哪些经济任务正在由AI完成？来自数百万次Claude 对话的证据

Kunal Handa∗, Alex Tamkin∗, Miles McCain, Saffron Huang, Esin Durmus Sarah Heck, Jared Mueller, Jerry Hong, Stuart Ritchie, Tim Belonax, Kevin K. Troy Dario Amodei, Jared Kaplan, Jack Clark, Deep Ganguli Anthropic Abstract Despite widespread speculation about artificial intelligence’s impact on the future ofwork, we lack systematic empirical evidence about how these systems are actuallybeing used for different tasks. Here, we present a novel framework for measuringAI usage patterns across the economy. We leverage a recent privacy-preservingsystem [Tamkin et al., 2024] to analyze over four million Claude.ai conversationsthrough the lens of tasks and occupations in the U.S. Department of Labor’sO*NET Database. Our analysis reveals that AI usage primarily concentrates insoftware development and writing tasks, which together account for nearly half ofall total usage. However, usage of AI extends more broadly across the economy,with∼36%of occupations using AI for at least a quarter of their associatedtasks. We also analyzehowAI is being used for tasks, finding 57% of usagesuggests augmentation of human capabilities (e.g., learning or iterating on anoutput) while 43% suggests automation (e.g., fulfilling a request with minimalhuman involvement). While our data and methods face important limitations andonly paint a picture of AI usage on a single platform, they provide an automated,granular approach for tracking AI’s evolving role in the economy and identifyingleading indicators of future impact as these technologies continue to advance. 1Introduction Rapid advances in artificial intelligence suggest profound implications for the evolution of labormarkets [Brynjolfsson et al., 2018b, Acemoglu, 2021, Trammell and Korinek, 2023, Hering, 2023,Comunale and Manera, 2024, Maslej et al., 2024].Despite the importance of anticipating andpreparing for these changes, we lack systematic empirical evidence about how AI systems are actuallybeing integrated into the economy. Existing methodologies—whether developing predictive models[Webb, 2019, Eloundou et al., 2023, Kinder et al., 2024], conducting controlled studies of productivityeffects [Peng et al., 2023, Noy and Zhang, 2023], or administering periodic surveys of users [Humlumand Vestergaard, 2024, Bick et al., 2024]—cannot track the dynamic relationship between advancingAI capabilities and their direct, real-world use across the economy. Here, we present a novel empirical framework for measuring AI usage across different tasks inthe economy, drawing on privacy-preserving analysis of millions of real-world conversations onClaude.ai [Tamkin et al., 2024]. By mapping these conversations to occupational categories in theU.S. Department of Labor’s O*NET Database, we can identify not just current usage patterns, but also early indicators of which parts of the economy may be most affected as these technologiescontinue to advance.2 We use this framework to make five key contributions: 1.Provide the first large-scale empirical measurement of which tasks are seeing AI useacross the economy (Figure 1, Figure 2, and Figure 3)Our analysis reveals highest use fortasks in software engineering roles (e.g., software engineers, data scientists, bioinformaticstechnicians), professions requiring substantial writing capabilities (e.g., technical writers,copywriters, archivists), and analytical roles (e.g., data scientists). Conversely, tasks inoccupations involving physical manipulation of the environment (e.g., anesthesiologists,construction workers) currently show minimal use.2.Quantify the depth of AI use within occupations (Figure 4)Only∼4%of occupationsexhibit AI usage for at least 75% of their tasks, suggesting the potential for deep task-leveluse in some roles. More broadly,∼36%of occupations show usage in at least 25% of theirtasks, indicating that AI has already begun to diffuse into task portfolios across a substantialportion of the workforce.3.Measure which occupational skills are most represented in human-AI conversations(Figure 5). Cognitive skills like Reading Comprehension, Writing, and Critical Thinkingshow high presence, while physical skills (e.g., Installation, Equipment Maintenance) andmanagerial skills (e.g., Negotiation) show minimal presence—reflecting clear patterns ofhuman complementarity with current AI capabilities. 4.Analyze how wage and barrier to entry correlates with AI usage (Figure 6 and Table 2).We find that AI use peaks in the upper quartile of wages but drops off at both extremes ofthe wage spectrum. Most high-usage occupations clustered in the upper quartile correspondpredominantly to software industry positions, while both very high-wage occupations (e.g.,physicians) and low-wage positions (e.g., restaurant workers) demonstrate relatively lowusage. This pattern likely reflects either limitations in current AI capabilities, the inherentphysical manipulation requirements of these roles, or both. Similar patterns emerge forbarriers to entry, with