There have been a lot of discussions on HackerNews in the past about EEGs and other potential input devices, but these non-invasive devices lack the required throughput to be used as a full keyboard replacement. EEGs, for example may allow you to transfer on the order of a few bits per second. I have always been interested in the potential of humans to "learn" (with the help of some sort of feedback training) how to control these macroscopic oscillations in a way that gives them a higher bit rate.
Most people conclude that it's not possible and that in-vivo solutions (e.g. invasive BCIs) are required for anything interesting. IIRC, there is research out there that suggests humans, with practice, can learn to "control" the amplitudes of certain large scale oscillations (alpha, theta, etc.). I think this has been done in certain ADHD studies, but it's been a while since I read about it.
The funny thing about this whole discussion is that invasive BCIs are also quite low-bandwidth. Even a Utah array under ideal circumstances recording 50ish neurons is still much worse than a mouse or a keyboard.
People seem to assume that it’s possible to get a higher bandwidth connection by tapping into neural activity directly, but it’s not obvious to me that should be true. After all, our whole nervous system is optimized for controlling our bodies. I’m not saying it’s impossible to build a neural interface that’s better than the one already plugged into your brain (your body), but I am saying nobody has come close yet.
"Rainbows End" by Vernor Vinge has all the younger people (and those grown-ups who are hip) "wearing". Their clothes are (in some unspecified way) I/O devices, so that they learn from quite young how to communicate and control things through imperceptible movements. Teaching her grandfather how to "wear" is part of how the main protagonist gets caught up in the events of the novel. The novel is pretty vague about how much bandwidth is involved, but the convenience is the thing - you are wearing clothes all the time (or at least during waking hours if you prefer to sleep naked) and so if you've learned it then replying "Yes!" to a friend's question about whether you want them to save you some apple pie becomes no harder than nodding is, except it works at a distance.
Most people conclude that it's not possible and that in-vivo solutions (e.g. invasive BCIs) are required for anything interesting. IIRC, there is research out there that suggests humans, with practice, can learn to "control" the amplitudes of certain large scale oscillations (alpha, theta, etc.). I think this has been done in certain ADHD studies, but it's been a while since I read about it.