Will artificial intelligence take our jobs and replace human workers with machines? Although recent headlines would lead you to believe that AI systems will substitute humans as soon as possible, recent research from MIT says otherwise.

MIT conducted a study funded by the MIT-IBM Watson AI Lab and used online surveys to collect data on roughly 1,000 visually assisted tasks across 800 occupations. It explored the integration of computer vision systems in these various work settings and found that the cost of implementing such technology often does not justify the potential savings in human labor expenses.

The cost-benefit analysis of computer vision AI in the workplace

The study found that 36% of jobs in U.S. non-farm businesses have at least one task that is exposed to computer vision, but only 8% have at least one task that has enough of an economic benefit for it to be attractive for a firm to automate the task. The report says that since only a tiny fraction (2%-30%) of any occupation can be classified as a vision task, the more relevant metric to consider when analyzing operations to see if it would make sense to implement AI would be the cost benefits of AI on the company.

In that regard, the report found that in many occupations with low wages, where there are few tasks per person and many workers doing the same task, the cost savings introduced by automation aren’t significant enough to justify making the investment and incurring the expense of creating (or buying), implementing and training AI. In other words, even though most AI systems are more than capable, not all tasks that AI can perform are economically viable to replace human labor. For large businesses, the high cost associated with fine-tuning AI systems for specific tasks may be feasible and worth the investment, but for small to medium-sized businesses, the cost is prohibitive.

An example from the study that showcases this is when the researchers looked at the potential use of computer vision in a bakery. One task that bakers do is visually check their ingredients to ensure they are of sufficient quality, for example, to ensure the ingredients haven’t spoiled. This task could be replaced with a computer vision system by adding a camera and training the system to detect food that has gone bad. To see if this is worth it for a small baker, the report calculates the cost savings that come from this sort of computer vision system.

The study found that quality-checking ingredients accounted for ~6% of the duties of a baker and that the average small bakery has five bakers making around $48,000 per year. Therefore, the potential savings from automating this task would be equivalent to roughly $14,000 per year. However, the researchers believe that the amount saved—14,000 per year—is far below the cost of developing, deploying, and maintaining a computer vision system and, therefore, is not an economical substitute for human labor in this scenario. In other words, it will cost the bakery more than $14,000 to get its AI system up and running. Therefore, the savings they would experience due to the automation are canceled out by the cost of owning and operating the AI system that makes it possible.

Computer vision AI vs. multimodal AI

It’s important to note that this study only focused on computer vision, which is much different than more dynamic systems like multimodal large language models, including
OpenAI’s GPT-4. While computer vision is very specific, multimodal large language models are much broader in their ability to execute tasks, and the tasks that they can run closely resemble the cognitive tasks that humans complete throughout the day.

recent study by OpenAI estimates that 19% of U.S. workers could see 50% of their workplace tasks impacted by advanced AI systems like GPT-4, which is much more significant and is bound to have much more of an effect on the workplace than computer vision.

While the report suggests that AI has transformative potential for the workplace, it acknowledges that the widespread integration of a computer vision system is not as imminent as some might expect. The study found that computer vision is currently capable of automating tasks that represent 1.6% of worker wages in the U.S. economy (excluding agriculture). However, only about 0.4% of the economy could benefit from cost savings through such automation. This is because the less dynamic an AI system is, the fewer jobs and tasks it is likely to effectively automate. This makes it a sub-optimal implementation for businesses, especially smaller businesses, due to the cost of implementing and operating the system.

I think the one area where the report falls short is that it does not look at the cost savings experienced by computer vision automation on long time horizons. It isn’t unusual to hear that a new system does not pay for itself during year one, similar to how the report describes the small bakery example, but usually, as time goes on and the costs to operate a system significantly decrease, there does comes a point where the system reaches break even and becomes a tool that saves the company more money than it costs to operate the tool leading to an overall increase in efficiency and cost savings.

It also does not explore the value that could be created if each worker could effectively use that 6% of the time they use to examine ingredients in other areas, which could likely lead to more revenue being generated for the business and slightly offsetting the costs of the AI.

Regardless, as the year progresses, we are bound to see more case studies and reports that examine AI’s impact on the workplace and try to answer the question of whether AI is indeed a threat (substitute) to human workers that will make them redundant by fully automating their jobs, or if AI is more of a complimentary tool that increases productivity and allows human workers to spend more time on the tasks that can’t be automated.

In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI.

Watch: Artificial intelligence needs blockchain

YouTube video

New to blockchain? Check out CoinGeek’s Blockchain for Beginners section, the ultimate resource guide to learn more about blockchain technology.

Thank you for engaging with us at SmartLedger through 'MIT study explores computer vision AI in workplaces' - https://smartledger.solutions/mit-study-explores-computer-vision-ai-in-workplaces/. We hope you found the insights valuable.

For more thought leadership and updates, delve deeper into our resources and stay ahead with the latest innovations.

 

This post was originally published on this source site: this site

image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog
image-blog