In many professions, human labor is still cheaper than AI: research

At present, there are no economically efficient methods to replace human labor with artificial intelligence (AI) technologies. This conclusion was reached by experts from the Massachusetts Institute of Technology (MIT) who conducted a study aimed at dispelling concerns about AI replacing humans in various industries.

In one of the early in-depth studies on the viability of replacing human labor with artificial intelligence, the researchers modeled the economic feasibility of automating various tasks in different sectors in the United States. The focus was on positions where the use of computer vision technologies would be required to replace humans, such as teaching or real estate appraisal. It was found that only 23% of representatives of the considered professions could be effectively replaced by AI when comparing their salaries and the cost of implementing the corresponding technologies into the production process. In other cases, using human labor remains more economically advantageous, as the integration and operation of visual recognition systems are costly.

Last year witnessed an acceleration of AI technology adoption across all industries, driven by the emergence of ChatGPT from OpenAI and other generative algorithms that vividly demonstrated the potential of neural networks. Technological companies, from Microsoft and Alphabet in the United States to Baidu and Alibaba in China, launched new AI-based services and announced plans to accelerate the development in this direction. All of this raised concerns that widespread neural network implementation might lead to a significant loss of jobs.

The researchers note that such concerns arose after the introduction of large language models. However, currently, less than a quarter of jobs could be replaced by systems with computer vision, as in other cases, it is not economically justified.

It is worth mentioning that computer vision enables machines to extract meaningful information from digital images and other visual data. It is used in various fields, including autonomous driving, photo classification on smartphones, and more. The cost-benefit ratio of implementing computer vision systems is most optimal in industries such as retail, transportation, and storage. The use of such systems may also be relevant in healthcare.

According to the researchers, currently, only 3% of 1,000 examined tasks in 800 professions can be automated through the use of AI systems with computer vision. However, by 2030, this figure could rise to 40% if the cost of data usage decreases and the accuracy of AI systems improves.