24bit. God Damn it.
Right. Digital audio is described with two sets of numbers:
The SAMPLE RATE, which is a measure of the number of times an analogue wave form is measured per second, i.e frequency. This is your 44.1, 48, 96 KHz number.
The BIT DEPTH. (NOT, repeat NOT 'Bit Rate'.) This is your 16 bit or 24 bit number, and details the amplitude of the analogue waveform at the point where it was sampled. A 16 bit sample has 65,536 possible values (2^16), a 24 bit sample has 16,777,216 (2^24) possible values.
Because Bit Depth is a measure of a wave's amplitude, it can be used as an indicator of a sound's loudness, and this is where using 24 bit makes sense. All audio systems have two boundaries: the lower being the point where a signal becomes indistinguishable from background hiss (the noise floor) and the upper, reached when the signal level becomes so high that the system becomes unable to process it, causing distortion ('clipping', as the peaks and troughs of the waveform are clipped.)
A greater bit depth offers a wider Dynamic Range (more 'headroom') when digitally recording a sound - less background hiss (i.e. a lower noise floor) means that you can record extremely quiet sounds clearly; more numbers (2^24 at 24 bit compared to 2^16 at 16 bit) means you can record much louder sounds without the signal clipping.
In the olden days, even the best studio equipment was noisy, meaning that there was much less headroom available when recording. Studio engineers spent a lot of time managing input levels so that a recording would cleanly capture a performance without the quietest parts being lost in background noise or the loudest parts being clipped. 24 bit recording means that studio engineers don't have to do that any more, as the format offers them enough room to cleanly capture a performance, and if a guitarist decides to turn an amp up to 11, the engineer doesn't particularly have to worry about the signal clipping.
If not more importantly, considering that the vast majority of music made today is created almost entirely through software (including those beloved classic album remasters), the extra headroom that 24 bit offers means that a producer is able to pile on the signal processing effects without being forced to degrade the signal quality of the piece to do so.
24 bit audio is now indispensible in audio production, and people are already moving up to 32 bit floating point workflows.
What about for audio playback? As we know, 24 bit gives us a wider dynamic range - the difference between the quietest sound we can detect and the loudest we can cleanly process. Say you had a 24 bit audio file that consisted of a tone just distinguishable above the noise floor, which then increased to the point where it clips, and that you played this through a pair of capable speakers. By the time the playback finished, you would be in agony, and your hearing would be permanently damaged.
No piece of audio for playback would ever be put out that uses the entire dynamic range 24 bit offers. It would permanently damage customers' audio equipment and their hearing. All a 24 bit file gives you that is missing from a 16 bit recording is masses of dead, empty headroom in which there is nothing to hear, and that will never, ever get used.
Anyone who tells you that listening to a track rendered as a 24 bit file is 'better' than listening to the same audio rendered as 16 bit is a liar or a fool, and either selling you something or trying to justify something expensive they've bought.