If backflipping robots do the easier work, then what is left for humans is the more complex well paying work. When you remove the lower paying jobs from the average calculation, the average will rise. Indeed the article even states that at 89%, male prime-age [labour market] participation remains close to a record low. The article also has an opinion quote stating that some of the recent wage gains are misleading, because they have occurred in industries, such as textile manufacturing, in which employment continues to fall.
That's the point, it will take A LOT longer for the robots to replace these menial jobs. (And they are not all as low-paying.)
I remember articles from one or two decades ago about the attempts to automate a very dangerous and not even low-paying job of window washing in the high rise buildings. It is a real problem, it does not require HAL-9000 grade AI that can analyse Shakespeare, it is economically viable, and neither then nor now it is completely automated. I see that in Australia, window washing jobs are paid as high as low to midrange software development.
I am not even going to mention more complex tasks like a work of a plumber, a gardener, a janitor. Is the work of a janitor the more lucrative occupation?
It takes very long to get from the first demos to a prototype working most of the time, and even longer from the prototype to a system that can be relied upon in a mission critical environment (a category that most applications of robotics fall into). The Silicon Valley prophets got it all backwards.