Definitions:

Narrow AI - artificial intelligence capable of doing a narrow task at a level closely matching or exceeding that of a human.
Long-term unemployment - using the United States Bureau of Labor Statistics (BLS) definition: unemployment lasting 27 weeks or longer.
Automation - technology which increases the output of an employee. At the far end there is complete automation, where the employee is no longer required.

Historical effect of automation on unemployment

At first, let's take a look at how automation has affected employment in the past. The historical account will help us shape our argument: whether we should argue for continuation / acceleration of the trend or for its reversal.

Great historical shifts in automation

Farming: Over the last four centuries humans went from 95% of the population being farmers to 2% of the population being farmers (in agriculturally self-sufficient developed countries). We don't live in a world where 93% of the people are unemployed because 93% of the jobs went away.

Automobile: We didn't get mass unemployment when the Model T automobile mechanized the entire horse-and-buggy industry out of existence.

Great Recession

Jobs lost during the Great Recession (Q3-2008 until Q2-2009) have been slow to come back. The Robots, AI, and Unemployment Anti-FAQ argues that this is most likely due to issues with re-employment rather than issues with automation.

Germany labor laws

Germany saw unemployment drop from 11% to 5% from 2006-2012 after implementing a series of labor market reforms, though there were other things going on during that time. (Germany has [ twice the number of robots per capita][2] as the US, which probably isn't significant to their larger macroeconomic trends, but would be a strange fact if robots were the leading cause of unemployment.)

Theoretical arguments

Second, let's examine the theoretical arguments for why automation can or cannot affect unemployment.

Conventional economic theory argument

Conventional economic theory says this shouldn't happen. Suppose it costs 2 units of labor to produce a hot dog and 1 unit of labor to produce a bun, and that 30 units of labor are producing 10 hot dogs in 10 buns. If automation makes it possible to produce a hot dog using 1 unit of labor instead, conventional economics says that some people should shift from making hot dogs to buns, and the new equilibrium should be 15 hot dogs in 15 buns. On standard economic theory, improved productivity - including from automating away some jobs - should produce increased standards of living, not long-term unemployment. The idea that there's a limited amount of work which is destroyed by automation is known in economics as the " lump of labour fallacy".

Other arguments

Third, let's look at other arguments.

Speed of adoption

One factor that's significantly different today is that a lot of technologies can be adopted much faster. Software can be cloned and distributed almost instantly and for free. Hardware can be mass-produced and distributed a lot faster now. New products can be marketed and adopted by the consumers at a much faster rate. All of this leads to new automation solutions being adopted at a much faster rate. There is a difference between cars being in 60% over 25 years vs. 15 years for internet vs. (measuring until the technology is in 60% of the US households).

Specific technologies

Fourth, let's take a look at specific technologies that might affect long-term employment and see how they might interact with the above-mentioned evidence.

Self-driving cars