Product Categories

Contact Us

Address: Guangdong Province Shenzhen Futian District, #2504, Lei Zhen Building, FuMing Rd.

TEL: +86-755-88603570

Mob: +86-18938643630

FAX: +86-755-88603571

Home > News > Content
Facebook Is Developing Video Filter Chip May 30, 2018

Yann LeCun, Facebook’s chief scientist, said at the Viva Technology Summit in Paris that Facebook is developing its own computer chip and will be more energy-efficient in analyzing and filtering real-time video content. Lechuune said that traditional methods are inefficient and consume a lot of energy and computing power.
Last month, there was news that Facebook was developing its own semiconductor chip, and now the news has been confirmed.
In the AI service, the current CPU used by Facebook is from Intel. If you switch to a dedicated chip, Facebook may be able to increase the speed of video filtering and quickly find content that violates service policies. At present, Facebook has introduced various forms of computer vision technology, such as deleting the content sent by terrorist organizations.
AMD and Nvidia continue to introduce more advanced GPUs, and global unlabeled data (such as pictures and video data) is also increasing. Google has been accelerating the speed of AI training and deploying TPU. The Microsoft Project Brainwave project uses Intel Stratix 10, an FPGA chip.
LeCun said that more and more smart devices are equipped with powerful chips, allowing users to directly apply voice recognition, enhancement and virtual reality, as well as image and video processing functions on their devices. He said that this trend will become more apparent, and will urge more software-based companies to consider the direction of the hardware.
"Before Facebook invested in hardware: Designed its own server architecture, motherboard and its own communication chip for the data center." LeCun said. "So this is not a strange thing for Facebook."
Compared with other technology giants, Facebook's start can be said to be a lot late, Apple, Google, Amazon are currently also developing dedicated chips to improve the speed of AI services to support the development of hardware products.

In 2010, Apple began to launch its own chips and applied them to its multiple product lines. At present, A-series chips have been applied to the 11th generation in the market. Apple plans to use the self-developed Mac chips to replace Intel chips as early as 2020.
Google has independently researched and developed chips for many years. The AlphaGo system known for the Go game is a TPU tensor processor developed by Google. In addition, Google also specifically designed an image processor for Pixel 2.
Amazon has a 449 employee-scale chip team. The chip designed by this team will support Amazon's next-generation Echo products in the future and improve Alexa's response speed and service quality.