by Jonathan Kolber
Our new expert, Jonathan Kolber, wasted no time submitting this incisive article on the way artificial intelligence (AI) is eliminating jobs. Jonathan is the author of A Celebration Society.
You’re at your place of work. Your boss comes up to you, accompanied by a robot. She says, “Say hello to the Mark 24. It’s going to observe you do your job for 6 months. At the end of that time, it will replace you and you’ll be laid off. There’s a severance package if you cooperate.”
This isn’t fantasy. Accelerating automation threatens to displace multitudes of workers. It could wipe away whole lines of work in the next decade or so—possibly yours.
Three areas of technology are driving automation: artificial intelligence, robots and sensors. Most people don’t realize that computer power is doubling every 18 months, and each time it does new capabilities show up.
We already have artificial intelligence that solves many specific problems as well as people do. We talk to phones and other appliances, AIs drive vehicles and manage billions of dollars, AIs are even being used in surgery along with robots. They write magazine articles and have created patentable inventions.
Advances in sensors are giving robots sight, touch, and hearing that match and surpass human capabilities. They’re already stronger, healthier, and work 24/7 without vacations. They’re also cheaper and cheaper to employ.
Robots are now learning new skills the same way people learn—by observation, trial, and error. Once that’s perfected, we’ll just need to train one robot worker and copy it to replace that kind of human worker.
Many AI experts are expecting massive, permanent job losses by 2025. Very soon, self-driving vehicles will pull up and park, and a robot will deliver packages to your door. Humans will no longer drive for a living. This foreshadows the end of 2 million US driving jobs, to cite just one example.
Security guards are now being replaced. So are warehouse workers. Very soon, all manner of people who do manual labor may watch their jobs disappear.
Knowledge workers are also at risk. IBM has a supercomputer called Watson. Watson demolished the best human players of Jeopardy. Each of these men had gone undefeated before meeting Watson. They were certifiable geniuses at Jeopardy. That makes Watson a superhuman intelligence, at least at Jeopardy.
Watson is now being repurposed to handle medical diagnosis, to do legal research, and to master other human professions. IBM is signing up hundreds of partners for different applications. Each will tackle something that people now do.
It’s not just IBM. Amelia is a customer service AI. It interacts with customers like a person does. It notices emotional cues. It learns by observing human customer service experts and from its own mistakes. Already, multiple Fortune 1000 companies are using it.
Each time Watson or another AI replaces a specialty within a profession, the remaining professionals must compete harder to keep their careers. Wages will drop as the displaced professionals switch specialties. Automation is the ultimate in outsourcing.
Perhaps you think your profession is unique. But many AI researchers believe that most of the value-added in professions can be reduced to knowledge, rules, probabilities, and experience. If that’s true, then programs like Watson are going to perform more and more parts of your job faster and cheaper, until what remains is unrecognizable.
As you can see, this is a very serious problem. Yet for all the attention it’s receiving in the media, only two solutions are being proposed. The proposed solutions are that either we should pay a guaranteed minimum income to everyone or we should offer continuing education to all. Neither is adequate.
In much of the world, there is no social safety net nor will there be one any time soon. That fact aside, even in advanced economies these are but palliatives.
While a guaranteed income has much economic appeal, this is a classic case of “the devil is in the details.” For example, the societies now seriously exploring this (Holland, Switzerland, Finland) are homogeneous. It will be far harder to implement in more divided nations—especially the United States, where a “guaranteed minimum income (GMI)” will likely be derided as a “gimme.” It is, in my view, at best something to ease the pain while we move toward a more comprehensive, lasting solution.
Continuing education also is problematic. Suppose you went to school to become a website designer or a legal researcher. Now you find that software is replacing those jobs. So you go back to school and learn something else. You do this again…and again.
At what point is your student-loan debt too crushing to go on? We’ll see many people facing these issues soon. Probably people you know and care about.
A guaranteed income and retraining both depend on someone to pay the bill. Who will that be? Big businesses have become skilled at lobbying against new taxes, finding loopholes, and switching their tax residence to low-tax countries whenever they find it useful.
While some are proposing a new international tax regime to collect and pay such costs, that is likely a fantasy. Existing tax treaties are rife with loopholes, and they are far less onerous, complex, and intrusive than such a new regime would be. Likewise, “tax the rich” is a nice mantra that rarely works out in practice. Smaller businesses and those unable to play tax games will bear the brunt, along with the middle and upper middle classes.
The middle and upper middle classes are splitting into a big class of downwardly mobile people and a much smaller class of what Andrew McAfee and Erik Brynjolfsson call “winner-take-all superstars.” The latter will always do well; the former will be harder and harder pressed to pay existing taxes, much less new ones.
In my view, the most pressing question for policymakers right now is: how few working people can be left to support the unemployed before the system falls apart? We cannot afford to learn the answer.
Some argue that a proliferation of new professions is the answer. Yet the new professions tend to be highly technical, and only a small percentage of workers have the necessary competencies. Further, with AIs and robots now gaining the ability to learn through observation, there is no reason to presume that such AIs will in future take longer to get up to speed than it would take an average person. Also, once an AI/robot/sensor system knows a line of work at a human level, that competency can be rapidly duplicated. People will not be able to compete.
Others say that entrepreneurship is the answer. But not everyone has this competency, and today’s entrepreneurs produce far fewer new jobs relative to revenue than before. Also, the successful ones will use automation at least as well as established companies.
To recap, sensor/robot/AI systems are coming for huge numbers of jobs. A guaranteed income is dubious, and universal retraining is like the proverbial carrot in front of the horse.
Something truly new is needed to deal with technological unemployment. One such solution is the basis for my new book, A Celebration Society. We need to start discussing, debating, and testing such solutions in a serious manner before many more years pass, else the social disruption could be dire.
Editor’s Note: TechCast is working on a Critical Issue that presents the knowledge and arguments on this question, and then invites our experts to estimate the proportion of jobs in various fields about 2030. The study, titled “AI and Future Work,” will be out soon, hopefully to help resolve this nagging question.