Surprise: Using an AI to replace employees wouldn’t be cheaper

As in any industrial revolution, the fear of machines beating humans is an obvious fear. And the case of artificial intelligence is no exception to this. But it is one thing for certain jobs or tasks to begin to be done by machines and another for it to be more profitable. In fact, an MIT study has found that In most cases it would be more expensive to replace us with an AI.

The authors, led by Maja Svanberg, have analyzed the profitability of 1,000 “visual inspection” tasks in 800 occupations, such as inspecting food to see if it has spoiled or studying X-ray, ultrasound or ultrasound images. They found that only 23% of workers’ total salaries “would be interesting to automate”, mainly due to the “large initial costs of AI systems“, and that’s if automatable tasks could even be “separated from other parts” of jobs. The results go against the usual fears linked to AI.

“Overall, our findings suggest that AI job displacement will be substantial, but also gradualand, therefore, there is room for policies and retraining to mitigate the impacts of unemployment,” the authors conclude.

In their article, Svanberg’s team focused on tasks assisted by computer vision. One of the examples they mention is that of a person in charge of visually checking the ingredients used in a bakery to “make sure they are of sufficient quality.”

This type of job could theoretically be replaced by a computer vision system, adding a camera and training the system to detect food that has gone bad. However, installing and operating such a system would still be prohibitively expensivesince he would only be responsible for the task of checking the ingredients, which represents only six percent of the employee’s work.

But they also warn that if prices fall, as happened with other technologies that were initially prohibitive (such as mobile phones), “the AI ​​economy may become more attractive.” Obviously, this is a partial analysis focused on certain jobs. Nor does it analyze the collaboration that can occur between humans and artificial intelligence to improve performance. And to this we must add another ingredient from the study: has been funded by IBM’s Watson AI Laboratory. Which means that behind the results there may be an interest in continuing to invest in AI without fearing the consequences.

“Our results should serve those responsible for creating economic policies when confronting AI labor automation – says the co-author of the study, Neil Thompson, in an interview -. But our results also reveal that this process will take years, or even decades, to develop and, therefore, there is time for those political initiatives to be implemented.