The more people lose their jobs to robots, the less people will be able to buy what the robots are making.
Bernard:
So let's say we get a huge breakthrough, and robots cause 20% unemployment. Huge. i.e. economic recession/depression.
What do companies do at such times(like we've seen in the last recession) ? seek where to save money. If robots offer savings, and companies have the right financing(or robots are offered on-demand), companies will replace even more employees with robots(even though the global effect will be bad). To a certain extent they have no choice, because if they don't do that lose money and market share and die.
As for the article and the claim that Gordon doesn't see robots in any industry(besides warehouses and manufacturing):
* retail is becoming warehouses(E-commerce) and automated parcel lockers and self-driving cars.
* self-driving transportation
* Broad group has built 30 high rise pre-fab buildings which we're manufactured in a plant, hence could use robots for most of the work. But it's a slow industry so it will take a long time.
* Munchery is working on tranforming restaurants into large commercial kitchens(factory like), with labor costs per meal declining ~5x + already in some restaurant chains thousands of tablets replace some waiters's responsibilities
* We're starting to see automated teaching in online college courses, including complex stuff like grading.
etc, etc
In short: Gordon is a known skeptic because he doesn't see productivity growing in the economic numbers. But if you look into the details , in some sectors, like manufacturing and retail etc - producitivity does grow nicely, but than we silly humans decide to take all our extra money because of that , and waste it on low productivity industries (mainly getting better service, better experiences , all which take lots of labor) and than on average the numbers look bad.
This is from:
The productivity paradox: why we're getting more innovation but less growth - Vox - a very good read about the subject.
The only question is: what could and what will change that pattern towards full automation ? or will we continue this way ?