I work in industrial automation, so I do PLC programming, robot programming, control system integration, etc. I've been doing it over 15 years now. For the first 10 years I bought the whole "luddite" argument, and figured that automation only displaced people to other, ultimately higher paying jobs.
However, in recent years I've really started to worry. Imagine the person who is barely functional: they can follow instructions but you have to repeat yourself a whole bunch of times, and even then they still make lots of mistakes. My experience tells me this is around 30% of the workforce, at least. Back when everyone was in agriculture, these people couldn't really do too much damage, and if they were strong, they were useful. The magic of the industrial revolution was that we were able to both magnify the strength of everyone, *and* reduce the chance of making errors by (a) breaking things down into tiny tasks so people only had a very very simple thing to do (tighten nut A on bolt B all day long), and (b) designing things such that they couldn't assembled incorrectly (the modern term is poka yoke). This "lower" 30% of the workforce became very productive, and they joined labor unions and owned big houses and boats. They retired with nice company pensions. Their kids got much better educations than they did.
So, if you look at the things that these people made lots of money doing (something extremely simple, repetitive, and designed to be error-proof), then that's exactly what is simple enough to automate with a robot. We recently had a job that was taking 3 operators to do and produced parts at the rate of about 3 parts per minute, and they couldn't meet the production numbers even with 2 shifts (total 6 people). We replaced all 6 of those people with a single robot, and we're up to about 8 parts per minute so we probably only need to run about 1 shift.
The difference is that this new robot assembly cell requires a semi-skilled operator to run it. They need decent troubleshooting ability, with a bit of mechanical knowledge and decent computer skills (not programming, but basic stuff like navigating screens, understanding slightly more abstract concepts, etc.). They need to be able to look at the robot gripper and determine if anything's worn and needs replacement. We happen to have someone who's almost overskilled for the position. So we keep shuffling those other 6 people around in the plant, trying to find something for them to do, and almost always realizing that whatever they're doing could be automated. Plus, I really need to stress that these aren't people with decent troubleshooting skills, computer skills, etc. Any process we put them in requires us to remove all human decision making, because we can't tolerate errors (or they're very expensive).
My point is that unskilled laborers are a hassle to employ. We have a hard time thinking up things for them to do, and we'd love to find something because, well, they're so cheap! (And we already have $10/hr+ minimum wage here.) But so are robots. It used to be that a bare robot (uninstalled) cost $50,000. Integration costs might push that to $125,000 or $150,000. That really limited the choices... you pretty much had to eliminate one operator for 3 shifts to make it a valid investment. Now those costs are almost cut in half. The robots are well under $30,000 and integration is getting cheaper, plus we're just getting better at it.
As we transition into this "new economy" where there are no unskilled manufacturing jobs left, I really don't know where these people are going to find employment. I don't just see it happening in manufacturing either. I'm pretty sure that truck drivers and taxi drivers will be the first to get automated by the kind of auto-drive technology that Google's working on. We're already seeing automated forklift trucks in factories. I just don't know.