[alkonaut]

I think in 20 years it's still much easier and cheaper to rebuild a human workplace to be suitable for (very dumb) robots, than it is to use humanoid robots for human tasks.

We already replaced humans for many tasks such as assembly etc. The next step is cognitively challenging but physically easy tasks like driving cars.

A lot of "simple" human tasks such as making sandwiches in a restaurant require fine motor skills, creativity and other very difficult robotic traits. Those jobs will be done by robots long long after we as humans stopped driving cars or building them. Given the budgetary difference between making a sandwich and going into battle, we can be pretty sure sandwiches will come last. If anyone is still around to eat them by then that is.

[kenshaw]

Purpose-built robots for things like "making sandwiches" will always be cheaper than general purpose robots. There really isn't a culinary task today that can't be fully automated (you can look at the frozen food section of a grocery store for examples of almost every cuisine type). The reason this hasn't happened at your corner diner is that manual labor remains much more inexpensive in the short and long term for low volume production (ie, restaurant vs food factory). Additionally, before it becomes cost efficient to build a customized culinary robot for a restaurant, it is still less expensive to design a custom culinary tool for use by the human staff that will net 99% of the efficiency of the purpose-built robot.

Barring "replicators" (a la Star Trek), or labor costs rocketing through the roof (perhaps from something like a "minimum income") I don't expect the general composition of restaurant staff to change in the next hundred or even two hundred years.