Large language models and the Danish labour market

Generative artificial intelligence (AI) tools such as large language models are spreading rapidly. The most prominent example is ChatGPT, which gathered more than 100 million active users within two months. This type of generative AI has the potential to change the way people work, creating opportunities for innovation and productivity gains. However, the opportunities and challenges will most likely be unequally distributed across the workforce.

This analysis explores the unequal economic impact of large language models (LLMs) on the Danish Labour Market. The analysis uses the so-called AI Occupational Exposure (AIOE) scores from a study of the American labour market and merges these scores with administrative data from Statistics Denmark. The AIOE scores reflect the relatedness between AI applications and human abilities connected to different occupations. Thus, the scores express potential economic impact of AI applications across occupations through either labour-augmenting or labour-displacing effects.

Main conclusions:

• Occupations dominated by cognitive routine tasks have the highest potential to change through large language models. Legal Professionals is the occupation with the highest LLM score. The occupation with the lowest score is Painters, building structure cleaners & related trades worker.
• Economic activities influenced by cognitive abilities have higher LLM scores than activities dominated by physical tasks. The activity with the highest LLM score is Higher Education. The activity with the lowest score is Building completion and finishing.
• Employed females altogether have more potential to apply large language models than employed males. However, within Human Health & Social Work activities women have a slightly lower LLM score than males.
• Employees with high personal yearly income generally have more potential to use and take advantage of large language models than employees with lower income.

The analysis is available in Danish here: Store sprogmodeller og det danske arbejdsmarked