When mathematicians are obsolete, computer scientists will be obsolete. When computer scientists are obsolete, we have by definition entered the Singularity.
So my take is that he's not necessarily wrong, but if he's right then lack of employment for mathematicians won't be our biggest concern.
I actually think Watson would be far more useful to lawyers than to mathematicians. With a little tweaking, he could probably even identify all the non sequiturs in your comment.
Imagine a future where, in addition to checking your spelling and grammar, your computer could check your logic too!
To me, singularity is hypothetical nonsense that shows a clear misunderstanding of engineering problems. Computer systems have not become much more sophisticated in the past 40 years, we have just made them bigger. We keep running into the same mathematical and physical limitations that will not suddenly disappear when machines can improve themselves. In fact, even if such a machine was possible it probably would not get much further than a few generations after which the limitations have become orders of magnitude more complicated compared to the sophistication of the machine. (welcome to engineering)
I'm not saying that the singularity isn't some kind of "nerd rapture", but it may a bit disingenuous to say computer systems haven't gotten much more sophisticated. Sure, we're still using Von Neumann architecture and transistors, but software has gotten massively more interesting in several different directions.
So my take is that he's not necessarily wrong, but if he's right then lack of employment for mathematicians won't be our biggest concern.