Will computers automate our lives? Will all of our coworkers be robots, or invisible programs, whirring away inside banks of computers? Will robots steal our jobs? Will super-intelligent AI wipe out humanity? These questions aren’t reserved just for science fiction writers anymore.
If you read the headlines about robots and artificial intelligence, then it’s likely that for the past few years, you’ve been vaguely worried that a robot will eventually steal your job. And more recently, you’ve started to ponder what it would look like if scientists create artificial intelligence that’s actually super intelligent. What jobs will be left when software can drive cars or translate speech better than humans can? What happens when machines become more intelligent than we are?
All of these questions make it sound like the world as we know it is quickly coming to a definite and dystopian end. But many researchers who work with artificial intelligence are weary of hearing apocalyptic projections about the effects of super intelligence, and think that they detract from discussions of what some would characterize as more serious — perhaps plausible — scenarios. Among these researchers is Baidu chief scientist Andrew Ng.
Wired’s Robert McMillan reports Ng said at a recent artificial intelligence conference that he considers such discussion “a distraction from the conversation about…serious issues.” A group of AI experts who recently convened at a retreat in Puerto Rico, came to the consensus that while there are both short-term and long-term artificial intelligence issues to worry about, it’s the long-term questions that are getting all of the press. So what are the short-term effects, and the serious issues we should be worried about? It turns out that artificial intelligence really could take your job — and that could have a serious effect on our society within the next few decades.
The second machine age
Two MIT academics, Erik Brynjolfsson and Andrew McAfee, argue that we’re entering a “second machine age,” in which the accelerating rate of change has the potential to leave millions of medium- and low-skilled workers behind. It will be much harder for these workers to find new jobs and new industries. Many of the industries of the future will be automated from the start, and require much less labor than those of the past.
As Farhad Manjoo wrote for Slate back in 2011, even highly educated workers who make high salaries will be battling machines for their jobs. Those at risk include doctors, lawyers, pharmacists, scientists, and even writers. Manjoo cites evidence that information technology has already damaged workers in a large sector of the economy, those workers who are “middle skilled.” Their jobs include many that are often recognized as antiquated — administrative workers, repairmen, and factory workers — in categories where the number of jobs has rapidly declined. These jobs consist of tasks that are both routine and geographically portable, ideal for automation.
Murray Shanahan, a professor of cognitive robotics with Imperial College, told Wired that artificial intelligence is likely to start having a significant effect on society over the next five to 10 years. Ng says that it took the U.S. about 200 years to switch from an agricultural economy, where 90% of the country worked on farms, to the current economy, where only about 2% of the country works on farms. The shift required people to move off farms and learn new skills, and thus contributed to massive growth.
But things are expected to be different this time around. Shifts brought by artificial intelligence promise to take hold much faster. For example, artificial intelligence technologies like the self-driving car are poised to be massively disruptive over a much shorter period of time than the industrial revolution needed to take hold. McMillan points out that there are three million truck drivers in the United States — all of whom could be put out of a job if driverless vehicles take hold in a matter of years.
Ng says that if artificial intelligence begins taking jobs in earnest, retraining large numbers of workers would present a major challenge, and that “our education system has historically found it very difficult.” Complicating the situation even further is the fact that many young, high-profile companies need to employ relatively few people in the process of providing their technology-enabled services, a trend unlikely to change with the new companies that rise to prominence in the next several decades.
So will the details of this second machine age play out? MIT’s Technology Review reported in 2013 that researchers at the University of Oxford projected that 45% of American jobs are at high risk of being taken by computers within the next two decades. The researchers believe that this takeover will happen in two stages. In the first, computers will start replacing people in vulnerable fields like transportation and logistics, production labor, and administrative support. Jobs in services, sales, and construction could also be automated during this stage. After that, the researchers project that the rate of replacement will slow down due to a “technological plateau” in fields that are harder to automate, like engineering.
That plateau will be followed by a second wave of computerization, which will depend on the development of good artificial intelligence. It’s at that stage that jobs in management, science, engineering, and the arts will be at risk. The researchers write, “Our findings thus imply that as technology races ahead, low-skill workers will reallocate to tasks that are non-susceptible to computerization — i.e., tasks that required creative and social intelligence. For workers to win the race, however, they will have to acquire creative and social skills.”
Creativity and common sense
PCWorld reports that Andrew McAfee, co-founder of the Initiative on the Digital Economy at MIT, says that many knowledge workers today get paid to do things that computers will soon be able to do. Software has replaced human payroll processors, and artificial intelligence will move up the skill ladder to replace middle-class workers, like financial advisors. But there are some jobs that artificial intelligence won’t find it so easy to take over, particularly those that require very human attributes like creativity, social skills, and common sense.
Paul Cohen, program manager in the Information Innovation Office at the U.S. Defense Advanced Research Projects Agency (DARPA), says that artificial intelligence has limitations to overcome. Scientists need to overcome the “problem of common sense,” which simply means equipping a computer with the same broad base of knowledge that a typical five-year-old has at his disposal. Peter Bock, an emeritus professor of engineering at George Washington University, projects that scientists will be able to build an artificial intelligence device that matches the processing power of a human brain within 12 years. Then the AI device would take several years to learn the information it needs to function like a human brain, just as a child needs to learn over time.
The effects of artificial intelligence are already starting to show in some lines of what we like to think of as creative work, taking on basic tasks that can be automated and completed more efficiently by a machine than by a human. The Verge reports that the Associated Press now publishes 3,000 stories on companies’ earnings per quarter that are written and published by a computer system by Automated Insights. When the program first began in July, every automated story was reviewed for errors by a human editor. Full automation began in October, when stories “went out to the wire without human intervention.”
Both the Associated Press and Automated Insights say that no jobs have been lost due to the service, even though the automated system logs fewer errors than the human-produced equivalents of the past. But the automation frees up writers to spend their time on more nuanced reports with unique angles, and to think more critically about the bigger picture behind the numbers in companies’ quarterly earnings reports. That’s a task that artificial intelligence won’t be up to anytime soon.
Similarly, DARPA has automated the process of assembling code from resources like StackOverflow, in the hope that artificial intelligence programs will eventually be able to write complex, original code. So far, humans still need to tell the program what the final code should do, and a time when a program could actually write code is still years off. McAfee says that to deny the possibility of AI writing code, “you’d have to believe there’s something ineffable about the human brain, that there’s some kind of spark of a soul or something that could never be understood. I don’t believe that.”
Regardless of your philosophical view on the difference between a human and a machine, McAfee does note that there are some complex, creative tasks that humans can complete that have “proved really, really resistant to understanding, let alone automation.” He explains, “I think of programming as long-form creative work. I’ve never seen a long-form creative output from a machine that is anything except a joke or mishmash.”