Contact
Edge AI is a technology to process AI in a location close to the terminal and the field rather than the Cloud.
For example:
- スマホ
- Automotive
- Security Camera
- Factory Robot
- IoT Sensor
などで使われます。
If Cloud AI is a “mechanism for AI processing in data centers”, Edge AI is a “mechanism for AI processing near users and machines”.
In this article,
- Definition of Edge AI
- Cloud AI
- Merit/Demerit
- Important Reasons as an Investment Theme
to beginners.
What is Edge AI?
Edge AI
AI processing data close to the terminal or where data is generated
を指します。
"Edge" means the edge of the user and the site, not the cloud in the center of the network.
具体的には、
- Smartphone
- Camera
- Car
- Sensor
- Factory
- Store terminal
This is the edge.
Edge AI
“Technologies that can not only place AI on the cloud, but also on the field”
Home
Cloud AI
What is the difference between Edge AI and Cloud AI?
| 項目 | Edge AI | Cloud AI |
|---|---|---|
| Processing place | Terminal side and field side | DataCenter |
| Contact Us | Less | More |
| Speed | Low latency | Communication delay |
| Power | Power saving is important | High performance calculation |
| Application | Im ate judgment, field control | Large-scale learning and large-scale analysis |
| weakness | Constraints on terminal performance | Depends on the communication environment |
Cloud AI is suitable for learning and large-scale reasoning using large amounts of data.
On the other hand, Edge AI is suitable for real-time environments.
Why important?
Edge AI
Since there are more scenes that require real-time processing
Home
For example, if you send brake decisions to the cloud every time, it will be too late.
Even with a security camera, if you continue to send all the videos to the cloud, the amount of communication becomes larger.
Even in factory equipment, if abnormal detection is delayed, it leads to line stoppage and defective products.
Therefore, edge AI that judges immediately on the field is important.
Edge AI
Smartphone
On your smartphone,
- Voice recognition
- Camera correction
- English
- 文字起こし
- Face recognition
There are more scenes to process such as on-board AI.
By processing everything to the cloud, there are advantages of response speed and privacy.
Automotive
By Car
- Detection of surrounding objects
- Lane recognition
- Pedestrian Detection
- Driving Support
Edge AI is used.
It is not always possible to communicate.
Therefore, AI processing that can be judged in the car is important.
Security Camera
Security Camera
- Character Detection
- Severe behavior detection
- Congestion detection
- In sion detection
can be processed on-site.
It reduces the amount of communication by sending only necessary information, rather than sending all video to the cloud.
Factory & IoT
Factory
- Detection of equipment failure
- Image inspection
- Robot Control
- Temperature and vibration monitoring
is used.
The purpose is to quickly find the prosecution of failure and reduce the stop time.
メリット
High speed
It is easy to get low latency due to low communication waiting.
We have strengths in applications that require instant judgment, such as autonomous driving, robot and factory equipment.
Cost reduction
No need to send all data to the cloud.
Especially because of the large amount of video, audio and sensor data, it is a great advantage that it can be pre-processed on the terminal side.
Privacy Policy
You may be able to process your personal information or sensitive data on the device without sending it to the outside.
Important points for medical, financial, surveillance cameras and factory data
Easy to move even if communication is unstable
Even in an environment that can not be connected to the cloud at all times, if you have AI on the terminal side, you can perform certain processing.
It is a strength in Yamama, factory, car, store, disaster site etc.
デ
High performance chip required
In order to process AI on the device side, you may need not only CPU but also GPU, NPU, and AIクセelerator.
In other words, edge AI is not only software but also semiconductor performance.
Power
It is important to have a battery in small devices such as smartphones, cameras, andセンサー.
Even with high performance, too much power will reduce the utility.
For this reason, power saving performance is very important in Edge AI.
It is difficult to update model
With the cloud, you can update the AI model on the server to reflect many users.
Edge AI requires updates, compatibility, and security management.
Operation after introduction is not easy.
Reasons to Focus on Investment
When Edge AI spreads, the relevant market will spread.
especially important
- Semiconductor
- Sensor
- Camera Module
- Contact Us
- IoT Devices
- Embedded software
- Industrial Equipment
Home
It is a semiconductor dedicated to AI.
In Edge AI, unlike the huge GPU for the cloud,
- Small
- Low power consumption
- Low latency
- Low cost
- High efficiency
AI chip is required.
This is why not only GPUs, but also NPUs, AIクセelerators, and sensor processing chips are important.
How beginners can easily misunderstand
Not only AI=Cloud
Because of the strong impression of the generated AI, AI tends to move in a huge data center.
In the future,
- Cloud AI
- Edge AI
- Terminal AI
- Hy d AI
is combined.
Not all clouds.
Edge AI is not an alternative to cloud AI
Edge AI does not replace Cloud AI completely.
We are good at cloud for large learning and heavy analysis.
On the other hand, we are good at edge AI.
What is important
Which is reasonable to process instead of “cloud or edge”?
Home
Edge AI is an infrastructure investment
Edge AI is not just an app.
Back
- Semiconductor
- Contact
- Sensor
- Terminal design Power control
- Security
Yes.
If you look as an investment theme, you need to focus on not only AI software, but also hardware and parts manufacturers.
- Edge AI is a technology to process AI on the terminal and on-site
- The difference between cloud AI is the processing location and volume
- High speed, low latency, reduced traffic volume, and privacy
- Issues for terminal performance, power consumption and update management
- Semiconductors, Sensors, IoT, and Communications
In the AI field
Where to Process
will be an important theme for the future.
AI processing in the cloud and edge AI that judges immediately on site.
By separating these two, AI-related symbols and semiconductorテーマ are easy to understand.
出典
- IBM「What Is Edge AI?」
- IBM「Edge AI vs. Cloud AI」
- NVIDIA Blog「What Is Edge AI and How Does It Work?」
- Intel「Edge AI」
- Red Hat「What is edge AI?」