neurobiologist, Brown University, and cofounder of Cyberkinetics
He hopes his research in fundamental neuroscience will lead to devices that can help people who have lost the ability to move their limbs.
On whether creating a brain-machine interface will open the door to mind control:
“We do that all the time already. Advertising is mind control. Even pharmaceutical agents are a form of mind control. When people have behaviors that deviate extremely far from the norm, they are given medications that bring their mind back into the realm of behavior that we call normal. So we do it now. If a child were to have a seizure and became unconscious because of the seizure, and we controlled his mind so that he didn’t have seizures, that would be a wonderful thing. We want to do that.”
RICHARD ANDERSEN, professor of neuroscience, Caltech
COREY GOODMAN, CEO, Renovis
TOMASO POGGIO, professor of brain and cognitive sciences, MIT
WISE YOUNG, director, W. M. Keck Center for Collaborative Neuroscience, Rutgers University
John Donoghue is building a brain decoder that could transform the lives of people paralyzed by injury or disease. Those who have lost the ability to move their limbs often have perfectly intact brains, so Donoghue hopes to implant a chip that can monitor their brain activity and convert their intentions into computer commands. In its current version, the chip’s 100 hair-thin electrodes listen to neurons firing in an area that controls arm movement and translate the activity into electronic signals. A program then decodes the brain signals into commands that direct a cursor on a computer screen. Donoghue hopes the chip can eventually control appliances or even robotic limbs. “We’re effectively rewiring the nervous system—not biologically but with real wires,” says Donoghue. So far, more than 20 monkeys have been equipped with the implanted chip, and four of them have successfully willed a cursor to follow a moving target. Now Cyberkinetics, the company Donoghue cofounded to develop the device they call BrainGate, is preparing to test it in five paralyzed humans. “People with these kinds of injuries are perfectly capable of leading full and productive lives,” says Donoghue. “They just can’t get their signals out.”
Give me the big picture first. How did the idea originate and what problem were you looking to solve?
D: I’ve had a long-standing basic science research program that has really been directed at how the brain computes information. In a simpler sense, how do you turn thoughts into action? The way you get at the fundamental activity is by recording with electrodes in the brain. And since it’s a procedure that requires you to introduce that electrode, you have to use a monkey or an experimental animal. A monkey has a motor cortex like ours, and its behaviors are a lot like ours, so we use it as a model.
So the basic science took us to a point to think about why we should develop a neuromotor prosthesis. Some people talk about brain-computer interfaces, brain-machine interfaces. Certainly an automobile is a brain-machine interface. You use your brain to drive a car. But a neuromotor prosthesis is specifically something that takes signals from the brains of people who can’t get those signals out and then connects them either to a device—like a computer or a robot—or even to that person’s own muscles.
How do you translate the firing of neurons into movement? How do you crack the brain’s code?
D: The monkeys play video games with a mouselike thing. We put something up on the screen, and we build this encoded model by having the monkey track a dot by moving its hand around. What we’re doing is having the monkey move in various positions and in various directions. And we ask: Is there information about which way? How far? It turns out there’s tons of information about the speed, velocity, and position of the hand in space. So we have the monkey just track the dot for a while, and we make an observation. We observe firing, and we observe hand motion. We have lots of neurons firing, but we just take the spike counts. We just take the spikes, and we say what the relationship is between this spike count and the current hand position, and then we build a model. It’s a multivariable linear regression. And we find a filter coefficient.
The idea is that you have to have a number of neurons. And the answer could have been millions because clearly when you do this, there probably are millions and millions of neurons acting. But the answer is, starting with about six or something, we can actually get some meaningful information.
And then once we build that model, we can now build a decoder. Once we have this little filter coefficient, we can now read out what each neuron is doing, multiply it by its appropriate filter coefficient, and then say what it tells us. Now we can use that information. This has been confirmed by a bunch of labs.
How do you see the boundaries between humans and machines changing?
D: The boundaries have changed a lot. When you move your hand, that’s an interface out of your brain. Driving cars, flying airplanes—these things become extensions of our nervous system and our body and our ability to interact with the world. So we already have changed. Look at how it changed from 1800 to now. There’s been an extraordinary change in the way we have now integrated machines into our body. If you look around at cars, there’s a little homunculus inside each one making it go. So that’s a machine interface.
But this is another step in which we’re going directly from the brain. Every time we think about brains, it has a magical or almost mystical component because many of us think the brain is at our essence. Most of our thinking about actively coupling to the brain is thinking about how we could help restore lost function.
So as we understand brain processes and we realize that they’re encoded in spikes, we should be able to pick up signals and interpret them. We’re trying to make sense out of what that pattern of activity is. Now we know when you see an image we can read it out and say that you are now seeing a red vase sitting on the table. Or we could go to the motor cortex and see that you’re just about to move and are moving your hand to that location.
Every time scientists record from the brain, they are attempting to understand what that part of the brain is processing, and we rarely have much insight into whether it is conscious or not.
If the brain code can be cracked, does that mean that my thoughts could eventually be read?
D: Yes, if you believe that spiking activity and populations of cells are the essence of brain activity—which they might not be. There might be more to it than that. But if that’s it, and you could pick up all those by millions of electrodes, then in theory you should be able to reconstruct everything that’s going on in your head and see your dreams and know your thoughts.