Do you understand the problem that OpenCL is trying to solve? Literally printing "Hello World" would be a useless example. As others have noted the example is appropriate, as it preforms calculations on dataset, something that can potentially benefit greatly by accessing multiple CPUs.
But my second thought was "Oh my GOD, how many pages of text for a single loop to run?". I'm sure they'll improve the api (with a wrapper layer), but right now OpenCL will only be used by the more demanding of applications.
EDIT: Maybe that sounded wrong. For huge classes of problems this isn't an issue, but to really pull that power down to the everyday developer, an easier wrapper api will be needed.
I can see spreadsheets using this stuff, but not as it stands now. On the other hand a gpgpu raytracer could use opencl as it stands now.