Artificial intelligence cannot effectively replace most jobs at the moment, according to a study by the Massachusetts Institute of Technology (MIT), which aimed to address fears of AI replacing humans in a number of industries.
In one of the first detailed investigations into the viability of replacing labor with AI, researchers modeled the cost attractiveness of automating various tasks in the US, focusing on jobs where computer vision was being used – for example, teachers and assessors of properties.
They found that only 23 percent of the people doing these tasks, measured in terms of dollar wages, could be effectively replaced. In other cases, due to the high costs of getting AI-assisted visual recognition up and running, humans have been more efficient at performing those tasks.
AI adoption in various industries accelerated last year after OpenAI’s ChatGPT and other generative tools showed the technology’s potential. Technology companies such as Microsoft Corp. and Alphabet Inc. from USA or Baidu Inc. and China’s Alibaba Group Holding Ltd. have launched new AI services and ramped up their development plans — at a speed that some industry leaders have warned is too fast. Fears about the impact of AI on jobs have long been a major concern among the public – at every level.
«Machines will steal our jobs’ is a sentiment frequently expressed in times of rapid technological change. Such anxieties have resurfaced with the creation of generative language models,» researchers from MIT’s Laboratory for Computer Science and Artificial Intelligence said in the 45-page paper titled «More Than the AI Exposition.»
«We found that only 23% of workers’ compensation ‘exposed’ to AI computer vision would be cost-effective, because AI systems cost a lot initially — to implement and train them, especially.»
Computer vision is an area of AI that enables machines to extract meaningful information from digital images and other visual data, with the most common applications appearing in object detection systems for self-driving cars or categorizing photos on smartphones.
The cost-benefit ratio of computer vision is most favorable in sectors such as retail, transportation and warehousing – all areas where Walmart Inc. and Amazon.com Inc. are prominent, for example.
It’s also feasible in the context of healthcare, according to the MIT paper. A more aggressive expansion of AI deployment, particularly through subscription AI services, could also increase other uses and make them more viable, the authors said.
The study was funded by the MIT-IBM Watson AI Lab and used online surveys to collect data on about 1,000 visually assisted tasks from more than 800 occupations. Only 3 percent of such tasks can be effectively automated today, but that percentage could reach 40 percent by 2030 if data costs fall and accuracy increases, the researchers said.
The sophistication of ChatGPT and its rivals, such as Google’s Bard, has revived concern that many jobs will be replaced by AI, as new chatbots demonstrate proficiency in tasks previously only humans were capable of.
The International Monetary Fund said last week that nearly 40 percent of jobs globally would be affected, and that policymakers should carefully balance the potential of AI with the negative consequences.
At the World Economic Forum in Davos last week much of the discussion focused on the replacement of labor by AI. Co-founder of Inflection AI and Google’s DeepMind, Mustafa Suleyman, said AI systems are «fundamental tools to replace labor.»
A case study in the paper looked at a bakery, as an example: Pastry chefs visually inspect ingredients for quality control every day, but this is only 6 percent of their duties, the researchers said.
The time savings and wages saved by implementing video cameras and an AI system are still far from the cost of such a technological upgrade, they concluded: «Our study looks at the use of computer vision in the economy, examining its applicability to every occupation in almost every industry and sectors,» Neil Thompson, director of the FutureTech Research Project at MIT’s Computer Science and Artificial Intelligence Laboratory, told Bloomberg . «We show through this study that there will be more automation in retail and health care, and less in fields like construction, mining or real estate.»
