In recent years it has finally dawned on the general public that their jobs may end up taken over by robots. They say “robots”, but what they really mean is Artificial Intelligence– the taking of jobs by mechanical robots has already run its course in manufacturing; the next wave will be not about robots but about artificial intelligence software, which may also be embedded in robots but will primarily be implemented in a variety of non-humanoid computer systems.
The scope of this employment revolution is not fully understood by the public, perhaps because of the difficulty we humans have in grasping exponential changes. Driven by Moore’s law, computer power continues to double every couple of years or so, allowing computer scientists to move the leading edge of AI capabilities ever faster. We already see numerous jobs being replaced by AI, but this is nothing to what’s coming down the road in ten years, and that will be nothing to what we’ll see in twenty… and so on (or maybe not, if the “technological singularity” scenario comes about and breaks everything).
And because people have difficulty grasping this exponential trend, they find comfort in a model where a large class of jobs will be taken over by AI, and a whole other class of jobs will not. They feel that the incursion of AI is bounded, that the computers will only go thus far and no further, that the problem will therefore run into an impenetrable wall and stop. They also feel they know where the wall is, saying things like “the jobs that will succumb to automation are those involving repetitive tasks; but jobs requiring strategic thinking, creativity, and human interaction are safe”. The former class includes lower-level white collar jobs, like bank clerks and insurance underwriters and pharmacists, where the work follows predefined patterns. The latter class is thought to contain the jobs of healthcare professionals, teachers, therapists, artists, politicians and creative writers, among others. After all, they reason, how could a “cold, unfeeling machine” ever grasp and apply human emotions and empathy? Can a computer fall in love?
But they are wrong. I’ll grant you computers may never fall in love (though the jury’s out on that one too), but the insanely accelerating progress in Machine Learning, NLP (Natural Language Processing), image analysis, pattern recognition, and more, is cracking one “what a computer will never do” after another. It is a safe bet that if a human brain can do it, a computer will do it sometime in the present century. We already have computers driving cars in chaotic traffic and composing symphonies that people can enjoy without realizing who wrote them. There is no reason to expect AI systems won’t be able to replace poets, psychiatrists, surgeons and airline pilots before long. If we want something more challenging, we should begin to come to terms with the possibility of AI-based CEOs, supreme court judges and government officials… and, perhaps most significant, AI based software developers capable of developing AI systems better than themselves – the key to a runaway intelligence explosion.
Which raises the question, what is one to do?
If you’re a young person, you should take a hard look at making career choices. Some jobs will survive longer than others; in some fields there will even be new jobs created by the move to AI. For at least a while the AI systems will need to be programmed, optimized, and debugged by humans; and – also for a limited time – robots will need to be designed and repaired by us.
So, which jobs are AI-proof? In the long run, nobody knows; but for the next 20 years, there are two clusters that seem safer than most.
The first is technological jobs, and especially those supporting our “new robot overlords”. Studying STEM subjects in college is indicated; a good and fascinating career choice would then be AI R&D, which is a hot sub-field of computer science. A bright student of NLP or ML can expect to be snatched by the Googles and Facebooks of the world upon graduation.
The second is trades and professions requiring manipulation of disorganized systems in this messy world. Plumbers and electricians are likely to stay around for many years; so are physicians, surgeons, and their assistants – consider how difficult it will be to fully automate dentistry…
And whatever your age, you should drop the complacency and become flexible. The notion that “a machine will never be able to do X” is dangerous, because it is a pretty good bet that a machine will in fact do X, whatever X may be. You should say “when a machine does take over my job – and it will – I must have an alternative skill set to offer”. And because there is little doubt that the AI revolution will create many new jobs (beyond just repairing robots), a flexible person with a versatile skill set could end up employed on the leading edge of a new and fascinating revolution!