AI Challenges: Jobs

This article is part of the series "How AI challenges humanity". In the article we'll look at how machine learning will probably affect the world of jobs. First through a pessimistic and then through an optimistic lens.

Threats

An algorithm that’s able to perform a cognitive task better than a human at a comparable price will take that task away from the person. The estimates for western economies range from 35% to 49% of jobs that can be automated in the near future.

The transition period will depend on the necessary capital investments. It would take longer to transition to fully autonomous transport than to automate all translations. Driverless cars require a suite of sensors, computational power, actuators and probably even new cars. At the current replacement rate of vehicles a full transition would take 15 to 20 years. Fully automated translations would mainly require software that can be deployed across the globe in weeks.

Whether the transition to automated cognitive tasks will generate as many new jobs as it will destroy is currently debated. People on the one side argue that this time is different and when machines take away routine cognitive tasks there won’t be any other routine cognitive tasks left to perform. Whether that’s true or an equal amount of new jobs will indeed be created, these transition periods never went smoothly in the past. Even if everything worked out in the end.

The earlier this transition is managed the better people will fare.

Opportunities

A repetitive task lost to automation is time freed up for creative and more valuable work. People who believe no new jobs will be created fall victim to the lump of labor fallacy. The amount of work available in an economy is not fixed and so people are free to learn new skills, take on new jobs or simply work less.

As we’ve explored in the threats-section the transition in some areas might happen slower than expected. Automating the operation of a machine requires adding sensors that sometimes are expensive. For planes or long-range buses the cost of the human operator gets split by so many passengers that the actual share in the ticket price is minimal. The main costs are often capital expenses. Automation will only add to those.

Ways Forward

There is a lot of debate and a lot of uncertainty about the amount of endangered jobs, whether or not new jobs will be created and about the timeframe of this transition. There is more or less consensus on what kind of jobs are threatened: Routine manual and routine cognitive tasks with little or standardized human interaction. Typical and often cited examples include radiologists, checkout operators, accountants, telemarketers, technical writers, typists, translators.

Governments need to monitor the progress of machine learning closely and not only evaluate technical feasibility but also economic factors to predict the timeframe of automation in different groups of jobs. They also have to prepare to (re)train up to 50% of the population.

In the best-case scenario as many jobs are created as are lost. But then the real work starts for people, educational organizations and governments. People have to acquire the skills necessary to transition to the newly created jobs.

Luckily the technology that will then have put people out of work can also help make education more efficient. Institutions need to change accordingly or societies will pay a hefty price.