A friend of mine mentioned that he had read somewhere that Autotune is actually not a vocoder. Is this true? If so, what is the difference?
A vocoder is a specific effect: you pass the signal through a multi-band filter, each band to an envelope follower, and then the output from the envelope follower is used to perform filter control and additional synthesis. "It turns an analog audio signal in to something that can control a synthesis engine" is a simplified way of saying that.
Where as "autotune" is really more a process and it may employ a vocoder to that end. But not all autotune algorithms use vocoder-based techniques to do their thing. Example: Melodyne.
Your friend is right; Autotune is not a vocoder.
A vocoder works by taking a musical sound input (typically a synthesizer tone with a lot of harmonics), and passing it through a multi-band dynamic EQ filter, where the setting of each EQ frequency band is based on the relative spectral volume of sonic content in your voice within each frequency band at any given moment in time. In other words, if you say "aaaah", that sound has a predominant frequency; it is that frequency that is emphasized in the synthesizer's output.
Autotune works by pitch-shifting the audio signal to tones that you specify with a musical keyboard, while holding the formant information steady. This prevents the "mickey-mouse" or "kidnapper" effect that can occur with an ordinary pitch shifter.