Two key propositions seem to be:

1. The world is at risk from a superintelligence-gone-wrong;

2. The SIAI can help to do something about that.

Both propositions seem debatable. For the first point, certainly some scenarios are better than others - but the superintelligence causing widespread havoc by turning on its creators hypothesises substantial levels of incompetence, followed up by a complete failure of the surrounding advanced man-machine infrastructure to deal with the problem. Most humans may well have more to fear from a superintelligence-gone-right, but in dubious hands.