When it comes to the future of intelligent robotics, the first question people often ask is: How many jobs will they kill? Whatever the answer may be, the second question will likely be: How can I guarantee this for me Work is not part of it?
In a study just published in Scientific robots, a team of robotics from EPFL and economists from the University of Lausanne provide answers to these two questions. By combining the scientific and technical literature on robotic capabilities with employment and wage statistics, they developed a method for calculating the current jobs that machines are likely to perform in the near future. In addition, they have developed a way to suggest career transitions to jobs that are less risky and require less retraining effort.
“There are many studies that predict the number of jobs that will be automated by bots, but they all focus on bots, such as speech and image recognition, financial bot advisors, chat bots, etc. Moreover, these forecasts fluctuate greatly depending on the How to assess job requirements and software capabilities. Here, we consider not only artificial intelligence software, but also real intelligent robots that perform physical labor and have developed a method for systematic comparison of human and robotic capabilities used in hundreds of jobs”, explains Professor Dario Floriano, Director of EPFL Laboratory Intelligent Systems, who led the study at EPFL.
The study’s main innovation is the mapping of new robot capabilities to business requirements. The team considered the H2020 European Multi-Year Roadmap for Robotics (MAR), a strategic document from the European Commission that is reviewed periodically by robotics experts. MAR describes dozens of capabilities required of the current robot or that future capabilities may require, which range, organized into categories such as manipulation, perception, sensing, and interaction with humans. The researchers examined research papers, patents, and descriptions of robotic products to assess the level of maturity of robotic capabilities, using a well-known measure of technological sophistication, the “Technological Readiness Level” (TRL).
As for human capabilities, they relied on the O*net database, a database of resources widely used in the US labor market, which ranks about 1,000 occupations and breaks down the skills and knowledge most important for each.
After selectively matching human capabilities from the O*net list with robotic capabilities from the MAR document, the team was able to calculate the probability that each current transaction would be executed by a bot. Suppose, for example, that a job requires a human to work with millimeter precision of motion. The bots are very good at this, so the TRL for the corresponding ability is the highest. If a job requires enough of these skills, it is more likely to be automated than a job that requires skills such as critical thinking or creativity.
The result is a classification of 1,000 jobs, where “physicists” are least at risk of being replaced by machines, and “slaughterhouses and meatpackers” are most at risk. In general, jobs in food manufacturing, construction, maintenance, construction, and extraction seem to present the highest risks.
“The main challenge facing society today is how to become resilient to automation,” says Professor Raphael LaLife. who co-directed the study at the University of Lausanne. “Our work provides in-depth career advice to workers who face high risks from automation, enabling them to take on safer jobs while reusing many skills learned in the old job. With these tips, governments can help society become more resistant to automation.”
The authors then created a method for finding alternative jobs, for a given job, with much lower risk of automation and reasonably close to the original in terms of the skills and knowledge they would need—thus reducing retraining efforts and making a career transition possible. To test how this method performs in real life, they used data from the US workforce and simulated thousands of occupational moves based on algorithm suggestions, and found that it would actually allow workers in high-risk occupations to move into medium-risk jobs. He trades, while undergoing a relatively low retraining effort.
This method can be used by governments to measure the number of workers who may face risks of automation and to adjust retraining policies, by companies to assess the costs of increased automation, by robotics manufacturers to better adapt their products to market needs; It is by the public to determine the easiest way to re-position oneself in the labor market.
Finally, the authors translated the new methods and data into an algorithm that predicts automation risks for hundreds of jobs and proposes flexible career transitions with minimal retraining effort, publicly available at https://lis2.epfl.ch/resiliencetorobots.