technology is considered a new era of electricity.”A few years maybe it is difficult to understand this macro description.I feel deep now.Because AI is nowhere, no hole is incompetent,Like IT, electricity and other technologies, life is the daily life.Become a social infrastructure.

  Throughout the benchmark enterprises applying the AI applications of each line,There is a common feature when AI development: Online AI applications in a short time,There is no hiring a large number of top AI scientists or senior AI engineers.Professional AI technical talents are short and cost,Recruitment is difficult to solve the urgent experience in a short time.They do not have to choose third-party Ai technology services.It is not a tender to engage in a project, please ask the AI technology outsourcing company to develop,But use cloud services to build a special AI development platform.The general step is very similar:

  Step 1: Combine business needs in the cloud custom AI model,The AI model reaches a certain intelligence level after the cloud training;

  Step 2: Let the AI model run in the cloud,The data required by the business is passed to the cloud call AI-related interface to obtain the analysis results;

  Step 3: Return the analysis results to its own business system,Combined with business logic to form specific decisions.

  This process is very similar to “Last Cloud”.

  Cloud computing is easy to understand,Previously, the company’s website, APP or system should run,To purchase a bunch of servers yourself,Put your computer room or rent-free machine room,Removal of professional IT staff management,Cost is high,It is also necessary to handle power outages, hacker attacks, server damage, etc.With cloud computing,Enterprises can run the system on the cloud,How much resources have been used,The cost is low,Calculation has also become simple.

  Develop AI with deep learning open source frameworks,Developers want to spend a lot of time learning algorithms and training models,To purchase private servers or cloud resources to solve the needs of strength,To learn and use a variety of AI model tools,Difficult to ignore – let alone, many companies do not have professional AI technicians.

  Use cloud AI capabilities,Enterprises no longer need to hire professional AI technical talents,Ai uses very simple,The threshold and the cost are reduced,AI development is as easy as traditional IT development.”Ai on the cloud”,Become an inevitable trend of AI technology applications.


  Even if I don’t apply Ai technology,Under the horn of the digital transformation of the whole society,”Cloud” has become a must-have for business.”Cloud” is not just to reduce the calculation cost,More importantly, you can bring aggressive data, open business processes,It is the first step in the digital transformation of enterprises.Enterprise “Cloud” brings together data to the cloud,Preparing the raw materials to the AI app,It has the basis for intelligent transformation.

  Digital transformation and intelligent transformation are not a choice issue for companies.Instead, “both want”,More and more companies are walking “Digital” upgrades.Under “Enterprise Cloud” and “AI Yunhua” double trend,”Yun Zhi Intega” is a matter of success.It is because of this.Baidu smart cloud, the main “cloud-shaped” mode,Become the preferred platform for enterprise AI development.


  Baidu Intelligent Cloud is the first name “smart” cloud service provider,Later, other manufacturers have renamed or emphasize intelligent properties.It reflects that AI is increasingly important to cloud.Baidu Ai technology has developed more than ten years,The essence of the cloud is the ability to overflow,Baidu Smart Cloud also became the best form of Baidu open a variety of AI capabilities.

  As early as Baidu smart cloud in 2016, I proposed the “ABC Trinity” strategy.Fusion AI, cloud computing and large data.During 2016 to 2019,Baidu has been pushing clouds and Ai fuse.2020,Based on the 10-year AI technology and ecology thickness,Baidu Intelligent Clouds “is based on cloud computing,Take Ai as a starting point,The development route of important tracks,Further establish a development strategy of “Yunzhi Integration”.

  Baidu Intelligent Cloud’s “cloud-shaped” is not superimposed on the cloud computing,It is not to serve the AI technology to cloudize.But the cloud and AI depth,Carry AI with clouds,Use AI to define clouds.


  Let me say “cloud” first.AI calculates “big, more, miscellaneous, scattered”,Clouds that are most suitable for AI have new architectures.Baidu Intelligent Cloud integrates Ai elements in “AI development infrastructure (equivalent to traditional iaaS),Ai development infrastructure after special tuning,Can play a higher performance, high utilization,Become the most suitable clouds for running AI.

