A recent study from the Massachusetts Institute of Technology (MIT) has provided a new perspective on the ongoing debate about artificial intelligence (AI) replacing human jobs. Contrary to popular belief, the study reveals that human labor remains more cost-effective than AI in most job roles, especially tasks that require visual processing.
The study, a joint effort between MIT, IBM and the Productivity Institute, surveyed workers in a variety of sectors to determine the capabilities computers need to perform their tasks. The study then estimates the costs of developing and deploying such AI systems, comparing them to human wages. The findings are significant: only about 23% of workers’ wages paid for visual tasks would be attractive for automation with current AI technology. Essentially, AI systems, especially those involving computer vision, are currently too expensive to replace employees in over three-quarters of the jobs considered.
The MIT study, supported by the MIT-IBM Watson AI Lab, analyzed over 1,000 visually assisted tasks in 800 different occupations. Data shows that currently only 3% of these tasks can be economically automated. Even with a projected 20% annual reduction in AI system costs, it will still take decades for AI to become more economically viable than human labor in most companies. Additionally, AI’s high energy consumption and significant implementation challenges further limit its current viability as a replacement for human workers.
One of the key findings is that AI struggles with tasks that require implicit knowledge, intuition, or instinct—abilities deeply rooted in human cognition and critical to many job roles. While AI is expected to impact specific sectors such as banking, marketing, healthcare, and transportation due to the repetitive nature of tasks in these fields, its ability to completely replace human labor seems overstated, at least for now.
The study’s implications go beyond economic considerations, touching on broader societal impacts such as workforce retraining and policy development. It highlights the potential of AI to create new job categories focused on managing, maintaining and improving AI systems, as well as roles where human skills are indispensable to AI. This could lead to the emergence of new business models, including AI-as-a-Service platforms, democratizing access to AI technologies for smaller businesses and organizations.
In conclusion, the MIT study suggests a more gradual integration of AI across sectors, in contrast to the often assumed rapid AI-driven job displacement. It calls for a more systematic assessment of the feasibility of adopting new technologies across industries, taking into account the economic and practical limitations of AI systems.
Image source: Shutterstock