Facebook Focuses on an AI-Driven Future

January 7, 2021

Thousands of Facebook employees were treated to project announcements and product demos at the company’s end-of-the-year meeting. BuzzFeed News got its hands on an audio recording of the proceedings and shares a few highlights in, “Facebook Is Developing a Tool to Summarize Articles so You Don’t Have to Read Them.” As have many of us, Facebook has had a challenging year. However, company executives painted a positive picture at the meeting. We are not surprised to see the company pinning many hopes on AI. Writer Ryan Mac reports:

“Despite the turmoil, the company’s leaders said the social networking company has moved forward, adding some 20,000 new workers this year. With more people around the world at home, the company has experienced record usage, said Facebook Chief Technology Officer Mike Schroepfer. … Among the advancements touted by Schroepfer were the company’s commitments to artificial intelligence, which has often been seen internally as a panacea to the social network’s ills. He noted that Facebook’s data centers were receiving ‘new systems’ that would make them 10 to 30 times faster and allow Facebook’s artificial intelligence (AI) to essentially train itself. ‘And it is actually the key tool we are using right now today in production to fight hate speech, misinformation, and honestly the hardest possible content problems we face,’ Schroepfer said, noting a company talking point that Facebook now detects 95% of all hate speech on the platform. In recent weeks, departing Facebook employees have pushed back on the idea that AI could cure the company’s content moderation problems. While Facebook employs thousands of third-party human moderators, it’s made it clear that AI is how it plans to patrol its platform in the future, an idea that concerns workers.”

The employees are right to be concerned. Experience shows AI is still a long way from consistently discerning nuance in human language. Facebook will save money by creating algorithms over hiring people to moderate the platform, but that will do little to stem the tide of false information or hateful speech. Another new tool that leverages AI looks like another way to spread false, or at least incomplete, information—”TL;DR” (the popular abbreviation for “Too Long, Didn’t Read”) will summarize news articles into bulleted lists. No word on whether this is connected to Semantic Scholar’s tool by the same name designed for use with academic texts.

Other curious announcements include the development of a universal translator (a la Star Trek) and “Horizon,” a VR social network where users’ avatars can hang out together (inspired by pandemic isolation, perhaps?) Then there is the brain-reading device. Last year Facebook bought neural-interface startup CRTL-labs and has since made progress on a device that can translate thoughts into physical actions. Potential applications, says Schroepfer, include typing, manipulating virtual objects, or driving a character in a video game. Will they put that together with the Horizon project? Hmmm.

Cynthia Murrell, January 7, 2021

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta