Only 23% worker wages worth automation: MIT

Long way to go: The cost-effectiveness of artificial intelligence will play a key role in its proliferation, as per a recent study.

Massachusetts Institute of Technology (MIT)’s study also highlights that computer vision tasks, or automatic recognition of images, will take decades to be deployed even if costs are decreased 20 per cent per year.

The working paper on ‘Beyond AI Exposure: Which Tasks are Cost-Effective to Automate with Computer Vision?’ implies that majority of jobs cannot be automated cost-effectively, amid fears of the technology rapidly replacing jobs.

The study takes into account previous work done on artificial intelligence, wherein large language models (LLM) or AI algorithm for learning techniques can partially automate atleast 50 per cent of tasks. 

The present LLMs, in the form of Open AI’s ChatGPT, Google’s Bard and Elon Musk’s AI foray Grok present fears of rapid automation in the knowledge economy.

As per a McKinsey study, AI adoption is estimated to be between 4 per cent and 55 per cent.

Using an example of bakeries opting to automate a routine task of detecting spoilt goods, the study found that it would save $14,000 per year, since it would automate only 6 per cent of a baker’s tasks.

Accounted for when a baker makes $48,000 per year, the study concluded that the portion in savings is “far less” than the cost of developing, deploying and maintaining a computer vision system.

The study points that labour replacement will be gradual rather than abrupt, and that job loss from AI functions like computer vision  will be smaller than the existing job churn seen in the market.

Additionally, the study suggests that AI systems doubling the benefit of the human worker would increase the share of compensation that is attractive to automate from 23 per cent to 30 per cent, inferring that exponential changes in benefits are needed for linear changes in the share of automation.