Facebook builds chips to analyze live streaming videos

Facebook builds chips to analyze live streaming videos

Facebook is designing and producing its own energy-efficient chips to analyze and filter the content of Facebook Live Live Videos, Yan said. Yann LeCun, the company's chief artificial intelligence researcher, At the Viva Technology conference in Paris, traditional methods are no longer enough to require a lot of energy and computing.

"Let's imagine someone using Facebook Live to photograph his suicide or death," said Jan, speaking at the Paris Technology Conference. "You must be able to remove this kind of content immediately." He said that the use of systems based on traditional computers To monitor each video clip directly or recorded will require a huge amount of computing power, which costs a lot in terms of energy consumption.

Currently, Facebook uses Intel processors for many of its artificial intelligence services, so it can help the company move to a more specialized segment to filter video content more quickly when there is a breach of the terms of service, noting that the social networking platform is currently used for various forms of computer vision technology from Remove unwanted content.

According to Facebook officials, there is a strong incentive to design a company chip that is more energy efficient, and that there are a large number of companies working on this, including Facebook, so that hardware manufacturers such as Intel, Samsung and Infidia began this trend.

Bloomberg said last month that the social networking platform was building its own semiconductor design team, according to job listings and people familiar with it. The company, among other technology companies, is looking to provide itself and reduce its dependence on manufacturers Chipsets such as Intel and Qualcomm.

Google is another example of companies that build their own equipment to suit their own needs. Google surprised many companies when it built a Tensor processor to run the AlphaGo AI on machines faster in 2015 and announced it in 2016. The chip is able to help it jump forward with higher performance than the Tesla K80 graphics processing unit from NVIDIA.

It is worth noting that Facebook has previously designed its own server architecture, mainboards and communication chips for data centers. Social Networking seems to be on track to design its own ML solutions to achieve its goal of having the ability to filter live streaming video content in a timely fashion Actual.