MIT Research: Replacing Workers with Artificial Intelligence is Actually Much More Expensive

A team of researchers from MIT (Massachusetts Institute of Technology) has discovered that in many cases, replacing human workers with artificial intelligence is still significantly more expensive than employing human workers, a conclusion that contradicts current fears about technology supposedly taking our jobs.

As detailed in a new paper, the team examined the cost-effectiveness of 1,000 ‘visual inspection’ tasks across 800 occupations, such as inspecting food to see if it is spoiled. They found that only 23 percent of total worker wages would be attractive to automate, primarily due to ‘high initial costs of AI systems’.

However, they acknowledge that this economy could change over time.

– Generally, our findings suggest that the takeover of jobs by artificial intelligence will be gradual, thus there is room for policy and retraining to mitigate the impact on unemployment – concluded the team in their paper.

This topic, regarding the takeover of jobs by artificial intelligence, has become unavoidable lately, especially with the emergence of tools like OpenAI’s ChatGPT and Google’s Bard. However, this is still not true artificial intelligence, but rather a kind of beginning.

While many have warned of the dire consequences that significant job losses could have in the near future, technology leaders remain optimistic about such possibilities, claiming that these jobs will be replaced by new types of occupations, which they do not know.

According to OpenAI’s CEO Sam Altman, who spoke at last year’s Wall Street Journal Tech Live conference, this is an inevitable part of any ‘technological revolution’.

Impact on the Technology Sector

However, the greatest impact on jobs is currently occurring in the technology sector, which has been affected by ongoing layoffs despite significant investments in artificial intelligence.

During this month’s World Economic Forum in Davos, Switzerland, a survey among executives showed that a quarter intend to reduce their workforce by at least five percent ‘due to generative artificial intelligence’. While there seems to be a consensus that artificial intelligence will one day come for our jobs, it is still a topic of heated debate as to when such a change will actually occur.

In their paper, the MIT team focused on tasks assisted by computer vision, using the example of a bakery worker who visually inspects ingredients to ‘ensure they are of sufficient quality’. Such a job ‘could theoretically be replaced by a computer vision system by adding a camera and training the system to detect spoiled food’, the researchers write.

However, installing and operating such a system would still be too expensive, considering it would only deal with ingredient inspection, which represents only six percent of the employee’s job. Only when the costs of implementing such systems decrease can we talk about replacing workers with AI.

The feeling that machines will take our jobs is often expressed during times of rapid technological change, and such anxiety has resurfaced with the creation of large language models (e.g., ChatGPT, Bard, GPT-4) that demonstrate significant skill in tasks previously only human beings showed expertise in. While other experts have concluded that these fears are not unfounded, many of them fail to ‘directly consider the technical feasibility or economic sustainability of AI systems’, claim the MIT researchers.

However, many questions remain. What about jobs that do not involve visual analysis, like the bakery example the researchers cite? What about jobs that can be supplemented by artificial intelligence instead of being completely replaced? As TechCrunch points out, the MIT research was supported by IBM’s Watson AI Lab, which means there may have been a financial interest in downplaying the risks of job replacement by AI.

The authors, however, argue that it is simply about creating significant regulatory frameworks to prepare for the future.

– For policymakers, our results should emphasize the importance of preparing for job automation by AI. However, our results also reveal that this process will take years, or even decades, and therefore there is time to initiate political initiatives – said MIT researcher and co-author Neil Thompson to TechCrunch.