Do you believe AI and robots taking over jobs is, as Elon Musk recently put it, the “biggest risk that we face as a civilization?” Or mostly overblown fear-mongering? Somewhere in between?
There’s probably a research-backed prediction that supports your view. We’ve ranked them here, in order from ”certain doom” to “possible utopia,” for your convenience.
Doom rating: [protected-iframe id=”105222e93d8609de6eed7410a904d1a2-39587363-113766103″ info=”https://giphy.com/embed/btF5rwlEQMgQE” width=”480″ height=”221″ frameborder=”0″ class=”giphy-embed” allowfullscreen=””]
How they got there: Carl Benedikt Frey and Michael A. Osborne, researchers at Oxford University, asked machine-learning experts to assess whether a sample set of occupations were “automatable” or “not automatable,” which helped to inform how the presence of “engineering bottlenecks” that couldn’t be easily automated, such as creativity and social engineers, factored into whether an occupation could be automated. They created a machine-learning algorithm to estimate a probability of automation across each US occupation. The resulting paper published in September of 2013.
“We find that most workers in transportation and logistics occupations, together with the bulk of office and administrative support workers, and labour in production occupations, are likely to be substituted by computer capital. … As industrial robots are becoming more advanced, with enhanced senses and dexterity, they will be able to perform a wider scope of non-routine manual tasks. From a technological capabilities point of view, the vast remainder of employment in production occupations is thus likely to diminish over the next decades.”
Doom rating: [protected-iframe id=”0d4d1dc89c74ce71812ae23a5efea477-39587363-113766103″ info=”https://giphy.com/embed/l2JIhq7AR3J7Qq9Hy” width=”480″ height=”324″ frameborder=”0″ class=”giphy-embed” allowfullscreen=””]
How they got there: Whereas Frey and Osbourne analyzed entire occupations for the risk of automation, researchers at PWC took into account that some tasks within jobs can be automated, while some can’t, and that different jobs within occupations involve different combinations of tasks. For a March 2017 paper, they created a machine-learning algorithm for identifying automation risk that included data about workers doing the tasks, such as the education and training levels required.
Average pre-tax incomes should rise due to the productivity gains, but these benefits will probably not be evenly spread across income groups. The pay premium for higher education and non-automatable skills will also probably rise ever higher. There is therefore a case for some form of government intervention to ensure that the potential gains from automation are shared more widely across society through policies in areas like education, vocational training and job matching.
Doom rating: [protected-iframe id=”544acb188e9b29cf4d1e559be4d18942-39587363-113766103″ info=”https://giphy.com/embed/yy0D0Q4lOiR56″ width=”480″ height=”360″ frameborder=”0″ class=”giphy-embed” allowfullscreen=””]
How they got there: For an OECD working paper published in May of 2016, Melanie Arntz, Terry Gregory, and Ulrich Zierahn responded to Frey and Osbourne’s approach directly by considering how tasks may vary within occupations. For instance, according to Frey and Osbourne, the profession “bookkeeping, accounting, and auditing clerks” have a 98% potential for automation. But unlike Frey and Osbourne, the OECD working paper took into account that only 24% of them can perform their job without group work or face-to-face interactions, which aren’t as easy to automate as scanning documents. This approach accounted for that variation, resulting in a much lower estimate for the number of jobs at risk (The PWC researchers who came up with the much higher estimate of jobs likely to be automated, as high as 38% in the US, built upon this approach).
The main conclusion from our paper is that automation and digitalization are unlikely to destroy large numbers of jobs. However, low qualified workers are likely to bear the brunt of adjustment costs as the automatibility of their jobs is higher compared to highly qualified workers.
Doom rating: [protected-iframe id=”8965ef6c011578fb8801347471b41e2f-39587363-113766103″ info=”https://giphy.com/embed/3og0IJLDE9vDBhCVLW” width=”480″ height=”332″ frameborder=”0″ class=”giphy-embed” allowfullscreen=””]
How they got there: For an April 2017 report, Forrester researchers divided occupations into three types of tasks: physical, intellectual, and customer service. Then they created cannibalization rates for the percentage of job tasks that will be reduced by automation each year.
Using this methodology, they figured that 17% of current jobs would be lost to automation. For every 15 jobs the researchers predicted would be lost, they assumed that one job would be created “in software, engineering, design, maintenance, support, training.” This ratio was based on an EY survey of entrepreneurs, 81% of whom said that the investment in technology had increased their workforce, on average 13%. PWC adjusted it on the assumption that the Trump administration’s “America First” policies would increase the rate of automation.
Forrester estimated that automating technologies would create about a 10% increase in employment, which, when combined with the 17% decline in employment, would create a net effect of 7% of jobs lost.
“Although many scaremongers promise a future in which automation technologies create massive unemployment, these analyses are based on flawed assumptions. Instead, automation will replace some jobs and create others.”
Doom rating: [protected-iframe id=”4e606fa6d067a8280236bfe9d24678b6-39587363-113766103″ info=”https://giphy.com/embed/l4FGGAXRjAPGBvK1O” width=”480″ height=”266″ frameborder=”0″ class=”giphy-embed” allowfullscreen=””]
How they got there: For a January 2017 report, McKinsey analyzed tasks within occupations, breaking 800 occupations into more than 2,000 activities. Then, using machine learning, it determined which of 18 defined capabilities such as sensory perception and natural language understanding that humans use to complete those activities. It assessed the technical potential for each of those capabilities to understand what percent of each job could be automated.
“People will need to continue working alongside machines to produce the growth in per capita GDP to which countries around the world aspire. Our productivity estimates assume that people displaced by automation will find other employment. The anticipated shift in the activities in the labor force is of a similar order of magnitude as the long-term shift away from agriculture and decreases in manufacturing share of employment in the United States, both of which were accompanied by the creation of new types of work not foreseen at the time.”