CAMBRIDGE, MASS. -- Most people have yet to take a ride in a driverless car, but taking a flight on an automated...
airplane? That's practically a standard. "Commercial pilots, today, touch the stick during an entire flight for three to seven minutes," Associate Professor Mary "Missy" Cummings, director of the Humans & Autonomy Lab at Duke University, said at the MIT Sloan CIO Symposium. "And that's on a good day."
Cummings, a panelist at the event, is familiar with automation -- and not just as an academic. One of the Navy's first female fighter pilots, Cummings worked with automated systems that controlled the jet's takeoff and landing on aircraft carriers.
How automated? Pilots had to prove they couldn't possibly interfere with the automated control system in the cockpit by raising their hands above their heads and grabbing onto a pair of handlebars. "The planes are programmed to fly so close to the stall margin that if you touch anything, you can induce problems in the control system," she said.
This was back in the mid-1990s, before the technological revolution of social, mobile, cloud and big data. For Cummings, the writing was on the wall. "This was a big epiphany for me as to why I needed to look for another career," she said.
The 'combinatorial' effect of digital goods
Erik Brynjolfsson, director at the MIT Initiative on the Digital Economy and moderator of a panel on the impacts of automation, said that modern technology is "advancing all three of these areas that used to be uniquely human: dexterity, language and problem solving."
These advances are happening partly because digital goods aren't finite the way natural resources are, Brynjolfsson said. "There is a mindset from traditional economy -- from agriculture, from mining -- that as you do more and more, you're going to use up the good land, find the easy gold or oil," he said. "But that doesn't really apply to ideas."
Instead, digital goods are "combinatorial." When one of Brynjolfsson's students built an app in a few weeks, he was able to reach an audience of a million people in just a few months precisely because he could build on top of innovations that came before his -- on top of Facebook, the Internet and other digital networks with impunity.
"These earlier innovations, rather than making it harder to do things, rather than using up the low-hanging fruit, are actually making it easier to build other things," Brynjolfsson said. "And this is getting at combinatorial innovation. Combine building blocks in new ways, and the more building blocks you have, the more new things you can create."
Among the new things this aggregation of building blocks has created are automated systems … everywhere -- from the fast food industry to the financial industry, from the legal profession to journalism. "The next time you read an article … look a little more closely at who the author is," Brynjolfsson said. "You may be surprised that in thousands of cases, the author is Automated Insights."
Augmenting -- for now
Indeed, Automated Insights (Ai), acquired by private equity firm Vista Equity Partners in February, generated 1 billion stories last year with its Wordsmith platform, which included general earnings reports and sports recaps for the likes of the Associated Press. "But how we get to a billion stories is through personalization," Robbie Allen, CEO and founder of Ai, said during the panel.
Since 2012, for example, Ai has been teaming up with Yahoo Inc. to produce "personalized narratives for the millions of Yahoo fantasy football users" on a weekly basis, according to a press release. "The real promise of what a technology like ours can deliver on is the ability not to create one story that a million people read, but to create a million stories each person reads individually," Allen said.
Personalized fantasy football reports, to be sure, represent a new content area that likely wouldn't have existed without this kind of technology. But Ai's "robot writers" are also encroaching on traditional journalism areas.
Earnings reports, another specialty area for the Wordsmith platform, used to be written by journalists. At the Associated Press, an investor in the technology, Ai robots have upped production from the 300 quarterly earnings reports produced by AP journalists to more than 3,000.
Allen said that earnings reports aren't 100% automated, nor do they have to be. Journalists still produce reports on the most influential companies, he said, and, in some cases, the technology acts more as a research assistant. "We're not supplanting existing writers; we're augmenting," he said. At least for now.
For some industries, though, pushing the automated envelope further isn't possible just yet. "Robots have already automated many tasks in factories," said Daniela Rus, professor in the electrical engineering and computer science department at MIT and the director of the institute's Computer Science and Artificial Intelligence Laboratory. "But these robots are only used in very narrow domains."
The automobile industry, which automates 80% of factory tasks, can take advantage of automation because of the consistent repetition, Rus said. For the cellphone and general electronics industry, where products are evolving rapidly, automated systems only perform 10% of factory tasks.
"Part of the reason is that every automation line requires special tooling, requires special configuration, and this takes many years," Rus said. "So if your product is going to change every three months, there won't be time to change the tooling or reconfigure the factory plans -- at least not today."
One of the hurdles? Building automated systems has yet to be automated. "If you wanted to have lots of flexible manufacturing," Rus said, "you need to figure out how to make machines faster."
Smart robots are coming to the enterprise. What their arrival means for IT.