MP3 and AAC are two encoding schemes for audio based on the properties of human perception. One such property is that high levels of low frequency sound makes higher frequency sounds with lower levels inaudible. This is one example of a masking effect and as a result of that, fewer bits can be used to represent the audio signal.
When a PCM signal is converted to MP3 or AAC, supposedly inaudible parts in the original signal are removed and sometimes we refer to this as destructive or lossy encoding. An audio signal encoded as MP3 or AAC reduces the amount of data with about 90% and cannot be restored to the original PCM signal.