Comment by wanderingmind

All you need is plot -log(x) for x between 0 and 1 and you will see that log(x) transforms a uniform line into an exponential decay (towards 1). Its being said in a fancy way. This is also the origin of the log likelihood loss function in ML