recognition of sub-vocal communication eh....
Will it translate teen "ugh" and "meeeer" sounds as they get up at 2pm?
Will it be used to record all the other things you think about like a sort of non-verbal thought reader? Isn't it bad enough already that you have to shout at a grubby plastic dome to call down the shopping gods because you can't be arsed to get out of your armchair? No? you would rather go round your house looking like you are becoming an early adopter of the BORG having your under-the-breath mutterings captured by the hive?
No now you will voluntarily submit to having all those "sub-vocal" communications recorded and potentially replayed where you used to say under your breath "what a ****wit ****** is!" while you were in the kitchen, now it can be innocently blurted out by the machines for you.
Who on earth wants this technology? Given that most AI's think a Scottish accent is a fatal speech impediment I don't hold out much confidence for the people who have genuine and often "non-standard" challenges with speech or worse ... grammar. Good luck Yoda with your "four hundred toilet rolls there will be master Luke".
And as for using parts of the car as a speaker diaphragm. Although technically possible, has anyone thought this through? Anything touching the dashboard will start to move, keys, coins, the pile of general matter in the glovebox will start to bounce to your tunez. Even for relatively modestly powered systems that don't require underlighting on your car can bounce coins 4-5 centimetres if placed on the diaphragm. think about it just for a moment. You are approaching the toll bridge, fumble around for the change, put it on the dash and it's fired back at you and all over the car like confetti while you are doing 70 in the wrong lane.
Try door panels at least, they are marginally more likely to make a stereo output without doing more than shaking the can of lemonade you put in the door pocket.
Just because these things can be done, does not make them sensible or useful.