If you can't beat the robots, make them
A CEO explains how he knew artificial intelligence would eventually take his job.
In an old American folktale, John Henry is the strongest railroad builder ever. When a company invents a steam-powered drill that can supposedly drive down stakes faster than any human could, Henry demands a competition: him and his hammer versus the drill. Henry wins in the end, but he dies from exhaustion.
A strangely similar situation occurred in a little office in the Israeli city of Tel Aviv, when Israeli entrepreneur Or Shani became the guinea pig in his own experiment. It was 2012 and he'd just started building Albert, a program that uses artificial intelligence to craft digital marketing strategies. A gaming company agreed to let him test his creation; both Shani and Albert would have two weeks to find new clients. The winner could indicate the future of marketing.
Albert Technologies is now a much bigger company, with about 80 employees in Israel and 20 in New York. Despite his success, Shani never planned on starting a business. He was working as a media buyer and data analyst years ago, and he just happened to notice that a lot of the work he did was repetitive.
"It was very much like an assembly line," he told me at his New York office.
He realized that the process could be automated. Not only could be ... would be.
“I knew it was going to happen," he said. “Finally, I was like, ‘If no one’s doing it, I’ll do it.'" He saw the way the tide was turning and jumped in a boat. After all, he knew robots would take his job eventually.
Albert works by separating potential customers into hundreds of thousands of clusters. The program tries out different Facebook, email and other online strategies with different clusters. Albert figures out exactly when people should see which ad. If Albert is targeting you, for instance, it might intentionally show you a Facebook ad on Monday, send you an email on Wednesday and predict you'll actually click on Friday. If it predicts incorrectly, it just realizes it has placed you in the wrong cluster and corrects the mistake. No human could do this kind of work; it's impossibly time-consuming for anyone but a machine.
Shani knows he's part of a larger trend toward automation.
“Everything that you can mechanize will be mechanized," he said. "That’s been how things have been going since the Industrial Revolution."
Unlike many startup CEOs who insist their inventions will save the world, Shani has mixed feelings about the oncoming wave of automation. Though automation will likely make many new inventions and processes possible, Shani worries about people like his father, who was a bus driver for 47 years. Many people like him might not be able to jump on a new trade so easily, once things like self-driving cars and buses take over.
“It’s a brave new world," he said tentatively.
He also worries for future generations. It could be tough to get a job once human labor just ceases to be all that useful.
“I have no idea how we’ll evolve," Shani said. "It’s going to be interesting.”
Matthew Ruttley, a data scientist in New York who has worked at companies that use artificial intelligence like Mozilla and Frame.io, agrees with this outlook. "We’ll have a lot of time on our hands, and we might really have to think about basic income," Ruttley said. "Or the job market will shift."
Though Shani thinks some jobs, particularly ones that require creativity, will always be filled by humans.
“Marketing is all about a story," he told me. "Machines can’t come up with that.” No machine, for instance, could have thought up the Ice Bucket Challenge. “It was such a stupid idea, but it worked," he continued.
During that fateful competition years ago, when Shani went up against Albert, he settled down for his two weeks of digital marketing. But Albert won in only two days.
“It was a very happy moment, but also a very insulting moment," he remembers. After all, Shani wasn't just a regular marketer. He was quite experienced and, possibly more important, motivated to win.
Albert was just a newly minted prototype. But he was much, much better.
MORE FROM THE GRAPEVINE:
Related Topics: Science