--- title: "Anthropic's report should be interpreted as follows: high theoretical capability, but practical application remains limited" type: "News" locale: "en" url: "https://longbridge.com/en/news/281294500.md" description: "Anthropic's labor market research released on March 5 divides the impact of AI on jobs into \"theoretical capability\" and \"actual exposure,\" reigniting discussions on whether large language models can replace jobs. Although theoretically, LLM systems exceed 80% in task handling capabilities across multiple domains, this data is not based on the actual performance of Anthropic's models but rather on extrapolations from older studies. The research indicates a significant gap between the actual application of AI in the workplace and its theoretical capabilities, especially in the fields of creative and knowledge work" datetime: "2026-04-01T03:20:45.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/281294500.md) - [en](https://longbridge.com/en/news/281294500.md) - [zh-HK](https://longbridge.com/zh-HK/news/281294500.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/281294500.md) | [繁體中文](https://longbridge.com/zh-HK/news/281294500.md) # Anthropic's report should be interpreted as follows: high theoretical capability, but practical application remains limited Anthropic's labor market research on March 5 divides the impact of AI on jobs into two layers: "theoretical capability" and "actual exposure," reigniting discussions about how many positions large language models can actually replace. The blue block represents "theoretical capability," which seemingly suggests that LLM systems can theoretically handle over 80% of tasks in various fields such as arts and media, office administration, law, business finance, and even management, thus interpreting it as AI will significantly invade the white-collar workforce. However, upon examining the foundational data cited by Anthropic, it is not as intense as the outside world perceives. These "theoretical capability" figures do not come from the actual performance of Anthropic's own models, nor are they quantitative forecasts of future performance growth, but rather they are derived from an old study from August 2023 titled "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models." In other words, the blue block represents a considerable speculative academic estimate, focusing on the range in which AI may enhance human productivity, rather than equating to a complete takeover of human jobs. This set of research by Anthropic also highlights the significant gap between what AI can "do" in the workplace and what it is "actually doing." The March labor market study, which combines O\*NET occupational data, real usage patterns of Claude from the Anthropic Economic Index, and existing task feasibility ratings, attempts to depict the actual penetration of AI in job scenarios. The results show that AI usage is highly concentrated on tasks that are theoretically optimal for model processing, but there remains a significant gap from full replacement. **▲ Anthropic has not found a field where AI is more easily accessible that has a significant impact on the job market.** Taking creative and knowledge work as an example, the theoretical exposure rate for Arts & Media positions is as high as 83.7%, but the actual exposure rate is only about 19.2%. Roles related to arts, design, entertainment, and media account for 10.3% of Claude's query volume, second only to computer and mathematics occupations, indicating that creators are not avoiding AI but are increasingly using it as a work tool. Research also found that AI is more commonly used for repetitive, production-oriented tasks, such as data organization, formatting, and basic research; whereas strategy, originality, and directional judgment still rely on humans. Overall, the message that Anthropic's research aims to convey is not that "AI will immediately take away jobs," but rather to remind the public that what is actually happening is a reorganization of workflows and task breakdowns. For the workplace, the key is not just whether AI can perform certain jobs, but where it is currently being applied and which job contents will change first. - How did Anthropic measure AI’s “theoretical capabilities" in the job market? - Anthropic Just Dropped Their Biggest Report Yet. Here’s What It Actually Means for Creatives. （Image source: Anthropic） ### Further Reading: - More frightening than "job stealing" is "seriously talking nonsense," Anthropic's survey of 80,000 reveals the biggest shadow of the AI era. - Anthropic establishes the think tank "Anthropic Institute" to explore the long-term impact of AI on society and the economy. - Anthropic report: The impact of AI on the workforce is still limited, with programmers at the highest risk ### Related Stocks - [Block, Inc. (SQ.US)](https://longbridge.com/en/quote/SQ.US.md) - [iShares Global Tech ETF (IXN.US)](https://longbridge.com/en/quote/IXN.US.md) ## Related News & Research - [Why Leading Restaurants are Switching to Square | XYZ Stock News](https://longbridge.com/en/news/280196623.md) - [Cotality Launches a Universal Connector for AI: Unveiling Its MCP Server and AI-Ready Property Intelligence](https://longbridge.com/en/news/281208250.md) - [Image Sensing Systems, Inc. Expands Agreement with Econolite Control Products, Inc, Strengthens Partnership | AATC Stock News](https://longbridge.com/en/news/281071522.md) - [15:35 ETKubex Named a Leader in GigaOm Radar Reports for Kubernetes Resource Management and Cloud Resource Optimization](https://longbridge.com/en/news/280363384.md) - [SAP SE agreed to acquire Reltio, Inc.](https://longbridge.com/en/news/280838407.md)