A research paper published on arXiv in January challenges the timeline of AI impact on the labor market. Data from the US unemployment insurance system, career profiles, and university courses show that employment decline in AI-exposed professions began nearly a year before the November 2022 launch of ChatGPT.
The study indicates that unemployment risk in roles highly exposed to large language models started rising from a low point in early 2022. Furthermore, graduates from the 2021 cohort onwards entered these professions less frequently than their predecessors, with this trend emerging prior to the end of 2022.
Researchers suggest that broader economic and sectoral shifts were already affecting tech-heavy jobs before generative AI became mainstream. The study emphasizes that AI exposure measures the potential for tasks to be affected by language models rather than counting actual job losses directly caused by AI.
Regarding education, the analysis found that graduates whose academic programs had high AI exposure secured higher starting salaries and spent less time searching for their first job after the release of ChatGPT. The authors note this is an association rather than a direct guarantee of career outcomes.
