Anthropic Study: AI Replacement Potential Far Greater Than Actual Adoption, White-Collar Jobs at Risk of 'Great Attrition'

BlockBeats News, March 7th. The latest research released by Anthropic shows that although AI theoretically can cover the vast majority of tasks in business, finance, law, computer science, and other fields, its actual adoption rate is only a small portion. Taking the Claude model as an example, its theoretical coverage of computer and math tasks is 94%, but the actual usage is only 33%. The research introduces the "Exposure Discrepancy" index to compare theoretical capabilities with real-world usage data.

The results show that the group with the highest AI exposure is not blue-collar workers but highly educated, high-income female white-collar workers: compared to the low-exposure group, this group has a 16 percentage point higher female representation, 47% higher average income, and nearly four times higher proportion of individuals with graduate degrees. Researchers warn that as capabilities improve and adoption deepens, AI could trigger a "Great White-Collar Worker Decline" similar to the scenario where, during the 2007-2009 financial crisis, the unemployment rate doubled from 5% to 10%.

While this has not yet occurred, the risk is evident. The current impact is more reflected in slowed hiring rather than layoffs: in the post-ChatGPT era, the job application rate for exposed professions has decreased by 14% from 2022, and the employment rate of young workers aged 22-25 in related fields has declined by 16%. Some young people choose to pursue further education or temporarily avoid the labor market.