AI is developing rapidly, and there are numerous startups and established companies actively developing hardware for artificial intelligence applications. At this moment, strong artificial intelligence cloud services (such as Google 's Alpha Go) have emerged, and people also hopes to bring artificial intelligence to portable devices which enables the capablity of merging future IoT applications.
Traditional methods of artificial intelligence on portable device is to send data to the cloud through the internet, and after cloud has performed the calculation, it will send the results back to the portable device, such as Apple 's Siri service.
However, this approach will encounter several problems. First, delay will occur in the use of network …show more content…
Therefore, applications that require low latency computing and data security for sensitive applications require artificial intelligence algorithms to perform on the portable device, or at least be apable of doing some pre-processing transfer a small amount of calculation results (rather than a lot of raw data) to the cloud to complete the final calculation, which requires the portable device to quickly perform these calculations. On the other hand, the portable device need to perform the calcuation using a small amount of energy, or else the battery will run out quickly. (a phone certainly cannot carry the Nvidia Pascal graphic card which has a power consumption of 200W+)
Currently, many companies are actively developing portable hardwares which make portable artificial intellegnce possible. There are two methods for hardware implementation on portable AI, the FPGA and ASIC method. Representative of FPGA is Xilinx, which mainly uses Zynq as the platform, while the representitve of ASIC is Movidius. I will talk about the pros and cons of the two methods in the following paragraphs.
FPGA vs. …show more content…
FPGA stands for Field Programmable Gate Array, the basic principle is the integration of a large number of digital circuits and memory within the FPGA chip, and users burned into the FPGA configuration file to define the connection between the digital circuits and memory. This burn is not permanent, means that the user can configure the FPGA to a microcontroller MCU, and can edit the configuration file making the same FPGA an audio codec tommorrow. ASIC stands for Application-Specific Integrated Circuit, once the design and manufacture is complete the circuit is fixed and can not be
Traditional methods of artificial intelligence on portable device is to send data to the cloud through the internet, and after cloud has performed the calculation, it will send the results back to the portable device, such as Apple 's Siri service.
However, this approach will encounter several problems. First, delay will occur in the use of network …show more content…
Therefore, applications that require low latency computing and data security for sensitive applications require artificial intelligence algorithms to perform on the portable device, or at least be apable of doing some pre-processing transfer a small amount of calculation results (rather than a lot of raw data) to the cloud to complete the final calculation, which requires the portable device to quickly perform these calculations. On the other hand, the portable device need to perform the calcuation using a small amount of energy, or else the battery will run out quickly. (a phone certainly cannot carry the Nvidia Pascal graphic card which has a power consumption of 200W+)
Currently, many companies are actively developing portable hardwares which make portable artificial intellegnce possible. There are two methods for hardware implementation on portable AI, the FPGA and ASIC method. Representative of FPGA is Xilinx, which mainly uses Zynq as the platform, while the representitve of ASIC is Movidius. I will talk about the pros and cons of the two methods in the following paragraphs.
FPGA vs. …show more content…
FPGA stands for Field Programmable Gate Array, the basic principle is the integration of a large number of digital circuits and memory within the FPGA chip, and users burned into the FPGA configuration file to define the connection between the digital circuits and memory. This burn is not permanent, means that the user can configure the FPGA to a microcontroller MCU, and can edit the configuration file making the same FPGA an audio codec tommorrow. ASIC stands for Application-Specific Integrated Circuit, once the design and manufacture is complete the circuit is fixed and can not be