The concern is that automated systems will become better-than-human at enough tasks of economic importance that involuntary joblessness will significantly increase. The usual counter-argument is "human wants are infinite," implying that there will always be sufficient demand for new kinds of or additional quantities of paid human employment to counterbalance labor saving techniques. But that argument neglects time to reach a new equilibrium. It neglects that people displaced from old industries may not be able to fill newer jobs even if they had years to retrain. It neglects that a major human want is leisure time, and leisure time not spent consuming or producing does not contribute to paid employment.

> What happened in the past was that people stopped being handy with a scythe and started getting handy fixing tractors

What happened is that people stopped being farmers. The number of farmers per capita dropped by something like 95%, compared to a century ago.

Something like 20% of jobs are moving stuff around. Drivers, warehouse workers, truckers. What happens with self driving vehicles doing that? (the great depression was 25% unemployment).

People stopped being farmers and were able to pursue other industries. But there were still farmers, and a whole support economy of new jobs to support the new method of farming. AI is going to be the farmer driving the tractor, the person who repairs the tractor, the system that manufactures new tractor parts, the person that drives it, a large amount of the support designing new tractors (even if there were 100 competing brands of tractors there are a limited number of people required for research and design and maintenance of the automation (not enough to replace 2.1 million farmers, let alone the millions of factory workers, mechanics, and delivery people), and the AI can even do most of that themselves too soon).

What's to stop the masses from working for each other?