1 Müller, Vincent C. and Bostrom, Nick, ‘Future progress in artificial intelligence: A Survey of Expert Opinion, in Vincent C. Müller (ed.), Fundamental Issues of Artificial Intelligence (Synthese Library; Berlin: Springer), 2014.
During the 21st century, we will likely develop the capability to automate essentially all human skills. This new technological capability will profoundly change the structure of our economy and society by ending the need to work. With almost no human input, computers and robots will be able to produce any good or service that can be produced. When this happens, two challenges that are central to retirement today—providing income and meaningful activities without paying jobs—will become central challenges for the entire society. Retirement today is our laboratory for designing the post-work society of the future.
We do not yet know when computers and robots will be able to reproduce all human skills. It could be a long time (though a poll of computer scientists suggests 2050-2075 as the most likely period 1). However, the big effects will begin long before we can automate all human skills. This is because the most difficult human skills are not difficult just for computers—they are also often difficult for many people. Most people have mid-level skills and work in jobs that require mid-level skills (Figure 1). This is true particularly for older adults. Even though our societies make large investments in education and training, we have been unable to develop high-level skills in the entire adult workforce. As a result, the big transformation will not wait until we can automate all human skills but will instead start when we can automate mid-level skills.
Figure 1: Literacy Levels by Age in OECD Countries
Source: OECD Skills Outlook 2013: First Results from the Survey of Adult Skills (PIAAC), Table A3.3 (L) (online)
The transformation towards automation will change the discussion about retirement. When we begin to displace many workers in jobs using mid-level skills, we will start to have serious challenges with unemployed workers throughout the population. In this environment, it will become difficult to expect people of traditional retirement age to continue to work as one of the ways of providing income for a longer lifespan. As a result, today’s discussion about the challenge of supporting a growing number of retirees will be transformed into a discussion about supporting an even larger number of non-workers of all ages. This discussion will have to focus on possible solutions related to redistributing income because solutions related to continued paid work will no longer be viable.
During the transition to automation, it will make a difference which skills are automated first. Jobs often involve a mix of physical and cognitive skills (Figure 2), but there is an important difference between the two: most adults have strong mid-level physical skills without much special training, but their mid-level cognitive skills are the result of many years of development in formal education. If computers and robots are able to automate the mid-level physical skills first, there will continue to be strong demand for the mid-level cognitive skills developed during formal education (until those cognitive skills are also automated). However, if instead the mid-level cognitive skills are automated first, then the jobs of many workers who now use mid-level cognitive skills will change to focus primarily on physical skills and many workers will then be over educated for their jobs.
Figure 2: Distribution of Current U.S. Employment by Cognitive and Physical Skills
Source: Elliott, 2014, Anticipating a Luddite Revival, Issues in Science and Technology
We need to know how close computers and robots are to automating mid-level skills to understand when and how this transformation will take place. To do this, we can adapt the tests we now use to assess people’s skills so that the tests can also assess the skills that computers and robots have. If we evaluate the skills of computers and robots that are just now being developed by researchers, we can identify the automation capabilities that will be applied in the economy over the next decade or two. This evaluation will indicate how close we are to automating all mid-level skills.
The OECD is now conducting an exploratory study to measure computer capabilities using an existing test of adult skills. This exploration will address only some of the skills used at work. The approach could be expanded in the future to look at the full set of work skills.
Hear from Stuart Elliot about how societies will need to anticipate and address the looming challenge posed by the automation of physical and cognitive skills.
Stuart W. Elliott
Directorate for Education and Skills, Organisation for Economic Co-operation and Development (OECD)
Stuart W. Elliott PhD is an analyst in the Directorate of Education and Skills at the Organisation for Economic Co-operation and Development. He recently completed work on a thematic report on the test of problem solving with computers that is part of the OECD’s Survey of Adult Skills (PIAAC). He is currently using PIAAC to look at the relation between technology, skills and productivity, including a comparison of the capabilities of humans and computers on the PIAAC test.
Before coming to the OECD, Elliott directed the Board on Testing and Assessment of the National Research Council in the US, leading numerous studies on educational tests and indicators, assessment of science and 21st century skills, applications of information technology, and occupational preparation and certification.
An interesting perspective on why it's never too early to start saving for your golden years.
If you’re 55 to 74 years old, CoRI Indices can be used to get an instant estimate of the retirement income your current savings...
Stay abreast of all the latest news in the financial markets with timely analysis from BlackRock's investment strategists.