  Again, “wisdom”.Although there is a variety of AI technical service providers and corresponding cloud services,But they often do not have native cloud capabilities.Baidu Intelligent Cloud’s “AI Development Platform” integrates into the Ai native cloud computing architecture,Optimize from the chip, OS, server, etc.Therefore, the force and speed are more accurate and faster.

  Ai defined cloud, cloudy AI,Let Baidu Intelligent Cloud becomes the only cloud computing platform and AI development platform of “Yunzhi Integration”.Form all stack AI development capabilities,Better meet the needs of enterprise AI development.”Yun Zhi Intega” also became the core advantage of Baidu Intelligent Cloud.IDC “China Artificial Intelligent Cloud Service Market Research Report (2020H1)” report shows thatBaidu Intelligent Cloud has ranked first in the Al Cloud market three consecutive times.Ranked first in multiple segmentation fields such as image video, NLP natural language processing, human identification.


  More and more companies realize thatTo develop AI applications using the cloud AI platform,To use a cloud with AI capabilities.”Yun Zhi Intega” became the final form of the development of AI and cloud computing.

  Industrial intelligence is getting up,AI and industries are rapid, extensive and in-depth.How do companies build a future AI development platform?What should I pay attention to when AI is developing?Baidu Intelligent Cloud Released “” Yun Zhi Intega “Technology and Application Analytical Series White Paper – Enterprise Development” (hereinafter referred to as “” Yun Zhi Intega “White Paper”),It recommends that when the company is selected and building a AI development platform that adapts to the future development trend,Pay attention to the “Scene first, independent innovation, efficiency”, “Conduct unified planning and construction.


  Specifically,”Full Full AI Development Platform”,It is the optimal solution of enterprise AI development.

  first of all,Enterprise AI development presents “customization” trend.

  The business scene of different enterprises is extremely complex. Just like the example,National Grids should apply AI to power generation, transmission, power generation, power grid safety and control, business management, government social services, etc.Recommended different Ai technology such as natural language processing, deep learning, machine vision, intelligent voice, knowledge maps,The general model cannot meet its AI development needs.

  It is not just a national grid in customization AI demand.According to the joint research of Baidu and Boston Consulting, 2018,86% of the market in the market is customizes AI demand,The AI model in 2017 and 2020 surged 6 times more.


  To meet the customization of enterprises,The AI development platform must have a versatility of fundamental ability.And can be easily decoupled, embedded, transplanted and iterative.According to “Yun Zhi,” White Paper “,Baidu Intelligent Cloud AI infrastructure is divided into underlying (AI infrastructure), medium (AI development double platform: zero threshold AI development platform Easydl and full-function AI development platform BML) and upper layer (facing industry-oriented solution),Form a full stack AI product architecture,Can meet the AI custom needs of any enterprises.


  ”Full Stack” means “big and complete”,It means that the development of enterprises, means complexity,Therefore, Baidu Intelligent Cloud integrates AI capabilities to “AI”.Through the AI capabilities engine and the AI development platform,Provide enterprises to provide data management, service management, full-line resource operation and maintenance management system,Implement intelligent upgrade of intelligent intelligence management of intensive AI capabilities.


  Simply said: Baidu Smart Cloud has full stack AI capabilities,Enterprises can use and assemble their own AI development platform on demand.Through “AI”,Enterprise AI model customization, management and operational complexity have been greatly reduced.

  Different subsidiaries of the State Grid and the business unit must customize various AI applications.It is based on Baidu AI TED to customize the national network artificial intelligence middle platform.Integrated basic AI capabilities, automation, and visualization tools,Painted some common model services,Supports upper complex AI applications.With artificial intelligence middle platform,National network model developers, application developers, business applicants can easily use basic Ai capabilities, development tools, and various interface services development AI applications.


  Based on the AI middle station, the development model of full stack AI capabilities,It can meet the customization AI development needs of all companies.

  Secondly,Enterprise AI development is becoming more sensitive to efficacy.

