I'm really glad Intel is embracing OpenCL. I've learned enough SIMD vector instruction set extensions in my years. I'd be quite content never to learn either AVX or how to program for Xeon PHI's 512-bit vector engine.
As for the Xeon Phi, the less I have to think about how to communicate, coordinate, and schedule its many cores, the better. If your problem maps well to OpenCL, it greatly simplifies most of those issues.