Mark Zuckerberg believes AI will fix Facebook, but he can’t say how
Mark Zuckerberg said that over the next five to 10 years, artificial intelligence would prove a champion for the world’s largest social network in resolving its most pressing crises on a global scale
Artificial intelligence will solve Facebook’s most vexing problems, chief executive Mark Zuckerberg insists. He just can’t say when, or how.
Zuckerberg referred to AI technology about 23 times during his five-hour testimony before a joint Senate committee hearing on Tuesday, saying that it would one day be smart, sophisticated and eagle-eyed enough to fight against a vast range of platform-spoiling misbehaviour, including fake news, hate speech, discriminatory ads and terrorist propaganda.
Over the next five to 10 years, he said, artificial intelligence would prove a champion for the world’s largest social network in resolving its most pressing crises on a global scale – while also helping the company dodge pesky questions about censorship, fairness and human moderation.
“We started off in my dorm room with not a lot of resources and not having the AI technology to be able to proactively identify a lot of this stuff,” Zuckerberg told the lawmakers, referring to Facebook’s famous origin story.
Later in the hearing, he added that “over the long term, building AI tools is going to be the scalable way to identify and root out most of this harmful content.”
But Facebook’s AI technology can’t do any of those things well yet, and it’s unclear when, if ever, it will be able to.
Tech experts had a different opinion on why Zuckerberg spent so much time offering tributes to the much-hyped but largely unproven tech advancement: The shapeless technology could help the company pawn off blame from the humans creating it.
“AI is Zuckerberg’s MacGuffin,” said James Grimmelmann, a law professor at Cornell Tech, using the film term for a plot device that comes out of nowhere to help the protagonist to save the day.
“It won’t solve Facebook’s problems, but it will solve Zuckerberg’s: getting someone else to take responsibility.”
To many of us, AI is the amorphous super-technology of science fiction, animating friendly computers and killer robots.
Silicon Valley leaders often promote that vision of AI by suggesting that it will serve humanity as a quasi-magical force for good. But today’s artificial intelligence is now being used in far more basic forms: driving cars, tracking cows and giving a voice to virtual assistants like Siri and Alexa.
Facebook uses AI in understated but important ways, such as for recognising people’s faces for tagging photos and using algorithms to decide placement of ads or News Feed posts to maximise users’ clicks and attention.
The company is also expanding its use, such as by scanning posts and suggesting resources when the AI assesses that a user is threatening suicide.
Facebook, Zuckerberg said Tuesday, has also been “very successful” at deploying AI to police against terrorist propaganda.
“Today, as we sit here, 99 per cent of the ISIS and al-Qaeda content that we take down on Facebook, our AI systems flag before any human sees it,” he said.
Non-profit groups like the Counter Extremism Project have argued that Facebook has exaggerated its achievement and failed to crack down on well-known Islamist extremists.
Zuckerberg said he was optimistic that Facebook’s AI would, within five to 10 years, be able to comprehend the “linguistic nuances” of content with enough accuracy to flag potential risks. But those limited cases, experts said, were helped by geography and required human moderators to make the final ruling.
Robyn Caplan, a researcher at the think tank Data & Society, said Zuckerberg’s optimism seemed to clash with the more pragmatic conversations she has had with representatives from platforms similar to Facebook, who have stressed that AI can help flag questionable content “but cannot be trusted to do removal”.
Tech experts said AI could also create new problems. The same bad actors behind viral hoaxes and fake accounts may be able to use AI to evade Facebook’s filters, make fake videos or spearhead targeted harassment campaigns, they said.
AI also “can’t solve political problems; it can’t make people agree,” Grimmelmann said.
“What’s ‘fake’ news depends on who you ask: kicking the question over to AI just means hiding value judgments behind the AI.”
The real issue, experts said, is that the problems plaguing Facebook may be battles that no one can truly win. Instead of acknowledging that grim fact, though, Zuckerberg has taken to gesturing vaguely at the future – and at a version of AI that could eventually save the day.
As Ryan Calo, an assistant professor at the University of Washington law school, tweeted Tuesday, “’AI will fix this’ is the news ‘the market will fix this.’”