I like to compare machine translation to driving a car. We all know how convenient and useful it is to travel to various places with a car. And many of us can and have a license to drive a car. Driving a car is a skill that has to be learned. When one has learned to drive a car, it is easy and useful to use the skill. But driving has some crucial limitations. Even when one can drive a car well, it is not possible to drive with the car anywhere, for example in the forest or from Europe to Australia or to the Moon. One has to stay on the road and on the solid soil. The situation is very similar with machine translation. One must know when and how machine translation can be used.
Machine translation is not a turnkey solution to all possible translation problems. One must know when machine translation can be used and especially how it should be used. For example, if you expect a computer to translate literature perfectly from one language to another you will surely be disappointed. But if you want to communicate simple matters to another language you might be positively surprised.
I would like to emphasize that machine translation is potentially a very effective tool that must be used correctly. Like a car, you must first learn how to drive it and where you can drive with it. Same applies to machine translation. Luckily, using machine translation is far, far easier to learn than driving a car.