Take for instance, this article on forbes. (Yes, I know. Have noscript ready.)
It is rather short on details, but makes the salient point about alpha-go creating a wholly original move, through machine "creativity."
It is not really that big of a change in tactics required to train similar AIs to do, for instance, market trading strateges-- which has already caused a mass exodus of humans from stock trade floors.
Further refinements of such methods could eventually lead to radical shifts in how things like aircraft are designed, or computer chips are laid out. Skilled human minds that rely on intuition can be replaced with purely logically founded iterative software agents, with billions of prior tested design strategies to work with behind them.
To get an idea of how quickly the fallout of a major paradigm shift can rattle through an economy, take a look at this Atlantic article from last year.
It also has the following gem in it:
In 2013, Oxford University researchers forecast that machines might be able to perform half of all U.S. jobs in the next two decades.
which is on par with my initial statement. Since it was called out specifically, let's see if we can find it.
And here it is. (warning, pdf)
Like I said, the linked story is not the only group that has looked at this issue. I am vaguely recalling at least 2 others that have reached similar conclusions to Oxford and PwC, and who have given a rough estimate of hitting the tipping point within the next 20 years, give or take.
I have no reason to argue against people better trained in trend analysis than myself.