Silicon Valley is going back to an ancient technology: People

Tech companies have long been valued by investors for their ability to replace employees with technology. Now, alongside software and server farms, they are moving at a breakneck pace on find living, breathing human beings to staff their systems.

They’re doing so because of a high-profile series of failures of automation, which have prompted a wave of intense pressure from investors, the public, and governments.

Tesla’s highly automated production line failed to produce cars at the rate CEO Elon Musk promised, prompting questions about the electric-car maker’s solvency. Systems at Google’s YouTube failed to flag extremist and exploitative videos. Russian operatives have worked to influence elections using Facebook, whose systems separately created categories of users with labels such as “Jew hater” that it then allowed advertisers to target.

While companies such as Google and Facebook still insist that they’re just distribution platforms rather than content creators and bear limited, if any, responsibility for most of the content they host, they’re increasingly acknowledging they need to do something to curb abuses. In the short-term at least, that approach usually involves more humans.

“Human are underrated,” tweeted Musk, as the company struggles to ramp up production of its Model 3 sedan. Musk has blamed an overly automated production process. “We had this crazy, complex network of conveyor belts… And it was not working, so we got rid of that whole thing,” he told CBS.

Meanwhile, Google and Facebook have been hiring thousands of people to monitor content and advertising on their platforms, amid backlash against their hosting of extremist videos and messages, videos depicting the exploitation of children, propaganda, and content created to manipulate electorates in the US and elsewhere.

Facebook CEO Mark Zuckerberg reiterated to US legislators last week that the company planned to double its security and content moderation workers to 20,000 people by the end of the year—an investment that he acknowledged would hurt its profitability.

YouTube CEO Susan Wojcicki in December said the Google-owned video site aimed to have 10,000 people working to find and combat content that violates its policies, a 25% increase according to BuzzFeed.

Artificial-intelligence experts say Zuckerberg and other tech executives are over-optimistic about the timeline for computers identifying things such as toxic speech, and point to existing systems that fail at that task. A new Barclays research report says that humans are better than robots at “sensorimotor skills” and “cognitive functionality,” meaning humans are less clumsy than robots and are better at making decisions factoring in context and in cases where there’s incomplete information. There are reasons to be confident that humans will retain some of those advantages for decades into the future.

But any surge in hiring by tech companies is unlikely to significantly offset the toll on employment from the current wave of automation. And the jobs that such companies are hiring for at scale—such as people to watch videos for offensive content—tend to require lower skills, and pay lower wages.