Fears of AI automating jobs en masse are overblown, according to an MIT study on computer vision technology. By modeling costs, researchers found just 23% of US worker wages could be effectively replaced currently. In most cases, humans still do tasks more affordably than AI. Retail, transportation and healthcare are most ripe for automation. But sectors like construction remain far costlier to automate.
MIT examined nearly 1000 visual tasks across 800 occupations. Only 3% can be automated in a cost-effective manner today. By 2030, 40% of tasks could be viable as data and accuracy improve. But upfront outlays for most companies still outweigh potential savings.
One hypothetical bakery case showed quality control inspection by AI is still too expensive. The time savings don't offset the system's price tag. AI adoption accelerated after ChatGPT demonstrated new potentials. But implementation must be balanced with workforce impacts, IMF and others caution. For now, humans remain more economical for most visual jobs. But continued innovation could eventually tip the scales.
Comments