Thinking about moving machines

• Article
• Comments ()
• Click-2-Listen

By

As a kid I read fantastic accounts of space travelers controlling their ships by means of special mind-reading helmets. Now we're actually doing it -- sort of.

The idea of mental control of machines is simple enough. Suppose that you are an amputee. When you think about different things, such as moving your (missing) hand, changes take place in your brain -- blood flow, electrical activity, glucose metabolism, what have you. In principle, sensors can detect these changes, and use them to control an artificial hand.

Unfortunately, while the idea is simple, doing it isn't, which is why we are only now getting there.

But we're doing it. At Washington University in St. Louis, for example, researchers have managed to have a boy of 14 play Space Invaders completely by mind control. The game is the sort in which a laser cannon slides back and forth across the bottom of the screen and fires at bad guys coming down from space.

In this case, legally and with everybody's approval, the boy, an epileptic, had a grid of tiny electrodes implanted through the skull to the cortex of his brain. These were connected to the computer. He quickly learned to control the laser cannon by thought alone.

According to the Web site of the university, after implanting the grid, "They then asked the boy to do various motor and speech tasks, moving his hands various ways, talking, and imagining."

The team could see from the data which parts of the brain and what brain signals correlate to these movements. They then asked the boy to play a simple, two-dimensional Space Invaders game by actually moving his tongue and hand.

He was then asked to imagine the same movements, but not to actually perform them, according to the Web site. He then played without using his hands. Bingo.

Most of us do not want grids implanted in our heads. However, there are many ways of detecting the activity of the brain noninvasively. There is positron emission tomography, functional magnetic resonance imaging, and so on.

Today these run from phenomenally clunky to merely impractical for controlling machines. Their resolution isn't what one would like: They can't all measure changes taking place in a really small region of the brain. However, the problem is largely one of measuring and localizing small signals, and this is the kind of thing that technology does increasingly well.

In particular, Hitachi is experimenting with "optical topography," in which infrared light passes harmlessly through the skull and, reflected to sensors, allows measurement of the activity of hemoglobin. This is cute because it doesn't require implanted electrodes or bulky apparatus.

Worth noting is that mental control of a wheelchair does not exactly involve reading the patient's mind, though the line isn't clear. When the patient wants his hand to move, a certain part of his brain "lights up," You can detect this and make a mechanical hand move. When the patient thinks, "I need to see George tomorrow," there is not now, and maybe never will be, a way of knowing what he is thinking.

However, emotions cause specific changes in the brain and (in principle) these can be detected. I have seen security people speculating that it might be possible to use emotional changes in the brain to catch terrorists at airports.

But for controlling a quadriplegic's wheelchair or computer? Mental control looks to be closing in fast on practicality. Neat stuff.