Edge means the local or the near local processing. For example a local device like a computer, a smart tv, or a smart refrigerator, or even a local server that is located close to the area of working.
While for Cloud, the applications first gets the data locally then sends it to the cloud for processing and finally sending it back with a response.
An Edge can be used whenever low latency is required or when the network may not be available all the time. The local processing will mean no need for a cloud processing and sending back and forth of the information or data. When using Edge, one has the desire to obtain desired results in real time for the application.
Majority of the applications using cloud get ehri data sent to the cloud, the processed, before being sent back to the working area. That means, using Edge, one does not need to send any of the data to the cloud. This means more security of the data and less impact on a network.
Therefore, Edge AI algorithms can easily be trained on the cloud but then easily run locally on Edge. The table below shows some of the AI applications that use either Edge or cloud:
| AI Application | CLOUD or EDGE |
| Voice Assistant | Cloud |
| Self-driving cars | Edge |
| Remote Nature Camera | Edge |
| Sales Transactions Insights | Cloud |
| Fitbit devices/wearable devices | Edge |
| Robots for surgery | Edge |
| Tracking devices | Edge |
Voice assistants send query to the cloud for processing. While the self-driving cars do need to be able to perform computation at the edge. The remote cameras may not be able to send the data over a connection. Whereas the sales transactions insights will require gathering of data using a high computing in the cloud.
However, one needs to know that the edge does not mean no usage of cloud at all. The AI models can still be trained in the cloud and then get deployed at the Edge. For instance, the GPS map on your mobile phone can be a great example of an Edge application that uses an AI model that is on cloud.
So Why is the AI at the Edge so Important?
Network impacts – Network communication being costly, the power consuming, poor bandwidth, and poor connection.
Latency considerations – The importance of real-time decision making is vital for many applications for example the GPS and self-driving cars.
Security concerns – Using sensitive data like personal, proprietary IP, financial data, and health data on the cloud is crucial. It can be easily stolen on the cloud.
Optimizations for the local inference – Can be helpful in achieving great efficiency with Edge AI models especially for hardware devices.
Therefore, it is important to note that the AI models on Edge are useful for building IoT (Internet of Things) devices, wearable devices, and hardwares applications. It is also important to note that not all devices or applications need the Edge AI models.
If you have any question or comment, do not hesitate to ask us.
Quote: The moon looks upon many night flowers; the night flowers see but one moon. – Jean Ingelow
tech-authority 