[DigitalToday reporter Jinju Hong (홍진주)] Robots can already technically perform a large part of the manual labor done by humans, but when costs are considered, only a very small share of jobs can be replaced more cheaply than people, an analysis showed.

On Wednesday, online media outlet Gigazine reported that Anthropic released research results estimating the current range and economic viability of robots performing manual labor, based on an analysis of about 900 occupations and 19,000 tasks.

In the study, Anthropic defined robots as autonomous physical machines that sense their surroundings and move, and examined how much of existing jobs robots at today’s technological level can perform. The research team entered a task database into Claude and evaluated whether robots can actually perform each task and what it would cost to achieve the same output as humans.

Robot task capability was divided into four levels. Tasks robots cannot perform were classified as E0. Tasks that can be performed in environments designed around robots, such as factory assembly lines, were classified as E1. Tasks possible in relatively structured environments built for humans, such as logistics warehouses, were classified as E2. Tasks that can be performed even in complex and unstructured environments, such as urban roads, were classified as E3.

By total labor hours, cognitive and interpersonal work accounted for 54 percent and manual labor for 46 percent. Of this, E0 tasks made up about 12 percent, E1 23 percent, E2 10 percent and E3 1 percent.

Combining E1 to E3, about 34 percent of all labor fell into areas that robots can perform today. Looking only at manual labor, the share rose to about 74 percent. From a technical perspective, that means robots can already carry out most human manual labor.

But the result changed significantly once costs were applied. A representative case was packaging work. The study assessed the robot replaceability of packaging at 97 percent. The median annual salary in the United States was estimated at about $49,000, and the annual cost needed to process the same output with robots was calculated at about $45,400. In this case, robots are cheaper than humans, meaning automation could be economically viable.

In contrast, tasks such as taxi driving and dishwashing lacked cost competitiveness even if robots could technically perform them. Dishwashing in particular was assessed as 100 percent replaceable, but the cost of using robots was estimated at more than three times that of human labor.

As a result, tasks that robots can perform at a lower cost than humans accounted for only 0.3 percent of the total. That means being technically possible and being economical enough to replace humans in real workplaces are completely different issues.

Falling robot prices are also a variable. Based on market data, the study found robot prices have fallen by about 3 percent a year on average since the 1990s. But even assuming that rate of decline continues, it was estimated it would take about 40 years for the share of tasks in which robots are cost-competitive versus humans to reach 10 percent of the total.

Anthropic said large-scale automation does not occur simply because robot prices fall below human labor costs. From a company’s perspective, it must consider not only the cost of introducing robots but also maintenance, operations and facility changes. To materially raise productivity, robots need to perform tasks at a significantly lower cost than humans, it said.

The analysis also shows that the impact of advances in AI and robots on the labor market is not determined solely by technical replaceability. Robots have advanced to a level where they can perform a substantial amount of manual labor, but the pace at which they displace humans in real workplaces may be driven more by cost structures than by technological capability.

The key variable in the robot era is shifting from "what can be done" to "what can be done more cheaply." The speed of actual labor replacement is expected to vary depending on how quickly robot prices fall with technological advances and how far companies’ introduction and operating costs decline.