Important note: they were using monkeys and didn't take advantage of the speedups that humans would use, like adaptive letter-guessing/autocomplete:
>“Also understand that we’re not using auto completion here like your smartphone does where it guesses your words for you,”
That would reduce the "distance" you have to move the cursor on average.
(Several years ago I had the idea to combine the accessibility software Dasher with the "Neural impulse actuator", which tranduces brain waves, but never went through with it, though I bought the tech.)
Well they're not using autocomplete, but the software sitting between the electrodes and the cursor is a black box as far as the article goes. If the software is trained or calibrated, what's to say that it isn't picking up some information from the structure of English.
Also depending the setup of the onscreen keyboard, it too could have error reducing aspects built in.
Also it's unclear whether the 12 wpm is actual wpm with back spacing and correction, or whether it's an 'adjusted' wpm based on the error rate.
This isn't to say that it's not an impressive result regardless. Just that I'm more skeptical of the 12 wpm.
>“Also understand that we’re not using auto completion here like your smartphone does where it guesses your words for you,”
That would reduce the "distance" you have to move the cursor on average.
(Several years ago I had the idea to combine the accessibility software Dasher with the "Neural impulse actuator", which tranduces brain waves, but never went through with it, though I bought the tech.)