Sometimes mystical titles for mundane realities are useful. We say "machine learning" instead of "computational statistics" not just because "computational statistics" sounds boring, but because "learning" is the effect we want. Framing it that way makes it easier to see the problems it can solve, because some problems that aren't obviously statistical still make you say "I wish it could just learn to do the right thing here." Likewise, it probably isn't useful for your baby's diapers to talk to your microwave, but a pithy phrase to keep "stuff that isn't computers might want to be networked" in the back of your head can be handy when you come across a pair of things that really do want to talk to each other. The cost is that people who don't know what they're doing can throw the phrase around and sound like they do. I don't know of any jargon that doesn't have that problem.
I agree to a point. "Machine learning" is a way of explaining computational statistics to a group of people that don't understand computation and don't understand statistics. "The Internet of Things" is a way of selling telemetry to people who certainly understand instrumentation and sensors but are suspect of your desire to profit by putting them where they may not belong. 'cuz that's the other aspect of "The Internet of Things" - it implies that machines can't talk to each other unless they've got an IP address. This is true if you want to tie them into 4square or Twitter or Facebook or Google... but even a diaper doesn't need a twitter feed to tell you it needs changing.