DIGITAL LIFE
Facebook builds AI watchdog chips to filter live video

Facebook currently relies on a combination of automated assessment and filtering, and a human team of moderators. The latter work around the clock, responding to content that has been flagged by Facebook users as well as what’s surfaced by the automated systems. It’s unclear, though, just how scalable a solution that is.
As a result, Facebook has begun looking at alternatives. That includes working on a homegrown chip design that will be more effective – and more energy efficient – at analyzing and filtering live video. The project was revealed by Yann LeCun, chief artificial intelligence scientist at Facebook, Bloomberg reports, at the Viva Technology conference in Paris.
For Facebook, the potential is clear. “Let’s imagine someone uses Facebook Live to film their own suicide or murder,” LeCun theorized. “You’d like to be able to take down that kind of content as it happens.”
The idea is to use a dedicated artificial intelligence (AI) chip that would be focused on sifting through video and making real-time assessments of what the content shows. While that sort of analysis isn’t impossible with today’s hardware, it typically demands significant quantities of compute power. That has serious implications on energy consumption and cost.
“There’s a huge drive to design chips that are more energy-efficient for that,” LeCun confirmed. “A large number of companies are working on this, including Facebook.” The company already designs its own data center hardware, LeCun pointed out.
It’s not the first time we’ve heard chatter of Facebook looking internally for bespoke silicon. Back in April, reports suggested the social network was following in the footsteps of Apple, Google, and others, and developing its own chipsets. Multiple job listings for new roles focused on processors were spotted, covering a variety of topics.
At F8 2018, meanwhile, the company’s annual developer event, AI took center stage. That included a new set of video analysis tools, in addition to AIs that specialize in translation.
Chris Davies
No comments:
Post a Comment