“To be or not to be. That is the question.” That is also the text that Monkey J typed out using a brain implant to control a computer cursor.
To be clear, the monkey didn’t know it was copying Shakespeare, and it had no deep thoughts about Hamlet’s famous monologue. Monkey J and its colleague, Monkey L, were both trained to use their neural implants to move a cursor over a computer screen, hitting circles as they turned green. Stanford University researchers placed letters on those targets to simulate the typing task. So to tap out the line from Hamlet, first the “T” circle was illuminated, then the “O,” and so on.
What was the point of this exercise? Was it simply an excuse to let journalists trot out the “infinite monkey theorem”? Because here we go: This probability theorem states that if you give a monkey a typewriter and infinite time, its random keystrokes will eventually produce the complete works of Shakespeare. (If you’d rather goof off than read about science, please enjoy these excellent cartoons.)
No, the bioengineers had a more practical motivation. By simulating this typing task, they demonstrated that their brain-computer interface could greatly benefit people who can’t communicate otherwise. That category includes people in the late stages of amyotrophic lateral sclerosis (ALS), also known as Lou Gehrig’s disease, which leaves the mind intact but gradually paralyzes the body, including the mouth and other face muscles.
This experiment set a new record for typing-by-mind, with one monkey tapping out 12 words per minute. “To our knowledge, this is the highest communication level ever achieved,” says Paul Nuyujukian, a researcher at Stanford’s Neural Prosthetics Translational Lab. Nuyujukian is coauthor of the paper describing this research, published today in Proceedings of the IEEE.
Here’s how it works: The monkeys had tiny electrode arrays implanted in their brains, specifically in the part of the motor cortex that controls arm movements. Those electrodes measured the electrical activity of neurons while the monkeys were trained at the cursor-control task, first moving their actual arms while cameras carefully tracked the movements. Machine-learning algorithms found patterns in the stream of data, and translated those patterns into a monkey’s intent to move the cursor left, right, up, and down.
Then the monkeys were set to the task of moving the cursor with their minds alone. (They could still make the movements in the air with their arms, but they weren’t being tracked.) The computer picked up the same patterns in the brain data, and the cursor moved smoothly from target to target.
The prior record for brain typing, set by human patients with ALS last year, was 6 words per minute. That experiment was done by a larger group of researchers, including Nuyujukian, who are part of the BrainGate consortium. The big improvement from that prior study came from software; the system used by the monkeys employed two smart algorithms in tandem, one to decode the cursor’s movement, the other to decode the monkey’s intent to click on a letter.
And there’s plenty more room for improvement, Nuyujukian tells IEEE Spectrum in an interview. The monkeys used an interface in which they picked one letter at a time, he says, but future interface could borrow some tricks from smartphones. “I can imagine a smart interface that’s auto-completing the words,” he says. “Google and Apple have done a lot of work on how to maximize the input from our very inaccurate thumbs. We can leverage a lot of that.”
Human trials of the technology are already underway. In one experiment, which the researchers presented at a conference last year, a woman with ALS used the cursor-control system with an Android tablet, using it to browse the web and write emails. Such human studies make it clear that the researchers aren’t just trying to generate a wave of jokes about Shakespearean monkeys. They’re trying to give paralyzed people their autonomy and the ability to speak their minds.
But the researchers aren’t above having a little fun themselves. Nuyujukian divulged some of the early phrases they had the monkeys type out as they tested the system: “A banana, a banana, my kingdom for a banana!,” and “A banana by any other name would smell as sweet.”
Eliza Strickland is a senior editor at IEEE Spectrum, where she covers AI, biomedical engineering, and other topics. She holds a master’s degree in journalism from Columbia University.