I work in robotics automation. The industry is poised for massive improvements in orchestration efficiency. Robotics currently does not take advantage of several known and trusted technologies. That alone will be huge. If we count the automation of driving jobs and efficiency improvements due to machine learning, I’d say the number is low.

Absolutely.

Government agencies i’m familiar with dropped about half of their processing and tabulation staff in the 90s. An agency with 12,000 employees in 1989 has 5,000 today.

About another 50-75% will be gone in a decade or two. Most of the case management, IT, program staff and big legal shops are bloated with legacy paper processing and old business process. As that disappears, so will the jobs.

A lot of 'jobs' are just repetitive unschooled work. Hardly something you indefinitely need a human for.

Job is a rather loaded term. How, exactly, are we defining it here? If your job was to weld two pieces of metal together and a robot now welds those pieces of metal together, but you transitioned to oversee the operation for quality control, strictly speaking, you lost your job to automation, but you didn't lose your 'job'.

Will two million people per year give up tasks they currently do to robots? Probably. That doesn't seem unrealistic at all. Will two million more people per year find themselves in the unemployment line? Probably not.

What's interesting is that (in the US), while manufacturing represented a larger share of the workforce in the past, the absolute number of people involved in manufacturing is relatively unchanged. The so-called decline of manufacturing is more a result of population growth than anything. Automation has allowed individuals to produce more, but it hasn't eliminated the people. The jobs that they do have no doubt changed considerably due to automation.

Just think of how many IT departments will disappear into the clouds