New brain-computer interface writes up to 90 characters per minute with your thoughts

Originally published at: New brain-computer interface writes up to 90 characters per minute with your thoughts | Boing Boing


well, pandemics, now brain – computer interfaces. When do we get to choose between World War Z and Robot–O–Pocalypse ?


After reading the title, I wondered if there would be separate windows for all the thoughts people tend to have at the same time. :nerd_face: One letter at a time, while a huge improvement over what they had before, is still far away from what we see in fiction. Baby steps…




Ussing tis is nownow. Tis tecology is is is amaazazing thisthis tecolog yyyyy amaazingbly not no no not probhavelems has.

Soglad wirs inmy bran having no w tea wirs arr sharp gone inn bud nownow II likening wirs in in minebrane.


And it says “up to” 90 characters per minute, so I’m gonna guess it took you 4 hours to write that comment :wink:


Usually, after seeing something like this, someone predicts that ten years from now, everyone will be using brain implants for communication.

I wonder if it will be possible within my lifetime to send text messages brain to brain.

I also wonder if people though of something nice to say to each other by then.

Well, i won’t be a person to use my mind to communicate with the world when my body fails: I guess Alzheimer’s will get me first.

yeah - but what if you’ve got really shitty handwriting?


I couldn’t figure out from the Ars Technica article why they would study writing over typing. (They do talk about a typing-based system, but not one analogous to the writing-based one.)

The patient studied here has 2, 4 mm wide, 96-electrode arrays in his premotor cortex which is laid out somatotopically. So they are recording neural activity specifically related to planning hand and finger movements. In theory (not feasible with current tech) a lot of arrays across different brain areas could simultaneously decode activity related to all sorts of behaviors.

This paper is describing data from one single patient who had a severe spinal cord injury. These arrays require opening up the skull and sticking metal pins into the brain; an invasive surgery with major risks (your immune system can’t treat infections in the brain) that is only done in cases with serious medical need. The arrays themselves damage the surrounding brain tissue over time. So you can go ahead and forget any thoughts of any such elective implants in our lifetime.


Yeah, yeah, I wrote that. Can you override like GPS gives you bullshit directions and you know a much better way?

This is cooler, more interesting, and far more practical than whatever Elon Musk’s Neuralink is doing with animals.

As a neuroscientist (for serious), this kind of statement is incredibly frustrating. Regardless of what you think of Elon Musk (& his eccentricities/peculiarities), none of the work at Neuralink would be possible if not for pioneering research done (in non-human primates) with the very same “Utah arrays” used in this handwriting decoding research.

The only reason the Neuralink video was able to go viral is because the technology is wireless, so there were no giant trans-cranial data interfaces, like the pair depicted on the human subjects head in the handwriting decoding paper.

Similar advancements are being made on the electrode end; resulting in far better quality recordings, producing functional data for much longer durations.

When people pretend the astonishing medical advancements in front of them aren’t built on the backs & sacrifices of safe/responsible/necessary animal research, they diminish our ability to produce the next life saving/changing advancement.
(…not to mention throwing dirt in the face of the hardworking & underpaid workforce of academic research)


Great question.

The act of writing involves much more dexterous & coordinated movement of the fingers than a simple key stroke (hunt & peck typers excluded).

Imagine each muscle group of your fingers making a different musical tone. …imagine string instruments, if you’re feeling classy …or 1990s Casio synth fart tones, if you’re feeling really classy.

Now do your best air-signature, as if you were John Hancock sighing the hell outta the Declaration of Independence. If you’re fully playing along, you should have ‘heard’ a series of highly coordinated tonal sweeps & flourishes as you wrote. Believe it or not, this is exactly what the neurons in your pre-motor cortex are doing when you visualize those finger movements (“fine-motor planning”); a temporally coordinated bio-electric symphony!

Now do the same imaginary exercise while typing your name. …probably something more like a few short, punchy bursts that sound more like your 1990s Nokia ringtone than the Bossanova on your old Casio.

With the “correlated symphony” produced by handwriting, the BMI neural classifier has a much better chance at discriminating the neural patterns of a letter d vs e than it would using the subtle differences of hitting the [d] or [e] key on the keyboard.


Honestly, while there are some incredible, amazing things going on with this…

I think our best hope is more resolution in hat based sensors or sensors that are located under the skin but outside the skull. Opening the skull and placing wires in the brain will always be really, really risky. It feels like putting wires and sensors under the scalp is a lot less risky, and wearing a hat would be least risky of all.

I do think that it is OK to do stupid, silly stuff with the tech as long as it is non-invasive because it helps push the tech forward to where it can be really useful: helping differently abled people interact with other people more effectively and increase their quality of life by offering new assistance devices if they choose to use them.

(Note: I am not by any means an expert in this.)

1 Like

Reminds me of the palm pilot. It was the first widely available means of entering text without a keyboard. But eventually we all moved on to something without as much complication.

There is nothing inherent in writing strokes that makes it ideal for this sort of recognition. Surely other complicated but stable thought patterns would be easier to recognize. Screw skeugraphics.

Still a bit thin, considering research has been going on since the 1950ies…

1 Like

I suppose, but I suspect the point was that the study participant already knew how to write by hand from before their hand was paralyzed, so there was no need for them to learn to produce any new thought patterns. Moving the burden from the user to the software

1 Like

This topic was automatically closed after 5 days. New replies are no longer allowed.