The rise of artificial intelligence (AI) has many professionals fearing for their jobs. However, a new study from the Massachusetts Institute of Technology (MIT) shows that the cost of deploying technology makes it likely more economical for your employer to keep you on, at least for now.
When professionals think about jobs replaced by AI, most only consider the technology’s ability to perform human tasks. This is why many studies focus on tasks that AI could automate, which only exacerbates the feeling of job insecurity.
However, the MIT study titled “Beyond AI Exposure” deviates from this usual approach and takes into account a much overlooked factor: cost.
The example of a small bakery that would use computer vision to automate the quality of its ingredients
In this study, the researchers first interviewed professionals to understand what performance would be required from an automated system. They then modeled the cost of building such a model and, finally, they determined whether adopting AI was economically worthwhile. It should be noted that the study focuses on tasks that can be automated by computer vision, such as quality inspection.
An estimate of the cost of task automation is a good indicator of what the future of business will look like because, ultimately, implementing these sophisticated AI models comes at a high cost, and companies do not are interested in investing in technology only if it provides a return on investment.
To explain why consideration of cost is so essential, the study uses the example of a small bakery that could consider using computer vision to automate visual inspection of the quality of its ingredients.
77% of computer vision tasks are not worth automating
Since this task represents only 6% of a baker’s tasks, a small bakery with five bakers earning $48,000 per year can save up to $14,000 by automating these tasks. But the cost of deploying computer vision is much higher than the money saved, so it would not make financial sense to automate this task. And this even if the technology is available.
The 45-page document presents the framework and methods used to conduct the study, as well as the results. And all of this should help ease your employment concerns. “We find that only 23% of compensation for professionals ‘exposed’ to AI machine vision would be profitable for businesses due to the high upfront costs of AI systems,” the study says.
In addition, 77% of computer vision tasks are not worth automating due to the low profitability of deploying a single system, i.e. one that can only be used at one level. one and the same company.
Waiting for economies of scale
The study acknowledges, however, that as the technology develops, the cost of its deployment will decrease. However, even if costs fall by 20% per year, it will still take decades for the use of computer vision to become profitable for most businesses.
Although the study does not consider other AI use cases, such as text generation, the researchers plan to use their method to study areas other than computer vision.
And in this area, the first prices offered by Microsoft, for example, already make it possible to model ROI. Microsoft 365 Copilot, available for a few days at €30 per month, helps boost the online office suite. But at this price level, CFOs and business departments will have to make some calculations before replacing payroll with AI operating costs.
Source: “ZDNet.com”