Facebook told its employees on Tuesday that it is developing a tool to summarize news articles so that users don’t have to read them. It also explained early plans for a neural sensor to detect people’s thoughts and put them into action.
Those announcements and product demos were part of a company-wide end-of-year gathering at the social networking giant, whose year was marked by controversy, employee dissatisfaction and multiple state and federal antitrust cases. BuzzFeed News received audio from the meeting, which was not public, but was broadcast virtually to thousands of employees.
Led by CEO Mark Zuckerberg, the meeting featured a slew of pre-recorded business executives, some of whom called 2020 a difficult year as the company weathered a global pandemic and the backlash of the police killings on George Floyd, Breonna Taylor and other black Americans .
Despite the turmoil, the company’s leaders said the social networking company has made progress, adding some 20,000 new employees this year. With more people in-house around the world, the company has experienced record use, said Mike Schroepfer, Facebook’s chief technology officer. Traffic in March seemed like New Year’s Day, typically the busiest time of the year on Facebook, he added.
“Our investments in technology aren’t just about keeping our services running,” he said, comparing the speed of Facebook’s messaging improvements with advancements in the COVID-19 vaccine. “We’re paving the way for groundbreaking new experiences that, without exaggeration, will improve the lives of billions of people.”
Facebook spokesman Liz Bourgeois declined to comment on this story.
Among the claims touted by Schroepfer was the company’s commitments to artificial intelligence, which was often seen internally as a panacea for the social network’s ills. He noted that Facebook’s data centers received “ new systems ” that would make them 10 to 30 times faster and allow Facebook’s artificial intelligence (AI) to train itself.
“And it’s actually the most important tool we are using in production right now to combat hate speech, misinformation and frankly the hardest possible content issues we face,” said Schroepfer, noting that Facebook is now 95% of all hate speech on the platform.
In recent weeks, outgoing Facebook employees have pushed back the idea that AI could fix the company’s content moderation issues. While Facebook employs thousands of outside human moderators, it has made it clear that AI is how it plans to patrol its platform in the future, an idea that concerns employees.
“AI is not going to save us,” wrote Nick Inzucchi, whose December suicide note was obtained by BuzzFeed News. “The implicit vision that guides most of our integrity work today is one in which all human discourse is controlled by perfect, honest, omniscient robots owned by [CEO] Mark Zuckerberg. This is clearly a dystopia, but one so deeply ingrained that we hardly notice it anymore. “
Another outgoing employee earlier this month estimated that, even with artificial intelligence and external moderators, the company “deleted less than 5% of all hate speech on Facebook.” Facebook later got back to that claim.
During Tuesday’s meeting, the company also unveiled an AI assistant tool called “TLDR” that could summarize news articles in bullet points so that a user wouldn’t have to read the full piece. Named after the online acronym for “too long, don’t read,” the tool is also reportedly able to provide spoken commentary as well as a vocal assistant to answer.
The news about the tool, first shared by BuzzFeed News, did not go down well with members of the media who were frustrated with Facebook’s absorption of large amounts of online ads and the creation of a platform where news competes with misinformation and unreliable sources. In Australia, lawmakers have forced Facebook to pay media organizations for news, while in other parts of the world, the company has partnered with a number of “trusted” outlets, including BuzzFeed News, to pay for the use of their content.
“I sometimes feel like there is someone at FB headquarters whose job is to come up with new ways to completely destroy any semblance of intelligence in America,” tweeted WNYC Editor-in-Chief Audrey Cooper.
Other projects announced or demonstrated by Schroepfer on Tuesday included one Star Trek–such as Universal Translator and “Horizon”, a new social network in virtual reality where users can hang out with their avatars.
He also detailed a neural sensor for reading commandments from people’s brains.
Following the acquisition of neural interface CTRL labs in 2019, Facebook demonstrated its progress in the field with a sensor that “takes neural signals from my brain, down my spinal cord, down my arm, to my wrist” and enables a user a physical action Schroepfer noted that it can be used to type, hold a virtual object, or control a character in a video game.
“We all have the privilege of seeing the future because we make it,” he said.
Still, Facebook’s chief technology officer seemed to anticipate any criticisms of the products – or past failures – by proposing security measures.
“We must build responsibly to earn trust and the right to keep growing,” he said. “It’s imperative that we do this right so that people around the world get all these great technologies … without experiencing the drawbacks.”