‘Too expensive,’ AI not capable of replacing humans in workplaces, study reveals

Artificial Intelligence, as of now, is not capable of replacing humans in most of the jobs in a cost-efficient manner, a new study conducted by the Massachusetts Institute of Technology (MIT) has found. The study emerges amidst concerns that the rapid advancement of AI can potentially impact the livelihood of millions of job-seekers across the world.

The MIT study specifically focused on those jobs where computer vision was employed to assess whether AI can replace humans in an economically viable manner. Computer vision is a branch of Artificial Intelligence (AI) used for deriving meaningful information from digital images and other visual inputs, with its most ubiquitous applications observed in object detection systems for autonomous driving or in helping categorise photos on smartphones.

Only 23% of workers can be replaced: Study

Add WION as a Preferred Source

The study revealed only 23 per cent of workers can be replaced by AI in a cost-effective manner, if measured in terms of dollar wages. AI-powered virtual recognition devices and systems are so expensive to install and maintain that humans might end up doing the job more economically.

Watch:Challenges & opportunities in the retail renaissance with Gen AI

“‘Machines will steal our jobs’ is a sentiment frequently expressed during times of rapid technological change. Such anxiety has re-emerged with the creation of large language models,” the researchers from MIT said in their 45-page paper titled ‘Beyond AI Exposure’. “We find 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,” they concluded.

Sectors where the application of AI is economical

Computer vision's cost-benefit ratio shines brightest in sectors such as retail, transportation, and warehousing, notably dominated by Walmart Inc. and Amazon.com Inc. The potential extends to healthcare, according to MIT's paper. The authors suggest a more assertive AI deployment, especially through AI-as-a-service subscriptions, could enhance viability and broaden applications.

Funded by the MIT-IBM Watson AI Lab, the study, utilising online surveys, gathered data on around 1,000 visually-assisted tasks across 800 occupations. Presently, only 3 per cent of these tasks can be cost-effectively automated, but the researchers anticipate a potential increase to 40 per cent by 2030, contingent on falling data costs and improved accuracy.

IMF’s warning

The advent of advanced AI models like ChatGPT and its competitors like Google's Bard has raised concerns about massive job displacement, as these chatbots exhibit proficiency in tasks previously exclusive to humans. The International Monetary Fund cautioned last week that nearly 40 per cent of global jobs could be impacted, emphasising the need for policymakers to carefully balance AI's potential with potential negative consequences.

(With inputs from agencies)