I'm wondering what it would be like to keep the metabolism of a brain-dead organism going and controlling it via the nervous system, thereby creating a cyborg (but in reverse as opposed to most sci-fi ones).

Like, imagine an irrecoverable comatose patient whose every speech-related nerve is tapped. Using machine learning, one could create a neural network that maps all muscle movements to sounds and vice versa. The setup could then play any waveform to just about the best of the vocal tract's ability – in short, turning the body into a peak beatboxer. (With multiple such cyborgs or a looper, one could achieve arbitrary precision at recreating sound waves! There can be actual uses too, like letting paralyzed people speak again, but I all can think of is whether the paper of this research ends up using Bad Apple!! or Never Gonna Give You Up as the demo song, and whether there's going to be an acapella band touring with effectively propped-up corpses.)