As AI takes on more tasks, the gap in employment and wages across different professions may widen. Anthropic released an analysis report on September 9th, assessing the impact of AI development on employment, wages, and unemployment rates in the US by 2030. The analysis, based on US occupational task data, examines how much of human work AI can take on and calculates changes in employment and wages according to AI performance and adoption levels.
The report shows that the impact varies significantly across different professions. Occupations like developers and call center consultants, which have more tasks that AI can replace, are seeing a decrease in job roles. In contrast, professions like electricians and nurses, which involve more tasks requiring direct human involvement, are experiencing relatively smaller changes. Within the same profession, the extent to which different tasks are affected by AI also varies. For example, tasks such as recording a patient's blood pressure, pulse, or ordering ward supplies are classified as AI-replaceable work for nurses, while tasks like washing patients and providing direct care to patients are still performed by humans.
AI Development Will Lead to Changes in Occupational Structure
The number of people changing jobs is also increasing. In professions heavily affected by AI, displaced workers need to learn new skills and enter other professions, a process that takes time and consequently drives up unemployment rates.
In terms of wage changes, wages for knowledge work are declining while those for non-knowledge work are rising. The report indicates that when AI has the capability to perform half of the knowledge work, the wages of knowledge workers almost stop increasing. When AI outperforms humans in most knowledge work, the wages of knowledge workers in 2030 will be more than 10% lower than they would be without AI. Conversely, wages for non-knowledge occupations are seeing an increase. For example, when AI accelerates infrastructure design and construction approval processes, construction business grows, leading to increased demand and wages for construction workers.
AI's Impact Will Promote Economic Growth
The report shows that even with a reduction in employment in some occupations, the overall US economy is still growing. If AI develops slowly, GDP in 2030 will be 1.6% higher than it would be without AI; if AI develops more rapidly and is widely adopted, the GDP increase will be 8.3%; if AI surpasses humans in most knowledge work, the GDP increase will reach 32.4%. The more work AI takes on, the higher the productivity and the faster the economic growth. However, this economic growth does not bring proportional returns to workers. Currently, labor income accounts for about 60% of total income, but if AI takes on most knowledge work, this proportion will drop to 45.2%; the share of capital gains (corporate profits and investment returns) will rise from about 40% to 54.8%.
Anthropic utilized occupational task data from the US Department of Labor to break down professions into multiple tasks, distinguishing between work assisted by AI, work fully replaceable by AI, work that humans continue to perform, and new work created after the introduction of AI. For example, nurses reviewing AI triage content or reviewing AI-proposed treatment plans are classified as new work created after the introduction of AI. Additionally, in August, Anthropic conducted a survey on the future of AI with 10,980 Americans, asking about what work AI will be able to do in the future and the actual level of adoption, among other things.
After applying the respondents' expected future applications of AI to economic models, the US GDP in 2030 is projected to be about 10% higher than it would be without AI, with an overall unemployment rate of about 5%. About 10% of respondents expect AI to outperform humans in most knowledge work.