Let’s Make AI Work for Everyone
“Machines will be capable, within 20 years, of doing any work a man can do.”
Herbert Simon wrote this quote in 1965, yet five decades later, most of us still have jobs. Today, rapid progress in artificial intelligence (AI) and robotics has reignited speculation about a utopian future where all labor is done by robots and humans are free to pursue their dreams. Unfortunately, this dream is incompatible with our current social and economic system. If your job is automated, you don’t get to relax, you get to be unemployed.
Who benefits?
In the past, the financial benefits of automation have concentrated in the hands of those who created the technology, not in the hands of those who were displaced by it. We should not make the same mistake with AI, or we, as technologists, might find ourselves on the pointy end of an angry pitchfork wielding mob. This post is a call for tech-companies to drive social innovation along side technological innovation.
Simon’s claim was neither the first nor the last: more great minds have made similar claims in nearly every decade since. (Here is a more recent study that rates each job with a probability that it will be automated in the next 20 years.)
But before we write off Simon’s claim as laughably optimistic, let’s remember that that the kind of work people did in the 60s was different from the work people do now. For example, as agricultural jobs disappeared new jobs arose in the manufacturing and service sectors- a shift largely driven by technology. So in a sense Simon was right: many jobs of the the 1960s are now done by machines.
What happened ?
People still work today, just in different jobs (Scrum Master, Vlogger?). Bear in mind though that this kind of transition happens over the course of decades. And therein lies the rub; we forget the disruptive effect that can immediately follow rapid technological development. Take ATMs as an example.
This example is often quoted by proponents of automation because, after the wide-spread introduction of ATMs in the 90s, the number of human teller jobs actually went up. However, this is an observation about populations, not individuals.
Let’s look a little closer: In 1997 there is a rapid decline in the number of teller jobs, and this number does not reach its 1997 peak again until 2004, 7 years later. For an individual a 7 year employment gap can be devastating, forcing them to change jobs, which is costly and may still lead to lower paid work.
While those new teller jobs may have the same job-title they are not the same job. As James Bessen, a lecturer in Law at Boston University, notes “[banks today] are hiring more college graduates as bank tellers. And […] the nature of occupations is getting up-skilled in some fashion”. This is good news for college graduates, however, it also indicates that the individuals being hired for those new teller jobs in 2004 are not the same individuals that lost their jobs in the late 90s.
This observation reveals the real story of automation. At a population scale (i.e. for society as a whole) it is beneficial: more jobs and economic growth. However, on an individual level, it can have devastating life consequences like unemployment, and lower wages. In short, the financial benefits of automation are reaped by those who developed the technology and wages stagnate for everyone else. If the pace of technological development continues to accelerate the impact on individuals will as well.
How do we fix this?
Some creative solutions have been proposed, like taxing robots. But I would like to propose something more fundamental.
First, we should change the way we define a ‘job’. If rapid change through innovation (i.e. “disruption”) is the new normal, we can no longer think of a job as single employee-employer relationship. It should become something more fluid, defined by your skills and the tasks you can perform rather than by your location, job-title, or employer. Consider that the needs of a company continually change (through technological innovation) and your skills will not always match their needs. So we need to lose the stigma associated with part-time and short term jobs (I’m looking at you, corporate America).
This idea seems at odds with he current trend of longer job tenure, but as noted by Anthony Carnevale, director of the Georgetown University Center on Education and the Workforce:
This situation is harming workers who are forced to hang on to a job for too long even if it does not fit their skills or ambitions. This brings me my to my second proposal:
Universal basic income
A social safety net should enable people to upskill at times that they deem best. The current system pushes people to stay in a job for the sole purpose of having an income until that job is made obsolete and forces a costly retraining. A more flexible social system, such as that afforded by a universal basic income, would enable people to move on between projects more flexibly and enabling more continuous growth. While others may view UBI as a tool for social justice, I see it as an enabler of greater (societal) productivity.
Lifelong Learning
Of course, the missing piece here is the mechanism for continuous skill development. The traditional 4-year degree does not fit the fluid job model of the future. I may expand on this in a future post, for now, I’ll say that Lifelong Learning will be the new norm, and employers had better catch up.
These measures are not silver bullets. They will not solve every problem. But they are a start. Most importantly we, as technology builders, need to be more proactive in advocating for social change to go hand-in-hand with technological change. If we don’t we may pay a high price. Opportunistic politicians that now blame immigration, renewable energy, or global warming for stagnating wages may turn AI into the next bogeyman. If we, as a community, don’t ensure everyone feels the benefits of the AI revolution, we might find ourselves a target, and justly so.
If you agree that we all need to drive innovation responsibility, please share this post with your friends or join the discussion below.
Acknowledgement: Thanks to Ruta Danyte and Tara Lonij for the interesting discussions to help me shape my thoughts.
