Why AI Will NOT Take All Our Jobs

Why AI Will NOT Take All Our Jobs
by
Steven Shwartz

Over the last 200 years, technology automation has never caused a net loss of jobs. AI is a type 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, but robots also 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 somehow manage to develop robots with human-level intelligence, every job is in danger. Robots will 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.

As I explain in my book [1], the good news is that 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 can’t 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 [2].  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. The biggest impact might occur if autonomous vehicle technology matures to the point where it can have a widespread rollout. Such a rollout would impact several job categories, including truck drivers, taxi and bus drivers, as well as Uber and Lyft drivers. However, as I explained in my book, because we cannot build thinking and reasoning capabilities into autonomous vehicles, this is likely to remain an unrealized concern for the next twenty to forty years, giving existing workers time to retire and new workers the opportunity to choose other career paths.

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 [4].

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 [5]. 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.

MGI compared this level of job loss with 200 years of job declines in the US and found that the level projected is very much in line with previous trends. Another question studied by MGI is how many jobs the economy will create. They noted that in every ten-year period in US history, eight to nine percent of current jobs did not exist in the prior ten year period. This means that, over the fifteen-year period in which ten percent of jobs are lost, twelve to fourteen percent will be created. So, there will be a net gain in job creation.

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 [6]. 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 [7].

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 [8]. 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 do nothing but 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.

As I explained in my book, none of these systems can be considered evidence of progress toward human-level AI systems. AI is producing very useful systems that can do some specific tasks that formerly required a human.  However, the techniques used to develop these systems cannot produce human-level thinking and reasoning. As a result, the impact on jobs will be similar to earlier types of technology automation.

Jobs Taken by Conventional Software Should Not Be Attributed to AI

Unfortunately, it seems to be in vogue these days to blame all automation job losses on AI. For example, 2020 US Presidential candidate Bernie Sanders, when asked if technology companies should be responsible for the jobs they eliminate, responded by saying: “I will tell corporate America that AI and robotics are not going to be used just to throw workers out on the street.”[9]

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.  Some 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 [10]. 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 similar to that of 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.

References
[1] Shwartz, S.  Evil Robots, Killer Computers, and Other Myths:  The Truth About AI and the Future of Humanity.  Fast Company Press.  Expected publication date:  February, 2021.
[2] https://hai.stanford.edu/sites/default/files/ai_index_2019_report.pdf
[3] https://sea.pcmag.com/feature/18333/ais-implications-for-productivity-wages-and-employment
[4] https://sea.pcmag.com/feature/18333/ais-implications-for-productivity-wages-and-employment
[5] https://www.mckinsey.com/featured-insights/future-of-work/jobs-lost-jobs-gained-what-the-future-of-work-will-mean-for-jobs-skills-and-wages
[6] https://www.gartner.com/en/newsroom/press-releases/2019-22-10-gartner-unveils-top-predictions-for-it-organizations-and-users-in-2020-and-beyond
[7] James Manyika, Susan Lund, Michael Chui, Jacques Bughin, Jonathan Woetzel, Parul Batra, Ryan Ko, & Saurabh Sanghvi. (2017). Jobs lost, jobs gained: What the future of work will mean for jobs, skills, and wages. In McKinsey & Company. https://www.mckinsey.com/featured-insights/future-of-work/jobs-lost-jobs-gained-what-the-future-of-work-will-mean-for-jobs-skills-and-wages
[8] Brynjolfsson, E., & McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W. W. Norton & Company.
[9] https://www.vox.com/policy-and-politics/2019/12/3/20965464/2020-presidential-candidates-jobs-automation-ai
[10] https://www.wsj.com/articles/the-robot-tax-debate-heats-up-11578495608#:~:text=Bill%20Gates%20has%20called%20for,his%20short%2Dlived%20presidential%20campaign.&text=A%20robot%20tax%20could%20serve,taxes%20paid%20by%20human%20workers.