Goldman Sachs estimates that currently about two-thirds of jobs in the United States are to some extent threatened by AI automation, with generative AI likely to replace a quarter of positions. Globally, generative AI could put 300 million full-time job positions at risk of "automation"
Since the birth of AI technology, discussions about AI taking over human jobs have been endless. Despite the uncertainties surrounding the potential of generative AI, it is undeniable that it has been able to create works comparable to those of humans and break down communication barriers between humans and machines. Goldman Sachs believes that this marks a significant leap in AI technology and may have profound implications for the global macroeconomy.
Goldman Sachs analyst team led by Jan Hatzius pointed out in a research report released last week that if generative AI can deliver on its promised capabilities, the labor market may suffer a significant impact.
Based on Goldman Sachs' research on job data in the United States and Europe, about two-thirds of current jobs in the United States are to some extent threatened by AI automation, with generative AI likely to replace a quarter of positions. Globally, generative AI could put 300 million full-time jobs at risk of "automation."
However, at the same time, Goldman Sachs noted that the "automation" brought by AI can be more than offset by creating new positions, which will largely be the main source of long-term employment growth.
According to Goldman Sachs' estimates, generative AI is expected to increase annual labor productivity growth in the United States by nearly 1.5 percentage points over the next decade, and could ultimately raise the global GDP growth rate by 7%.
While the actual impact of AI will depend on its capabilities and the speed of its adoption, this estimate highlights the enormous economic potential if generative AI can realize its potential.
Are administrative and legal professionals at risk? Can manual laborers breathe a sigh of relief?
Goldman Sachs used task content for over 900 occupations from the ONET database (later expanded to over 2,000 occupations from the European ESCO database) to assess the share of occupations and industries that AI labor-saving automation may involve.
Based on a review of existing research on potential applications of generative AI, Goldman Sachs believes that 13 out of 39 job activities in the ONET database are at risk of AI automation. However, it is worth noting that workers engaged in outdoor work and manual labor will not be automated by AI.
We found that about two-thirds of jobs in the United States are at some risk of AI automation, with a significant portion of tasks (25%-50%) in most positions potentially being replaced by AI.
Weighting the employment shares of various occupations in the Occupational Employment and Wage Survey (OEWS) in the United States and aggregating at the industry level, we estimate that in the United States, AI has the ability to automate approximately a quarter of existing jobs, with higher exposure rates in administrative (46%) and legal (44%) fields, and lower exposure rates in physically intensive fields such as construction (6%) and maintenance (4%). According to Goldman Sachs, globally, 18% of jobs may be automated by AI based on employment weight.
Although the impact of AI on the labor market may be significant, Goldman Sachs points out that most occupations and industries only face partial automation risks, making it more likely for AI to complement rather than replace them.
Goldman Sachs believes that at least 50% of important and complex tasks facing automation may be replaced by AI, while jobs with automation risks ranging from 10% to 49% are more likely to be supplemented by AI. Jobs with an AI exposure rate of 0% to 9% are unlikely to be affected.
Under Goldman Sachs' baseline assumptions, 7% of the current total employment in the United States may be replaced by AI, 63% may be supplemented by AI, and 30% may not be affected.
Substitution or Supplement?
Goldman Sachs believes that there are two main ways in which AI-driven automation can increase global GDP: improving the efficiency of existing jobs and eliminating outdated positions to stimulate the creation of new ones.
Firstly, most occupations may be affected to some extent by AI automation. After adopting AI technology, employees are likely to use at least some of the saved time to enhance productive activities. Academic research shows that employees in companies that adopt AI technology early have generally increased their labor productivity growth by 2-3 percentage points per year after adopting AI. Although it is difficult to directly generalize these results to generative AI, these research results clearly indicate that generative AI can significantly improve productivity economically.
Secondly, workers who are unemployed due to AI automation are expected to eventually re-employ in emerging occupations, thereby increasing overall output. These emerging occupations may directly result from the adoption of AI or from the overall demand growth driven by the productivity improvement of non-unemployed workers. For example, innovations in information technology have brought new occupations such as web designers, software developers, and digital marketing professionals, while also increasing overall income, indirectly boosting the demand for labor in service industries such as healthcare, education, and catering.
Goldman Sachs points out that since the first half of the post-World War II period, technological changes have displaced workers and created new job opportunities at roughly the same rate. However, since the 1980s, the speed of creating new job opportunities has not kept up with the pace of worker displacement.
These results suggest that if the impact of AI on the labor market is similar to early information technology advances, then the direct impact of generative AI on labor demand in the short term may be negative, but its impact on labor productivity growth remains positive Goldman Sachs estimates that about 7% of the workforce faces the risk of complete unemployment, but most of them can find new jobs in slightly lower-output positions. Employees affected by AI automation are expected to experience productivity gains consistent with existing forecasts, with this impact expected to be evident in about 50% of companies adopting generative AI technology after 10 years.
Based on these assumptions, Goldman Sachs predicts that widespread adoption of generative AI could increase the overall labor productivity growth rate by about 1.5 percentage points annually, similar to the recent average growth rate of 1.5% per year, comparable to the impact of transformative technologies such as electric motors and personal computers.
In the United States, the productivity growth boost from generative AI could easily range from 0.3 to 3.0 percentage points. This range depends on the difficulty of tasks, the prevalence of automation, and the speed of adoption.
For example, in scenarios where AI capabilities are weaker, the predicted annual labor productivity growth rate will decrease to 0.3 percentage points. If AI is more powerful, the growth rate will increase to 2.9 percentage points per year.
Furthermore, assuming no labor is replaced, productivity growth is expected to decrease to 1.2 percentage points per year, while if a significant proportion of labor is replaced, productivity growth will increase to 2.4 percentage points per year.
In conclusion, Goldman Sachs points out that despite uncertainties, the development and application of generative AI are expected to significantly drive the global economy in most cases