In a recent research study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), the burning questions surrounding the potential automation of human jobs take center stage: Will AI automate jobs, and if so, which jobs and when? This study delves into these queries to provide a nuanced perspective on the impact of AI technologies.

Numerous attempts have been made to forecast the influence of today’s AI technologies, such as large language models, on people’s livelihoods and entire economies. Estimates from Goldman Sachs suggest a potential automation of 25% of the labor market in the coming years, while McKinsey predicts nearly half of all work to be AI-driven by 2055. A survey involving the University of Pennsylvania, NYU, and Princeton indicates that ChatGPT alone could impact approximately 80% of jobs. Additionally, reports from Challenger, Gray & Christmas indicate ongoing replacement of workers by AI.

Contrary to expectations, the MIT researchers reveal in their study that the majority of jobs previously identified as susceptible to AI displacement are not currently “economically beneficial” to automate. Neil Thompson, a research scientist at MIT CSAIL and co-author of the study, emphasizes that the impending AI disruption might occur more gradually and less dramatically than some predictions suggest.

It is important to note that the study exclusively focuses on jobs requiring visual analysis, neglecting the potential impact of text- and image-generating models like ChatGPT and Midjourney. The researchers leave this aspect to be explored in subsequent studies.

The study involved surveying workers to understand the tasks an AI system would need to replace their jobs fully. The researchers then modeled the cost of building such a system and assessed whether businesses, particularly non-farm U.S.-based ones, would be willing to bear the upfront and operating expenses.

An illustrative example in the study revolves around a baker. Despite a potential 6% time savings in automating food quality checks, the study estimates that only 23% of wages for vision tasks would be economically attractive to automate with AI. The researchers argue that humans remain the better economic choice for these job components.

The study acknowledges its limitations, including the exclusion of cases where AI augments rather than replaces human labor and the creation of new tasks and jobs. Despite these limitations, the researchers assert that their motivation was not influenced by the study’s backer, the MIT-IBM Watson AI Lab.

The study concludes by emphasizing the need for policymakers to prepare for AI job automation and suggests that this process will unfold over years or even decades, allowing time for the implementation of policy initiatives. For AI researchers and developers, the study underscores the importance of reducing the costs of AI deployments and expanding their scope to make AI economically attractive for firms seeking automation solutions.

By Impact Lab