Artificial intelligence may be coming for many people’s jobs according to some Fortune 500 CEOs and Silicon Valley leaders – but AI computer vision technologies are not yet cheap enough to make them worthwhile for most US businesses today. The finding comes from a study of human work – specifically tasks involving vision – that is exposed to the risk of machine automation.
In the study, researchers focused on whether vision tasks included in various human jobs are economically worth replacing using existing AI computer vision. “There are lots of tasks that you can imagine AI applying to, but actually cost-wise you just wouldn’t want to do it,” says Neil Thompson at the Massachusetts Institute of Technology, co-author of the study, which was published today as a working paper.
Thompson and his colleagues identified 414 vision tasks in US job categories that could potentially be automated by existing AI technology. These jobs included retail store supervisors who visually check on whether items have the right price tags attached, and nurse anaesthesiologists who are trained to watch the patients in their care for dilated pupils or changes in cheek colour that might be warning signs of potential problems.
The researchers calculated the costs of training and operating an AI computer vision model capable of handling those tasks with the required accuracy. They then compared AI costs with the costs of human labour – the latter represented as share of total worker salaries and benefits, because vision tasks typically make up a small portion of any given employee’s work duties.
They found that, although 36 per cent of US non-agricultural businesses have at least one worker task that could be automated through AI computer vision, just 8 per cent have a task that is cost-effective to automate using AI.
They also concluded that just 0.4 per cent of US non-agricultural worker salaries and benefits would actually be cost-effective for employers to automate.
The current costs of AI computer vision mean that even large US firms with 5000 or more employees – bigger than 99.9 per cent of all US companies – could cost-effectively automate less than one-tenth of their existing vision tasks.
Such findings about AI computer vision being too costly for most US businesses “might sound like a reassuring result” but “there might be other [AI] applications with lower automation costs”, says Gino Gancia at Queen Mary University of London.
The rush to adopt “generative AI” that can create new content has already negatively impacted the number of available jobs and earnings of human freelancers in online platforms such as Upwork. Gancia’s research has also shown that US regions with industries that are adopting AI more quickly – such as California – have already experienced greater job losses.
“In general, we know that new technologies diffuse unevenly,” says Gancia. “As a result, it is likely that automation and AI will contribute to increasing inequality across firms and workers.”
Thompson and his colleagues do expect a significant amount of human work to be automated in the long run. But that depends on how quickly the costs of training and developing AI technologies can drop.
“There is going to be a substantial amount of automation and governments have to start preparing for it,” says Thompson. “But there is enough time for us to be putting into place real programs that can benefit [displaced] workers.”
Felecia Phillips Ollie DD (h.c.) is the inspiring leader and founder of The Equality Network LLC (TEN). With a background in coaching, travel, and a career in news, Felecia brings a unique perspective to promoting diversity and inclusion. Holding a Bachelor’s Degree in English/Communications, she is passionate about creating a more inclusive future. From graduating from Mississippi Valley State University to leading initiatives like the Washington State Department of Ecology’s Equal Employment Opportunity Program, Felecia is dedicated to making a positive impact. Join her journey on our blog as she shares insights and leads the charge for equity through The Equality Network.