Why AI Will NOT Take All Our Jobs
Steven Shwartz
Over the last 200 years, there have been many types of technology automation from the cotton
gin to the personal computer. Each new technology has created more jobs than it has
destroyed.
AI is one of the latest types of technology automation. The question for this article is whether
AI automation will be the first technology to result in a net loss of jobs that is bigger than the
decline in the number of workers. Is AI just another form of automation that will create some
jobs and destroy others? Or will AI destroy more jobs than it creates resulting in unemployment
that will continually rise, and remain permanently high?
Unwarranted Fear of Intelligent Robots
Science fiction books, TV, and movies are replete with descriptions of intelligent robots ranging
from beneficial robots like C3PO from the Star Wars movie series to killer robots such as the
ones depicted in the Terminator movie series.
Perhaps as a result, people fear AI more than other types of automation because they envision
humanlike robots replacing people in many kinds of jobs. Not only are they able to work 24/7,
robots don’t get sick, don’t need vacations, and they don’t join unions. Most importantly, they
don’t even get paid!
This fear is enhanced by the many fantastic successes of AI over the last decade and their
emergence into our daily. We use AI to translate web pages, command our smartphones, and
identify the people in the pictures we take. It’s natural to assume that intelligent robots like
those depicted in science fiction will be here soon.
If we were to somehow manage to develop robots with human-level intelligence, every job
would be in danger. Robots would be able to read every book and user manual, take every
training course, and pretty much learn every job. Robots could even go to medical school and
become doctors.
We will probably never see AI systems with this level of intelligence, and therefore we don’t
have to worry about this level of job destruction. The fantastic AI systems that impact our daily
lives are based on technology that cannot be extended to a C3PO level of intelligence.
AI Will Create Some Jobs
There will be several new job categories created because of AI. For example, the research and
advisory firm Cognilytica forecasted in 2019 that new spending on services that label data for
supervised learning would reach one billion dollars by 2023.
You can find many job creation statistics related to AI in the 2019 AI Index report put out by
Stanford University. In the US, the share of AI-related jobs grew from 0.3% of all jobs in 2012 to
0.8% of all jobs. Investment in AI startups increased from $1.3 billion in 2010 to $40.4 billion in
2019. Globally, autonomous vehicles received the most funding in 2019, with $7.7 billion
invested, followed by medical research, facial recognition, video content analysis, and fraud
detection.
AI Will Destroy Some Jobs
Some categories of jobs are certainly at risk from AI technology. Most tasks that can be
characterized as classification tasks can be learned by a machine. Tasks requiring visual
classification are especially vulnerable. Examples are spotting terrorists in airports, reading
MRIs, and sorting parts in a factory.
Additionally, customer service jobs that involve following a script are vulnerable because of
advances in speech recognition.
And it’s possible that self-driving cars and trucks might someday take taxi and trucking jobs.
Most Experts Believe AI Will Create As Many Jobs As It Destroys
Robert Gordon, Professor of Social Sciences at Northwestern University, said that, in the 250
years since the first industrial revolution, there has never been a technology that caused mass
unemployment, and there is no reason to believe that AI will change this pattern.
According to James Manyika, the Chairperson of the McKinsey Global Institute (MGI), MGI has
produced several reports on the impact of AI and automation in general on work and wages.
They analyzed the eighteen types of cognitive, sensory, and motor skills needed for over 2000
job descriptions and 800 occupations. They analyzed which skills are most susceptible to AI
automation and their reports project which jobs will disappear over the next fifteen years. They
found that fifty percent of the activity types will be automatable, which is a scary number.
However, that does not translate into a fifty percent job loss, because only ten percent of
occupations are composed of more than ninety percent automatable tasks.
In 2017, IT industry analyst Gartner Group predicted that by 2020 AI would have created 2.3
million jobs while eliminating only 1.8 million jobs. Rather than replacing humans entirely,
Gartner argues, AI will augment existing employment and thereby improve productivity. Along
the same lines, a 2018 study by consulting firm McKinsey polled 20,000 AI-aware C-level
executives in 3,000 large companies in ten countries. They said that, for the most part, their
primary goal in using AI was not to reduce the number of workers. Instead, it was to drive
initiatives that would be more likely to increase the number of available jobs.
Some Argue that AI Will Cause Rampant Unemployment
Some analysts disagree that AI is just another form of automation. MIT researcher Erik
Brynjolfsson and Andrew McAfee, in their book The Second Machine Age, argue that the job
loss due to AI will be far greater than the job loss seen in the past due to other revolutionary
forms of automation. They point to advances in language capabilities in systems like Siri, Google
Translate, and the Watson DeepQA computer that beat the Jeopardy! champions. Based on the
fast-moving progress to date, they assume that computers will continue to get smarter at a
rapid pace and take over more and more jobs.
Siri, Google Translate, and DeepQA are all fantastic achievements. However, the reality is that
these systems have no human-like intelligence. Personal assistants are great productivity
enhancers but they rely on workers who code canned responses to questions and
commands. One can ask the questions in many different ways but that is because workers
hand-code patterns for all the different ways they can think of to ask a question or issue a
command. The fact that personal assistants are useful does not mean that computers are
thinking and reasoning like people. Similarly, beating the Jeopardy! champions was a fantastic
accomplishment. However, the Watson DeepQA system was a collection of clever tricks with
no real language understanding capabilities. And Google Translate does a great job of
translating language but doesn't understand anything about what it translates.
None of these systems can be considered evidence of progress toward human-level AI systems.
AI is producing extremely useful systems that can do some specific tasks that formerly required
a human. However, the techniques used to develop these systems cannot produce humanlevel thinking and reasoning. As a result, AI will have far less impact on employment than
people expect.
Jobs Taken By Conventional Software Should Not Be Attributed to AI
The biggest technology driver of job loss today is not AI. Conventional software that uses
explicit coding of instructions and rules destroys far more jobs than AI systems. Word
processors have replaced many secretaries, tax preparation software has reduced the need for
accountants, automated toll booths have replaced human toll collectors, internet travel sites
have displaced many travel agents, Uber and Lyft are displacing taxi drivers, e-commerce
(especially Amazon) is taking a toll on brick and mortar retail, and self-checkout technology is
threatening the 3.6 million US cashier jobs. And those are just a few examples.
These conventional software technologies have all been responsible for gradual job loss.
However, they have not impacted the low US unemployment rate of the five- to ten-year
period before the onset of the COVID-19 virus. AI will continue to marginally enhance these
technologies, but there is no reason to believe it will change the long-term trend of no net job
loss due to automation.
How to Train Workers Is the Key Issue
Jobs will be lost, and jobs will be gained. People have been debating for centuries whether
more jobs are created or destroyed by technology. However, no one disagrees that job
displacement creates significant human suffering, and few disagree with the need for job
retraining programs.
To fund retraining, Bill Gates has called for a tax to be applied when a robot replaces a human
worker with the proceeds used to fund retraining activities. I am not a politician or a
policymaker, but I strongly believe that our society and government need to address the issue
of displaced workers with job training and a safety net.
Summary
People worry that AI will take all their jobs. If human-level AI were possible, this would be a
legitimate concern because AI systems would threaten nearly every job. However, we do not
need to worry about human-level AI systems taking jobs because it is highly unlikely that we
will develop human-level AI systems.
AI systems will take some jobs and will create other jobs. Most likely, the impact on jobs will be
less significant than earlier technologies such as computers, the internet, and e-commerce. Still,
loss of a job can be a devastating life experience, so we need to find a way to retrain people
who have lost jobs to technology.