Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Potential Impact of AI Innovations on U.S. Occupations

Ali Akbar Septiandri, Marios Constantinides, Daniele Quercia

TL;DR

Introducing the AI Impact measure, this work employs Deep Learning Natural Language Processing to automatically identify AI patents that may impact various occupational tasks at scale and reveals that some occupations will potentially be impacted, and that impact is intricately linked to specific skills.

Abstract

An occupation is comprised of interconnected tasks, and it is these tasks, not occupations themselves, that are affected by AI. To evaluate how tasks may be impacted, previous approaches utilized manual annotations or coarse-grained matching. Leveraging recent advancements in machine learning, we replace coarse-grained matching with more precise deep learning approaches. Introducing the AI Impact (AII) measure, we employ Deep Learning Natural Language Processing to automatically identify AI patents that may impact various occupational tasks at scale. Our methodology relies on a comprehensive dataset of 17,879 task descriptions and quantifies AI's potential impact through analysis of 24,758 AI patents filed with the United States Patent and Trademark Office (USPTO) between 2015 and 2022. Our results reveal that some occupations will potentially be impacted, and that impact is intricately linked to specific skills. These include not only routine tasks (codified as a series of steps), as previously thought, but also non-routine ones (e.g., diagnosing health conditions, programming computers, and tracking flight routes). However, AI's impact on labour is limited by the fact that some of the occupations affected are augmented rather than replaced (e.g., neurologists, software engineers, air traffic controllers), and the sectors affected are experiencing labour shortages (e.g., IT, Healthcare, Transport).

The Potential Impact of AI Innovations on U.S. Occupations

TL;DR

Introducing the AI Impact measure, this work employs Deep Learning Natural Language Processing to automatically identify AI patents that may impact various occupational tasks at scale and reveals that some occupations will potentially be impacted, and that impact is intricately linked to specific skills.

Abstract

An occupation is comprised of interconnected tasks, and it is these tasks, not occupations themselves, that are affected by AI. To evaluate how tasks may be impacted, previous approaches utilized manual annotations or coarse-grained matching. Leveraging recent advancements in machine learning, we replace coarse-grained matching with more precise deep learning approaches. Introducing the AI Impact (AII) measure, we employ Deep Learning Natural Language Processing to automatically identify AI patents that may impact various occupational tasks at scale. Our methodology relies on a comprehensive dataset of 17,879 task descriptions and quantifies AI's potential impact through analysis of 24,758 AI patents filed with the United States Patent and Trademark Office (USPTO) between 2015 and 2022. Our results reveal that some occupations will potentially be impacted, and that impact is intricately linked to specific skills. These include not only routine tasks (codified as a series of steps), as previously thought, but also non-routine ones (e.g., diagnosing health conditions, programming computers, and tracking flight routes). However, AI's impact on labour is limited by the fact that some of the occupations affected are augmented rather than replaced (e.g., neurologists, software engineers, air traffic controllers), and the sectors affected are experiencing labour shortages (e.g., IT, Healthcare, Transport).
Paper Structure (54 sections, 4 equations, 14 figures, 14 tables)

This paper contains 54 sections, 4 equations, 14 figures, 14 tables.

Table of Contents

  1. Introduction
  2. Results
  3. Discussion
  4. Materials and Methods
  5. Acknowledgments
  6. Funding
  7. Author contributions statement
  8. Data Availability
  9. Code Availability
  10. Supplementary Material
  11. Data
  12. Task-Patent Matching
  13. Previous Attempts to Link Tasks to Patents
  14. Most- and Least-Impacted Occupations and Industry Sections
  15. Impact of Robots and Software
  16. ...and 1 more sections

Figures (14)

  • Figure 1: AII score binned by the level of education: high school, associate degrees from community colleges, bachelor's degrees, and master's degrees in science. This binned score was obtained by averaging the scores across all occupations in a given education category, weighted by the total employment for those binned occupations.
  • Figure 2: The number of newly impacted tasks each year for the most affected occupations, combined with the most frequently occurring words in the patents influencing those tasks, are organized around the themes of healthcare, information technology, and manufacturing. These were derived qualitatively and describe the main themes emerging from the patents. Between 2016 and 2018, patents mentioned "patient", "image", "planning", "medical", "device" matched tasks in healthcare. Between 2019 and 2021, patents mentioned "data", "analysis", and "neural networks" matched tasks in information technology. Between 2021 and 2022, patents mentioned "control", "planning", "path", "user", "image", and "neural network" matched tasks in manufacturing.
  • Figure 3: Sector-level AII scores for: (a) all sectors; (b, left) sectors with lowest rate of change from 2015 to 2020; and (b, right) sectors with highest rate of change.
  • Figure 4: Automation vs. augmentation using patent similarity to tasks and micro-titles defined in the Census Alphabetical Index of Occupations and Industries (CAI) autor2022new.
  • Figure 5: Job vacancy rates by sector vs. sector-level AII. The sector of Accommodation and Food Services was positioned more than two standard deviations away from the regression line (i.e., considered as an outlier) and was removed. The original plot is in Supplementary Material.
  • ...and 9 more figures