  With the AI in more scene depth applications,More and more data to be processed, the model is getting bigger and bigger, and the customization is getting deeper.Therefore, the demand for power is getting higher and higher: the amount of calculation is getting larger, the accuracy requires increasingly high, and the processing aging requires increasingly high.The result is that the cost is getting higher and higher.

  at the same time,AI development involves data collection, data pretreatment, data labeling, model training, model assessment, and model deployment, etc.Long process, high complexity, more process,Every step is possible to become a function bottleneck,It may increase the cost of AI development.


  Maximize performance under the premise of cost control.It is a problem that AI development is important.The cost includes the cost and development cost,Validation includes data quality, model precision, processing effect, result accuracy, etc.therefore,”” Yun Zhi integrated “white paper” believes thatAdapting to the future AI development platform must have the following three capabilities:

  In terms of power resources,To dock and manage the performance, high utilization, high cost-effective AI power resources;

  In terms of research and development efficiency,To provide the optimization and development efficiency of the whole process;

  Aiming at the actual business scene and customization needs of the company,Provide overall solutions.


  Baidu Intelligent Cloud “Yunzhi Intega Ai Development Full Function”,It is just that you can meet the above needs.Baidu has both AI technology advantages,There are also cloudy events.Every link from the basic hardware, storage, container, development platform to the upper AI model, and the application of the “Yun Zhi Intega” concept,It has a low cost high efficiency.

  For example, “Baidu AI Development Infrastructure” is the cloud defined by AI,Baidu Intelligent Cloud In the industry, the AI native cloud computing architecture is proposed in the industry.Construct an infrastructure such as AI computing cluster, AI chip, etc.Special tuning for the server, AI acceleration, AI optimization, AI storage, AI container, etc.Has a higher performance (training speed, accuracy, model effect), higher utilization,Lower cost.


  For example, Baidu “AI development double platform”,PaddlePaddle, based on the autonomous development of paddle core open source framework (PaddlePaddle),It has powerful customization, portable and compatibility.Support by all of the tools,One-stop AI development and deployment requirement for developers full processes,It is easy to use, efficient and smooth.

  At last,Enterprise AI development is compatible with future ability.

  The AI industry ecology is still in high-speed growth,The techniques of different dimensions such as algorithms, power, cloud services, and applications are constantly evolved.”” Yun Zhi integrated “white paper” believes thatAI development is a dynamically evolving process,It is in the process of continuing to interact with the application scene,Get new data by continuously gaining feedback,The process of continuous iteration.Simultaneously,Artificial intelligent application scene dispersion, Need closely with business logic.therefore,For companies and developers,It is necessary to have enough flexible, easy to use platforms to make independent innovation.

  For example, the AI development platform is compatible with different computing platforms.Baidu Avantou Enterprise Edition is the most widely used AI development platform in the industry chip.Adapted to Feitan, Huawei, Baidu Kunlun and other domestic chips and servers.For example, the AI development platform should support business first race zero code application AI,Baidu Avantou Enterprise Edition adopts a “double platform” mode,Different users,There are both full-featured development models to meet the development needs of senior R & D staff.It also provides zero-threshold mode suitable for the first-line business personnel.Reduce the threshold of the AI application.


  Under the trend of “Yunzhi Integration”,More and more cloud computing manufacturers on the market emphasize the word “intelligence”,And provide different AI components on the cloud platform.But their clouds and AI are cutting,The cloud is not for Ai,Ai has not integrated with the cloud to achieve “integration”.Customization capabilities, AI calculations and development efficiency (efficiency + cost) and flexibility is not as good as Baidu smart clouds.They do not have full stack AI development capabilities,Enterprises must face different interfaces,The complexity is greatly improved, and the development efficiency is significantly reduced.

  Baidu financial report display,Baidu intelligent cloud camp in the first quarter reached 2.8 billion yuan,A large increase in 55% year-on-year,The growth rate in a cloud computing platform is a leading level.Baidu smart cloud rapid growth,A large part of the reason is its unique mode through “Yunzhi Integration”.I have eaten a smart dividend.

  As the industrial intelligence further accelerates,Each company is accelerating “智 智” transformation,”Yun Zhi Intega” value will be found by more companies,Baidu Intelligent Cloud’s advantages will be more exciting.Baidu Smart Cloud’s new round of rapid growth,Pull the dimension.

  Cloud computing market,It is going to change.

popular articles