AI's Challenge to Historical Job Creation Patterns in Automation Waves
The historical pattern of automation leading to job creation may be challenged by artificial intelligence. Research conducted by MIT economists Daron Acemoglu and Pascual Restrepo in 2019 revealed that from 1980 to 2015, half of the job growth occurred through changes in job titles or tasks. For instance, agricultural jobs in the U.S. decreased from 41% in 1900 to 2% by 2000, yet displaced workers transitioned to new positions. Similarly, the introduction of electrification and ATMs resulted in new employment opportunities despite some job losses. Nevertheless, a 2024 study by MIT's David Autor indicates that computerization has diminished middle-skill jobs. Acemoglu and Simon Johnson's 2023 research cautions against 'so-so automation' that displaces without creating new jobs, while Autor posits that AI could either devalue or enhance human skills.
Key facts
- AI challenges historical pattern where automation created more jobs than destroyed
- Agricultural workforce dropped from 41% in 1900 to 2% by 2000 in U.S.
- MIT economists Daron Acemoglu and Pascual Restrepo documented displacement-reinstatement cycle in 2019 paper
- Half of 1980-2015 employment growth occurred in occupations with changed job titles/tasks
- Electrification share in American manufacturing grew from 10% to 80% between 1900-1930
- ATM implementation reduced tellers per branch from 20 to 13 between 1988-2004 but increased branches 43%
- Nearly half of workforce used computer keyboards at work by 1993
- Acemoglu and Simon Johnson coined 'so-so automation' term in 2023 IMF analysis
Entities
Institutions
- MIT
- Journal of Economic Perspectives
- White House Council of Economic Advisers
- Federal Reserve Bank of Philadelphia
- USDA Economic Research Service
- Centre for Economic Policy Research
- Boston University
- IMF
- IBM
- NBER
Locations
- United States
- Sweden
Sources
- Quartz —