The Massachusetts Institute of Technology (MIT) conducted a study that revealed currently, in terms of cost-effectiveness, artificial intelligence (AI) is not poised to replace a significant portion of jobs.
This research, one of the earliest comprehensive assessments of AI’s potential to take over human roles, focused on the United States job market, particularly roles that utilize computer vision, such as educators and real estate appraisers.
The findings indicated that only about 23% of jobs when evaluated in terms of wage expenditure, are viable candidates for replacement by AI. In many instances, the high costs of implementing and maintaining AI systems for visual recognition make human labor more financially practical.
Last year marked a significant uptick in the integration of AI across various sectors, following the demonstration of its capabilities by OpenAI’s ChatGPT and similar generative technologies. Major technology companies, including Microsoft and Alphabet in the United States, as well as Baidu and Alibaba in China, introduced new AI services and escalated their development strategies.
However, this rapid advancement prompted warnings from some sector leaders about the precipitous pace. Concerns regarding the effect of AI on employment have consistently been at the forefront of industry discussions.
The field of computer vision in artificial intelligence allows machines to extract significant insights from digital images and other visual inputs. This technology is most commonly seen in applications such as object recognition in autonomous vehicles or the classification of images on smartphones.
In terms of cost-effectiveness, computer vision proves most advantageous in industries like retail, transportation, and warehousing. Companies like Walmart and Amazon are key players in these areas. The technology is also practical in healthcare, according to MIT. The authors of the paper suggest that a broader deployment of AI, particularly through AI-as-a-service subscription models, could expand its utility in other sectors and enhance its feasibility.
Conducted by the MIT-IBM Watson AI Lab, this study employed online surveys to gather information on approximately 1,000 tasks involving visual assistance across 800 different jobs. Presently, a mere 3% of these tasks can be automated cost-effectively. However, the researchers project that this percentage could escalate to 40% by 2030, contingent upon reductions in data costs and enhancements in accuracy.