Elementary Idea / Basic concept behind Machine Learning

Consider two variables x and y . We have a requirement to find a mapping between x and  y . Notice that this notion is not at all new to us. In mathematics, on a surface level ( not getting into crude details ) we call such a mapping as a function. Given an input variable x , function f  , is a mapping by which we can return an output variable y .

y = f(x)

In real world problems, sometimes it is not very straightforward to define any function ( in modern computer science sense, an algorithm ) that can map some input variable to some output variable to a good level of precision in some definite well defined steps. 

Here, we use machine learning to find such a mapping.

Generally, x is referred to as a set of features dependent on the problem domain and description, and y is referred to as the target label or value

In loose terms, ML is a set of concepts, algortihms and tools which we use to find such a mapping, in other words ML algorithm is an algorithm to find another algorithm (mapping). 

Note: Although it may appear as machine learning is the cure all panacea, but we must undertand everything behind the disguising layer of Machine Learning is still pure algorithms coupled with mathematical tools like Algebra, Statistics, Probability and Calculus.