There is evidence that we are now at the stage where we can identify through data some decisions, previously based on human intuition, that we humans are more likely to get wrong. In some cases we can produce decision support systems and even completely automated systems that are more accurate by far.

Of course, all the above is based on problems that we can codify in some way alongside a set of data which can be mined for enough information for this purpose.

As for the reference to machine intelligence, nowadays I wonder if it is not the artificial intelligent abilities but the volume of local data storage required that will be the most impressive feat achieved...unless they run it in the cloud 😉