Despite its impressive capabilities, artificial intelligence (AI) remains a costly option for certain tasks. As per insights from Fortune 500 CEOs and Silicon Valley leaders, AI is viewed as a potential threat to various job roles. However, the current pricing of AI computer vision technologies does not make them economically feasible for the majority of businesses in the US.
A recent study explored the automation potential of human work, particularly tasks related to vision, through AI technology. The analysis concentrated on assessing the viability of replacing vision-related tasks across different job categories with existing AI computer vision solutions. While there is a multitude of tasks suitable for AI integration, the high costs often hinder practical automation.
Researchers, led by Neil Thompson from the Massachusetts Institute of Technology, identified 414 vision tasks within US job sectors that could be candidates for automation using current AI capabilities. These tasks span from retail store supervisors validating price tags to nurse anaesthesiologists overseeing patients for physiological changes. The study involved evaluating the expenses linked to training and implementing AI models for these tasks, comparing them to the costs of human labor including salaries and benefits.
The study findings indicated that while 36% of non-agricultural businesses in the US have tasks suitable for AI automation, only 8% of these tasks are cost-effective for automation. Furthermore, employers could feasibly automate only 0.4% of worker salaries and benefits. Even large firms with over 5000 employees, constituting less than 0.1% of US companies, could automate only a fraction of their vision-related tasks due to the current high costs of AI adoption.
Despite offering some relief to US businesses, Gino Gancia from Queen Mary University of London suggests the existence of other AI applications with lower automation expenses. The swift uptake of “generative AI” for content creation has already impacted freelance opportunities on platforms like Upwork, resulting in job scarcity and diminished earnings for human workers. Regions at the forefront of AI integration, such as California, have witnessed significant job losses, hinting at a potential surge in inequality among industries and workers.
Thompson and his team foresee a substantial automation of human tasks in the future, contingent on reducing training and development costs for AI technologies. Despite the impending wave of automation, Thompson stresses the significance of governments implementing programs to aid displaced workers in adapting to the evolving job landscape.
