Board of Governors of the Federal Reserve System

Figure 1. Measurement of Exposure to Generative AI

This is a flow chart that depicts how the generative artificial intelligence occupational exposure (AIOE) measures from Felten et al. (2023) are linked to the National Survey of College Graduates 2013-2021 biannual data. From left to right, it first depicts “O*NET occupational abilities” that are associated with the AIOE measures with an arrow pointing to the right at “SOC occupation codes.” The arrow demonstrates linkage of these variables. Then, there is an arrow from “SOC occupation codes” pointing to the right at “Graduates’ principal jobs’ occupation codes,” which is provided in the NSCG 2013-2021 data. Finally, there is an arrow from “Graduates’ principal jobs’ occupation codes” to “Field of study or Carnegie classification.”

Return to text

Figure 2. Relationship Between Demographic Characteristics and Exposure to Generative AI

This is a figure with five panels depicting the relationship between demographic characteristics and exposure to generative artificial intelligence. Panel A has two graphs side by side, each depicting a line graph on top of a scatter plot. The Panel A graph to the left plots the image generation artificial intelligence occupational exposure (AIOE) z-score of a college major on the y-axis and percent of white graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line starts at about 0.80 z-score with about 55 percent white and decreases to about -0.50 with about 95 percent white. The Panel A graph to the right plots the language modeling AIOE z-score of a college major on the y-axis and percent of white graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line begins at about 0.40 z-score with about 55 percent white and decreases to about -0.25 with about 95 percent white.

Panel B has two graphs side by side, each depicting a line graph on top of a scatter plot. The Panel B graph to the left plots the image generation AIOE z-score of a college major on the y-axis and percent of black graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line starts at about 0.40 z-score with about 0 percent black and decreases to about -1.25 with about 30 percent black. The Panel B graph to the right plots the language modeling AIOE z-score of a college major on the y-axis and percent of black graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line begins at about -0.10 z-score with about 0 percent black and increases to about 0.50 with about 30 percent black.

Panel C has two graphs side by side, each depicting a line graph on top of a scatter plot. The Panel C graph to the left plots the image generation AIOE z-score of a college major on the y-axis and percent of Asian graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line starts at about -0.50 z-score with about 0 percent Asian and increases to about 1.60 with about 40 percent Asian. The Panel C graph to the right plots the language modeling AIOE z-score of a college major on the y-axis and percent of Asian graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line begins at about 0 z-score with about 0 percent Asian and increases to about 0.10 with about 40 percent Asian.

Panel D has two graphs side by side, each depicting a line graph on top of a scatter plot. The Panel D graph to the left plots the image generation AIOE z-score of a college major on the y-axis and percent of Hispanic graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line starts at about 0 z-score with about 0 percent Hispanic and increases to about 0.25 with about 45 percent Hispanic. The Panel D graph to the right plots the language modeling AIOE z-score of a college major on the y-axis and percent of Hispanic graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line begins at about -0.30 z-score with about 0 percent Hispanic and increases to about 1.5 with about 45 percent Hispanic.

Panel E has two graphs side by side, each depicting a line graph on top of a scatter plot. The Panel E graph to the left plots the image generation AIOE z-score of a college major on the y-axis and percent of female graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line starts at about 1.10 z-score with about 10 percent female and decreases to about -1.50 with about 95 percent female. The Panel E graph to the right plots the language modeling AIOE z-score of a college major on the y-axis and percent of female graduates in a college major on the x-axis. There is also a line graph imposed on top that shows the trend line. The line begins at about -0.10 z-score with about 10 percent female and increases to about 0.10 with about 95 percent female.

Return to text