Visual AI may change the job market slower than expected because it’s too expensive


A recent study by MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) challenges the assumption that AI will quickly replace large numbers of jobs.

The study, which looks at the economics of replacing human labor with AI, suggests a gradual transition. But keep in mind that the researchers only looked at visual tasks, such as final inspection of products for defects or quality control of food.

While 36 percent of non-agricultural jobs in the U.S. have at least one task that could be automated with computer vision, only 8 percent (23 percent of visual tasks) have at least one task that would be economically attractive for companies to automate.

Image: MIT, Thompson et al.

According to the study, such visual AI systems could currently perform tasks that account for 1.6 percent of employee compensation in the U.S. economy (excluding agriculture). This is equivalent to 0.4 percent of the total economy.



Visual automation is not economically attractive enough for many companies, and may remain so for some time

The MIT study points to the high costs associated with fine-tuning AI systems for specific tasks, which could be uneconomical for smaller businesses.

The researchers illustrate this with the example of a bakery, where the high investment and maintenance costs of an AI system for food quality control would exceed the potential savings from automating the task. It is estimated that such quality checks account for about six percent of a bakery’s labor hours.

In an interview with TIME, lead author Neil Thompson emphasized that while AI has the potential to significantly impact the job market, the transition to automation through AI is not so imminent that it will cause immediate chaos. Policymakers have a window of opportunity to take steps such as retraining.

For AI researchers and developers, the study underscores the importance of reducing the cost of deploying AI and expanding the scope of AI to make automation more economically attractive to businesses.

“Much has been written about the future impact of AI on the labor market, primarily using measures of exposure. However, these estimates often rely on the assumption that if a job can be automated, it will be,” says Antonin Bergeaud, Associate Professor of Economics at HEC Paris.


Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top