As research on artificial intelligence continues to become more and more prevalent, and with new successes being made regularly in the field, there is growing concern amongst researchers that artificial intelligence could very well lead to a disaster for humanity, though not in the way sci-fi movies portray it. The concern is not so much that a super-intelligent computer will rise up against its masters and go full Terminator on humans; it’s considered much more likely that AI leads to mass unemployment, damaging the world by replacing workers in jobs, than AI directly causing death and destruction.
Though computers have become more adept at solving complex problems, the system of intelligence required to make Terminator-style decisions has been completely non-existent in computers, even as computers have passed other intelligence thresholds that were thought to be untouchable.
Over the last few years, computers have mastered a number of tasks that many people thought could never be fully addressed by a computer alone. For example, look at the game of Go, which has recently been in the news due to Google’s AI program AlphaGo.
Even just three or four years ago, many AI experts were claiming that the game was a decade or more away from being solved, while many masters of the game believed that computers would never be able to beat a master of the game.
Of course, AlphaGo defeated professional Go player Lee Sedol in a high-profile match last March, and managed to take down the world’s highest ranked Go player, Ke Jie, in three unofficial matches early in January. Now, AlphaGo is not thought to be a threat to humanity, and expertise in Go is probably not a great baseline for high-level decision making by computers. AlphaGo has been designed and trained for years by a team at Google to handle one very specific task.
Most science fiction portrays the AI threat as the exact opposite, adaptable to every situation and more intelligent than human beings. But the ways in which computers could upend society are more related to AlphaGo than not.
Take self-driving cars as an example. A decade ago, the idea of a computer being able to drive as well as, or even better than, a human was considered science fiction in itself. Google’s autonomous vehicle project changed all that in 2009, and since then virtually every company that produces anything automobile-related has been forced to dive head-first into autonomous vehicle development or risk becoming obsolete as soon as self-driving cars hit the road. The results of this rapid development have astounded even AI experts, who didn’t expect self-driving cars to be able to drive with humans anytime soon.
At this point, multiple companies expect to have computers driving humans around in the next two or three years. Though many people have reservations about riding in a car driven by a computer because of fear for their own safety, the real danger in self-driving cars is thought to be through the jobs they will replace.
The United States Bureau of Labor Statistics puts the number of heavy truck drivers at 1.8 million, and the number of taxi and limousine drivers at over 400,000. Autonomous vehicles are expected to replace those 2.2 million jobs eventually, dramatically shifting the picture of American employment. Elon Musk, the CEO of Tesla and SpaceX and one of the world’s most renowned technologists, agrees that self-driving cars could threaten a large amount of jobs.
On Monday, at the World Government Summit in Dubai, Musk discussed how self-driving cars will both be useful to society, but also create some unforseen problems. He says autonomous vehicles are “going to happen much faster than people realize and it’s going to be a great convenience.”
But he also spelled out the possible increase in unemployment, continuing that “driving in various forms” might be “the single largest employer of people.” Musk shares unemployment concerns with many other economists and AI researchers. Currently, the BLS puts the unemployment rate at 4.8 percent, with 7.6 million unemployed Americans seeking jobs. Immigrants and American manufacturers moving overseas are often seen as the main causes of millions of Americans being unable to find a job.
The reality, however, is that automation is a driving cause. Autonomous vehicles could contribute to dramatic unemployment in another industry. The founder of Google’s self-driving car program, Sebastian Thrun, sees this effect as even more dramatic, claiming that “no office job is safe” from automation.
As an artificial intelligence and robotics expert, former Stanford professor and creator of the project that spearheaded the evolution of AI to fit self-driving cars, Thrun may be uniquely qualified to make such a claim.
In Japan, for example, one company is already working to replace workers in what is typically thought of as an office job and not something at risk of disappearing to work by AI. Fukoku Mutual Life Insurance started the year by announcing it will use an application of IBM’s Watson AI to eliminate 34 insurance claim workers. The new system is expected to save over $1 million in employee salaries annually, as well as improve productivity by 30 percent. There are over 300,000 insurance claims workers in the U.S. according to the BLS, and those jobs could be just as much in jeopardy as truckers and taxi drivers. The software being used by Fukoku is just as accessible to American companies.
In fact, there is no reason the software only has applications in the insurance business. Any job that involves a large amount of reading, searching and categorizing of information is just as risky. This means jobs like law clerking, typically considered good, high-paying jobs, may not exist as we know them for much longer.
As computers continue on this trajectory, conquering more problems thought to be untouchable, the only consensus on how society will be affected is that no one actually knows how society will be affected. “Anyone making confident predictions about anything having to do with the futurte of artificial intelligence is either kidding you or kidding themselves,” said MIT economist Andrew McAfee in a recent interview with Wired. Systems like AlphaGo and autonomous vehicles are often seen as threatening because they solve problems seen as far too complex for computers to master.
This belief is in part due to the creativity challenge in the tasks of driving and playing Go, since computers are not usually thought to be able to utilize creativity in the way that humans can. However, AI research has begun addressing creativity problems; computers were thought to be incapable of mastering Go, the argument goes, who’s to say they will never be able to think creatively like humans?
Creativity in artificial intelligence could displace entire industries. For example, there are now a number of websites that can create a logo for a company or organization instantly using AI.
In fact, I just tried out one of these sites, called Logojoy, to see how easy the process is. By selecting a handful of logos I already liked and giving a small amount of information about the Concordiensis, I was given dozens of possible new logos for this newspaper. The cost of the logos was significantly cheaper than paying for a professionally done custom logo, but this is probably mitigated by a difference in quality.
Though this difference in quality currently keeps the graphic design industry safe, its unlikely that experts like Thrun, Musk and McAfee would bet that computers will never meet the human-quality threshold. If driving jobs, office jobs and creative jobs are all at risk of being lost to artificial intelligence, is any job safe? Many point to high-education jobs in industries like medicine as some of the safest bets. But, there are a number of startups working to diagnose illnesses using AI.
One company, the Israeli startup Beyond Verbal, is testing its system by using voice recordings of people scheduled for a coronary angiography and attempting to use just this recorded conversation to diagnose the caller for artery disease. A number of other startups are working to do similar diagnoses by voice only, for a wide range of medical issues. Mental problems like ADHD and PTSD are some of the most promising possibilities.
As this technology becomes more accessible to companies and, more importantly, more accurate, the necessity of doctors could decrease, especially for reasons like annual checkups. While this would currently be seen as a much-needed reprieve for an industry plagued by a shortage of doctors, a medical degree is seen as challenging but all-but-certain route to a well-paying job for life. Upending that employment ecosystem could be damaging to a population that would be very unprepared for unemployment.
Few experts would argue that mass unemployment by artificial intelligence will happen within the next year or two, but even fewer would suggest its unlikely to ever happen. With new applications of AI challenging the limits of computers in all types of industries, many people believe it’s only a matter of time before computers come for many jobs. As AI has conquered challenges that it was never expected to be capable of solving, it would be foolish to assume any job is safe. With enough time, Musk, Thurn and McAfee all believe humans are at risk of losing their status as the most productive workers on the planet. Now, how long until they come for newspaper writers?