Facebook has put its hands on designing in-house chips that can analyze and filter video content, according to Yann LeCun, the company’s chief AI scientist.
Social media is more commonly used for sharing personal life with friends and family. Although video posted on Facebook may require little filtering or analysis, Facebook sees it in a different way. “Let’s imagine someone uses Facebook Live to film their own suicide or murder, you’d like to be able to take down that kind of content as it happens,” LeCun said.
From the economical perspective, AI-enabled chips are also much more energy-efficient, compared with using conventional computers to monitor videos. Facebook currently uses Intel CPUs for many of its AI services.
Intelligent video analysis can not only help Facebook filter content for violations of terms of service, but also remove content posted by terrorist organizations like ISIS.
Facebook is just one of the companies that are throwing down resources to design energy-efficient, AI-enabled chips, LeCun said at the Viva Technology industry conference held in Paris recently. “You’ve seen that trend from hardware companies like Intel, Samsung, Nvidia. But now you start seeing people lower in the pipeline of usage having their own needs and working on their own hardware,” Bloomberg quoted LeCun’s statement in a report.
Along with Facebook, many technology companies want to rely less on chipmakers like Intel and Qualcomm, and to supply themselves. Apple, Google and Amazon are reportedly planning on making specialized chips to make better smart products.
Many of today’s smartphones are equipped with chips that enable speech recognition, as well as image and video processing. This trend is growing, and as a result, more and more software companies are paying more attention to hardware development, LeCun said.
Facebook has done this before. It made its own server design, motherboards and communication chips for data centers, LeCun said, adding that the AI chip venture “is not completely new for Facebook.”
Qualcomm Said to Improve AI Chip
Still, major chipmakers are fully capable of improving their products’ AI capability. Qualcomm, for example, recently provided a glimpse into its work on a new voice recognition program that can enable edge processing with 95-percent accuracy.
Chris Lott, an AI researcher at Qualcomm, said the new technology works on two neural networks, according to a VentureBeat report. The first one is called the neural network, which uses memory to process inputs. The other one is called neural network, which mimics the connectivity pattern between neurons in the human brain.
“It learns from patterns [and] from your use of the device,” Lott said, adding that “it can personalize its behavior to you.”