Facebook has used Intel CPUs for many of its artificial intelligence services, but the company is changing course to adapt to the pressing need to better filter live video content. At the Viva Technology industry conference in Paris, Facebook chief AI scientist Yann LeCun stated that the company plans to make its own chips for filtering video content, because more conventional methods suck up too much energy and compute power. Last month, Bloomberg reported that the company is building its own semiconductors.
VentureBeat reports that, “a transition to a specialized chip could help the company more quickly filter video content for violations of its terms of service, like if a person commits suicide in a livestream or performs another act of violence.”
The company “currently uses various forms of computer vision to do things like remove content by terrorist organizations like ISIS,” and “graphics processing units (GPU) from companies like AMD and Nvidia assist with the rapidly increasing amount of unlabeled data like images and videos.”
Other companies at work on their own chips to “speed up AI services for their hardware products” are Amazon, Apple and Google, the latter of which “continues to accelerate AI training and deployment with its tensor processing unit (TPU).” Microsoft’s Project Brainwave is relying on “Intel’s Stratix 10, a field programmable gate array (FPGA) chip.”
Bloomberg quotes LeCun as saying that, “there’s a huge drive to design chips that are more energy-efficient” for AI processes. “You’ve seen that trend from hardware companies like Intel, Samsung, Nvidia,” he said. “But now you start seeing people lower in the pipeline of usage having their own needs and working on their own hardware.” By doing so, companies such as Facebook would “depend less on chipmakers such as Intel and Qualcomm.”
LeCun also noted that, “more smartphones are equipped with powerful chips that let users take advantage of speech recognition, augmented and virtual reality, as well as image and video processing directly on their devices,” adding that the trend will grow. “Facebook has worked on hardware before,” he said. “It makes its own server design, motherboards, its own communications chips for data centers.”
At an earlier Bloomberg conference in Paris, LeCun said that, “while AI is helping Facebook tackle problems of extremist propaganda, fake accounts and hate speech, it’s not sophisticated enough to handle many of the most pressing issues facing the social network.”