> Automation reduces costs and prices, creates new job fields, and permits for specialization.
Currently automation has reached the point in which it replaces lower-skilled jobs by higher-skilled jobs. Many people who worked in the former will not be able to work in the latter (mostly because it takes years to specialize), so for them, those jobs are as good as gone.
Do you have any particular examples we could analyze? Yes, you're right, the jobs that robots can do better are gone, but any improvement in efficiency in an industry generally results in lower prices and an increase in job creation around that industry. Again, in the case of the cotton gin, while the gin itself reduced the need for manual labor to pick seeds from cotton, it blew the demand for labor to plant and harvest cotton through the roof, and the improved output of these larger plantations resulted in the creation of shipping ports and explosive growth in the textile industry. In fact, this is widely considered to be a socially negative effect of the cotton gin, since it dramatically increased the demand for labor, which was then filled by slavery. That was obviously not desirable, but the core point I'm trying to make there is that eliminating all those slaves' jobs resulted in a dramatically larger demand for labor in the supply chain.
Just because the old job is gone doesn't mean there isn't a new one to be done, and improvements in efficiency and costs of production result in economic growth that have positive impacts all over related markets.
> Do you have any particular examples we could analyze?
I'd look at every job that is being replaced by a robot or a computer system. I can't give you a particular example in form of "job A replaced by robots, workers can't move anywhere else", because so far we've been very efficient at reallocating labor. Throughout last 200 years, people out-automated in agriculture moved to manufacturing; optimized out from there they are moving to services, but the computer technology makes this sector fair game, and there doesn't seem to be anywhere else to go.
More machines are creating more opportunities for building them, but the manufacturing is already heavily automated, so this is the case of robots building robots. As a human, you can't compete there much.
Just imagine self-driving cars really taking off in few years and replacing most jobs in transportation industry. Where will those people go? What kind of jobs we can imagine they could have without years-long retraining that robots arleady can't do better?
> Just because the old job is gone doesn't mean there isn't a new one to be done
Yes, but my point is that jobs are not made equal, and just because you could do the old one doesn't mean you will be able to do the new one. It is increasingly not the case.
> I can't give you a particular example in form of "job A replaced by robots, workers can't move anywhere else", because so far we've been very efficient at reallocating labor.
Surely it makes sense to be able to cite examples of how we are failing to reallocate labor before complaining that we're not able to reallocate labor.
You're right. I can give you a fair share of personal anecdotes and general social lore related to rising unemployment for STEM-or-finance-educated people. I can point to first principles and proofs by enumeration. I can't provide examples of "elimination of an entire class of human skill and subsequent failure to move labour up the skill ladder" because this transformation is - I believe - underway, not yet finished.
Regarding the "skill class elimination", as far as I have read, basic manufacturing is already done; i.e. it's not completely automated only because robots are still a bit more expensive than low-wage workers. But even in China this seems to be changing (in favour of robots).
Personal anecdotes and social lore make for pretty crappy science. I would love you to point me at first principles and proofs by enumeration.
What I see, in looking at the numbers published by the government, is that the number of jobs has been consistently tracking with labor force since they started measuring it in 1925. If automation were wrecking jobs and making people unemployable, I'd expect that the 20th century - the century in which the human race made more significant technological advances than during the rest of human history combined - to have resulted in a consistent decrease in percentage of the employed workforce. Instead, we see that no such widening has occurred. There is certainly an argument to be made that people are being forced out of factory jobs and into foodservice jobs or whatnot, but I can't see any empirical evidence for the assertion that we're actually innovating ourselves out of the opportunity to work.
Currently automation has reached the point in which it replaces lower-skilled jobs by higher-skilled jobs. Many people who worked in the former will not be able to work in the latter (mostly because it takes years to specialize), so for them, those jobs are as good as gone.