Recently in artificial intelligence （AI）、 Great progress has been made in edge computing and the adoption of Internet of things devices , These all come together as edges AI Bring opportunities .

This is the previously unimaginable edge AI Opened up new opportunities —— From helping radiologists identify diseases , To drive a car on the highway , And then to help us pollinate plants .

Edge computing, which is talked about and implemented by countless analysts and enterprises , Its origin can be traced back to 20 century 90 years , At that time, the content delivery network was created , From the edge server deployed near the user Web And video content .

today , Almost every enterprise has the edge of adoption AI Benefit from the work function . in fact , Edge applications are driving the next wave of artificial intelligence , To improve our family 、 Work 、 Life in school and transportation .

This article will take you through the edge AI What is it? , Its advantages and working principle , edge AI Use cases and the relationship between edge computing and Cloud Computing .

What is the edge AI？

edge AI Is deployed in devices throughout the physical world AI Applications . It's called " edge AI", Because AI Computing is done near users at the edge of the network , Close to the location of the data , Instead of concentrating on cloud computing facilities or private data centers .

Because the Internet has global influence , So the edge of the network can be almost anywhere . It can be a retail store 、 factory 、 Equipment in hospitals or around us , Such as traffic lights 、 Telephone 、 And automated machines .
# edge AI： Why now ？

Enterprises from all walks of life are seeking to improve the degree of Automation , To improve the process 、 Efficiency and security .

To help them , Computer programs need to recognize patterns and perform tasks safely and repeatedly . But the world is unstructured , The scope of human tasks covers an infinite environment , These environments cannot be completely described in procedures and rules .

edge AI The progress of has opened up opportunities for machines and equipment , No matter where they are , Can use human cognition " intelligence " To operate . Support AI Smart applications can learn to perform similar tasks in different situations , Just like real life .

Deploy on the edge AI The rise of models stems from the recent 3 Progress in four areas ：

• The maturity of Neural Networks ： Neural networks and related AI The infrastructure has finally developed to allow generalized machine learning . Enterprises are learning how to train successfully AI
  Model , And deploy it to the edge production environment .

  • Advances in computing infrastructure ： Running on the edge AI Strong distributed computing capabilities are needed . Highly parallel GPU The latest progress of has been applied to executive Neural Networks .

  • Adoption of IOT devices ： The widespread adoption of the Internet of things has driven the explosive growth of big data . With the sudden ability to collect data in all aspects of the enterprise – From industrial sensors 、 Smart cameras 、 Robots, etc , We now have edge deployment
    
    AI Data and equipment required for the model . Besides ,5G Provide faster for the Internet of things 、 More stable and secure connections .

Why deploy on the edge AI？

because AI Algorithms can understand languages 、 Vision 、 voice 、 Smell 、 temperature 、 Unstructured information in facial and other simulated forms , So they are particularly useful where end users with real problems occupy . Due to and delay 、 Bandwidth and privacy related issues , Will these AI It is impractical to deploy all applications in a centralized cloud or enterprise data center , It's not even possible .

edge AI The benefits include ：

• Intelligent ：AI Applications are more powerful and flexible than traditional applications , Because traditional applications can only respond to the expected input of programmers . by comparison ,AI Neural networks are not trained to answer a particular question , But how to answer a specific type of question , Even if the problem itself is new . without AI, Applications cannot handle an infinite variety of inputs , Text 、 Speaking or video .

  • Real time insight ： Because edge technology analyzes data locally , Instead of analyzing data in remote clouds with telecommunication delays , Therefore, it can respond to users' needs in real time .

  • cost reduction ： By bringing processing power closer to the edge , Applications require less Internet bandwidth , Thus, the network cost is greatly reduced .

  • Increase privacy ：AI Can analyze the information of the real world , Without exposing it to humans , This greatly increases any need to analyze the appearance 、 voice 、 The privacy of people with medical images or any other personal information . edge AI By including this data locally , Only upload analysis and insights to the cloud to further enhance privacy . Even if some data is uploaded for training purposes , It can also be anonymized to protect user identity . By protecting privacy , edge AI Simplifies challenges related to data compliance .

