A recent study by the Massachusetts Institute of Technology shows that only 23% of work tasks exposed to AI would be cost-effective to automate. Although AI has the potential to radically transform the workplace, MIT's 'Beyond AI Exposure' study shows that there are still significant financial barriers to its wider adoption.
The statement "machines will steal our jobs" is an expression that often comes up in times of rapid technological change. This anxiety has resurfaced with the creation of large language models (e.g., ChatGPT, Bard, GPT-4) that demonstrate significant proficiency in tasks where previously only humans could deliver satisfactory results.
This includes tasks that require a degree of creativity and complex problem solving that were previously thought to be uniquely human domains. The study found that the switch is too expensive due to the large initial costs of the AI systems. 77% of vision tasks are therefore not economical to automate if a system can only be used at company level.
This contrast makes it clear that the cost-effectiveness of AI models will play an important role in the diffusion of the technology. The study pointed out that even if an AI computer system costs $1,000, there are still jobs that are not economically attractive to replace, such as low-wage jobs and work in small firms.
However, it emphasized that AI could be made more attractive through decreases in implementation costs or by increasing the scale at which implementations are made. "Overall, our model shows that the job loss from AI, even within the set of tasks, will be less than the existing job rotation seen in the market, suggesting that labor replacement will be more gradual," the study added.