MIT Researchers Claim It Would 'Take Decades' For AI Computer Vision Tasks 'To Become Economically Efficient' Compared To Human Workers

A main source of apprehension from people who are hesitant about AI usage is the notion that the popular technology could replace working professionals. As previously shared by AFROTECH, the World Economic Forum suggested that AI will replace 85 million jobs across the world by 2025.

Specifically, within the Black community, a 2019 McKinsey study shared an estimate of 4.5 million jobs performed by Black Americans would be disrupted by AI by 2030. The estimate was based on the fact that automation would likely replace fast food and service positions, which Black people “are overrepresented in.”

While there are a plethora of estimates and projections, new research has emerged that is bringing a different take. In a working paper study by Massachusetts Institute of Technology (MIT), titled “Beyond AI Exposure: Which Tasks are Cost-Effective to Automate with Computer Vision?,” the researchers used the “first end-to-end AI automation model” to help address “anxieties about AI,” per Forbes.

The outlet explains that the group, from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), utilized the model to factor in the level of proficiency required for vision-related tasks based on surveys from relevant workers, the cost of humans completing the work compared to AI systems, and “the economic decision by firms whether or not to adopt AI for a particular task.”

After using their model, the MIT researchers’ findings included that “only 23% of worker compensation ‘exposed’ to AI computer vision would be cost-effective for firms to automate because of the large upfront costs of AI systems.” What’s more, in regard to the scenario of cutting down the costs of AI development and/or deployment, “even with rapid decreases in cost of 20% per year, it would still take decades for computer vision tasks to become economically efficient for firms,” the report shared.

In other words, based on their findings, it may not be likely that AI is going to come in swinging and “taking jobs.” The MIT researchers also claimed that their research is relevant to generative AI, stating that “fine-tuning” and “customization” is still cost-restrictive. Such insight may put at ease the findings of another report.

As previously shared by AFROTECH, a study by McKinsey’s Digital Practice and the McKinsey Institute for Black Economic Mobility showed that generative AI could increase the racial wealth gap by $43 billion.