  • High availability ： Decentralization and offline functions make the edge AI More powerful , Because processing data does not require internet access . This is a critical task 、 Production grade AI Application brings higher availability and reliability .

  • Continuous improvement ： With the training of more data , AI models are becoming more and more accurate . When the edge AI When an application encounters data that cannot be processed accurately or confidently , It usually uploads data , In order to AI You can retrain and learn from it . therefore , The longer the model is produced on the edge , The more accurate the model is .

edge AI How technology works ？

In order for the machine to see 、 Perform object detection 、 Driving a car 、 Understanding pronunciation 、 speak 、 Walk or otherwise imitate human skills , They need to replicate human intelligence in function .

AI A data structure called deep neural network is used to replicate human cognition . these DNN After training , You can answer a specific type of question by showing many examples of this type of question and the correct answer .

This is called " Deep learning " The training process usually runs in the data center or cloud , Because training an accurate model requires a lot of data , And it requires data scientists to cooperate to configure the model . After training , Models can be used to answer real-world problems " Inference engine ".

On the edge AI Deploying , The inference engine can be in the factory 、 The hospital 、 automobile 、 Some kind of computer or equipment in remote areas such as satellites and homes . When AI When there is a problem , Troublesome data is usually uploaded to the cloud , To the original AI Model for further training , At some point replace the edge of the reasoning engine . This feedback loop plays an important role in improving the performance of the model ; Once the edge is deployed AI Model , They will only become more and more intelligent .

edge AI What application examples are there ？

AI It is the most powerful technological force of our time . We are now in an era when artificial intelligence is completely changing the largest industry in the world .

In manufacturing 、 Health care 、 financial service 、 The transportation 、 Energy and other fields , edge AI It is promoting new business achievements in various fields , for example ：

• Intelligent prediction of energy ： For key industries such as energy , Discontinuous supply will threaten the health and welfare of ordinary people , Intelligent prediction is the key . edge AI The model helps to combine historical data 、 Weather patterns 、 Grid operation and other information , To create complex simulations , So as to provide customers with more efficient energy production 、 Assign and manage information .

• Predictive maintenance in manufacturing ： Sensor data can be used to detect abnormal conditions early and predict when the machine will fail . If the machine needs maintenance , Sensors on the device will scan defects and manage alarms , In order to solve the problem as soon as possible , Avoid costly downtime .

• In health care AI instrument ： Modern medical instruments at the edge are being enabled by devices that use ultra-low latency surgical video streams AI, To achieve minimally invasive surgery and on-demand insight .

• Intelligent virtual assistant in retail ： Retailers hope to improve the digital customer experience by introducing voice ordering to replace text-based search with voice commands . Order by voice , Shoppers can easily search for products using smart speakers or other smart mobile devices 、 Ask for product information and order online .

What role does cloud computing play in edge computing ？

AI Applications can run in data centers in the public cloud , It can also run in the field of the network edge near the user . The advantages provided by cloud computing and edge computing can be deployed at the edge AI When combined .

The cloud provides with infrastructure costs 、 Extensibility 、 High utilization 、 Server failover capabilities and collaboration related benefits . Edge computing provides faster response time 、 Lower bandwidth cost and network failure resilience .

Cloud computing can support edges in many ways AI Deploy ：

 -  The cloud can run models during training .

 -  Cloud continuous operation model , Because it uses data from the edge for retraining .

 -  The cloud can run AI Inference engine , When high computing power is more important than response time , These engines can complement the model on site . for example , The voice assistant may respond to its name , But complex requests will be sent back to the cloud for parsing .

 -  The cloud provides the latest version  AI  Models and Applications .

 -  The same edge AI usually runs in the field of device clusters , The software is installed in the cloud .

edge AI The future of

Due to the commercial maturity of Neural Networks 、 The popularity of Internet of things devices 、 Parallel computing and 5G Progress , There is now a strong infrastructure for general machine learning . This enables enterprises to take advantage of this huge opportunity , take AI Introduce its application fields , And act on real-time insights , At the same time, reduce costs and increase privacy . We are just on the edge AI Early stage of , Possible applications still seem endless