Ok, yes. It is very interesting, but such systems rely on the nonlinearity of air at high pressure. Therefore, they don't produce sound in the audio band unless the ultrasonic SPLs are sufficiently high.
To achieve the requisite pressure levels at which this nonlinearity begins to produce audibly loud difference frequencies in the audio band, an ultrasonic transducer must produce well over 100dB SPL at 1m.
This graph comes from an
AES paper on ultrasonic transducers and shows the amplitude response of an ultrasonic transducer:
View attachment 22285
As you can see, the required ultrasonic SPLs are extremely high compared to the audio-band SPLs that are generated (an ultrasonic SPL of
[email protected] produces audio band SPLs of around 50dB@4m).
I can't imagine a real-life situation where (a) the amplifier is producing enough output at ultrasonic frequencies and (b) the loudspeakers are reproducing those frequencies at sufficiently high SPLs to cause the air to become sufficiently nonlinear for audible difference frequencies to be generated.
This is an interesting track to go down though, so if I've missed something in your reasoning I'm happy to keep thinking about it