A Paralyzed Man Is Using His Thoughts to Send Texts

Using brain implants and a machine learning algorithm, the system decoded brain signals linked to handwriting, and inserted the paralyzed man into the digital reality of modern communications. BrainGate is a consortium that's forged key advances in the development of brain-computer interfaces (BCIs) in recent years. This includes a complex brain-controlled robotic arm initially revealed in 2012, along with a more recent high-bandwidth wireless BCI for human use. The new project aimed to develop a new handwriting brain-computer interface moved forward under the leadership of Frank Willett, a Stanford research scientist. Another neuroscientist named Krishna Shenoy of the Howard Hughes Medical Institute, along with Henderson (a Stanford neurosurgeon) supervised the development.

Shenoy and his colleagues developed a thought-to-text system capable of besting earlier techniques in 2017 that allowed monkeys to text at 12 words per minute. This success became the foundation for further work on a brain-computer interface that showed up later that year, which offered people with paralysis the ability to type 40 characters (roughly 80 words) per minute. But at present, the "current work goes beyond the 2017 paper by more than doubling the speed of typing by a person with paralysis, and uses an entirely new and different method," wrote Henderson in an email to Gizmodo. This rings true because before this, no one had tried to capture and replicate the mental act of handwriting — and the novel experiment set out to specifically engineer a never-before-seen thought-to-text system.

The new brain-computer interface is promising but extremely invasive

The test subject for the experiment was a man of 65 years who had suffered a serious and paralyzing spinal cord injury a decade earlier. "Two sensors each measuring 4x4 mm, about the size of a baby aspirin, with 100 hair-fine electrodes, were placed in the outer layers of the brain's motor cortex — the area that controls movement on the opposite side of the body," said Henderson in the Gizmodo report. "These electrodes can record signals from about 100 neurons," and then processes the signals via computer "to decode the brain activity associated with writing individual letters."

It's really quite magnificent.

Mid-experiment, the man tried to move his paralyzed hand the way he would when physically writing words. In his mind, he visualized the act of "writing the letters one on top of another with a pen on a yellow legal pad." As he did this, a decoder typed every letter he imagined writing while it was "identified by the neural network," added Henderson. The "greater than" symbol was employed to signify spaces between words — which became the only way to detect the writer's desire or intention to leave a space between letters.

We can't stress enough how moving and frankly touching this is. But while the system could differentiate letters at a roughly 95% accuracy rate — enabling Henderson to reach roughly three-quarters the average typing rate of people over 65 using smartphones — the system has a few obvious limits. Brain surgery is explicitly invasive. Additionally, the machine-learning system isn't universal, and has to adapt to the cognitive nuances of every disparate user, comprising a "very computationally intensive" process that calls for a "specialized high-performance computer or a compute cluster." We're a long way from hearing doctors propose this new brain-computer interface to assist the average paralytic victim, but the potential for a system could someday transform how we write forever.