How Facebook's Telepathic Texting Is Supposed to Work

Published June 11, 2017
Dow Jones Newswires

You might've heard that Facebook Inc. is working on a way for you to message your friends and update your news feed...telepathically.

When the company announced this at its annual F8 conference in mid-April, it was pretty vague about how the feat would be accomplished. Turns out the plan includes building a technology that would, by itself, revolutionize how we study the human brain.

Are the methods crazy? Yes. Do neuroscientists and engineers outside Facebook express extreme doubt this will succeed? Yes. Facebook doesn't care and is investing millions in research that could produce a consumer gadget.

After I spoke with project members, based at Facebook's mysterious Building 8 incubator for moonshots, it became clear that the company's larger goal is to make a handful of long-term bets on technologies that could define the next era of computing.

When your face is stuck inside a VR headset or you're out walking around wearing a pair of augmented-reality glasses, you can't exactly reach for a keyboard or mouse, says Mark Chevillet, a physicist and neuroscientist who is Facebook's technical lead on the as-yet-unnamed project.

The initiative would give Facebook a way to control those systems hands-free. Messaging is just the beginning. Facebook isn't working on a brain implant -- though other Silicon Valley giants are. The answer could ultimately be as simple as a headband.

To pull it off, Facebook has enlisted a small in-house team, supplemented by 60 scientists and engineers from research institutions across the U.S., all receiving funding from Facebook. Their goal is to update an obscure, largely abandoned technology known as "fast optical scattering," aka "event-related optical signal." Basically, you shine a light through the head and into the brain, then measure the light reflected back.

It sounds bonkers, but in one way or another, scientists have been using light to peer into the body for nearly a century.

When this technique is used on lab animals, their brains are exposed and researchers can directly observe brain cells expanding and contracting as they fire. The challenge for Dr. Chevillet's team is to accomplish the same thing in intact humans, when there's a layer of skull, skin and hair in the way. It's a problem that to date has been impossible to surmount.

If you've ever pressed a flashlight into your hand and seen it glow, you know how light can make its way through human flesh. Facebook's researchers think they have a chance of success because they're developing sensors to identify the small number of photons that, after penetrating the skull and bouncing off a neuron, return to the detector instead of scattering in every direction.

Sensor technology that could in theory accomplish this -- developed at the Johns Hopkins University Applied Physics Laboratory and funded almost exclusively by the U.S. Department of Defense -- has to date benefited things like sonars, space exploration and observing the ground from the air through dense foliage. Thus far, the technology hasn't been pressed against anyone's head.

Krishna Shenoy, a pioneering neuroprosthetics researcher at Stanford University, says if there's any evidence Facebook's team could sense brain activity from outside the skull nearly as accurately as with implants, he hasn't seen it. Dr. Shenoy and his team have experimented with having humans type eight words a minute by moving a cursor on a virtual keyboard through thought alone -- but even that required a brain implant.

And repurposing U.S. military technology to observe the machinery of thought is only half of the Facebook research team's problem. The other challenge is transforming brain readings into actual words.

Alexander Huth, of the University of California, Berkeley, isn't connected to Facebook's project but has been working on how brains process language for eight years. His research revealed that words -- and the concepts that underlie them -- are spread across the surface of our brains. By observing which parts of a brain are active, you might be able to determine the word, or at least concept, that someone is thinking.

Dr. Huth argues that the greatest challenge in "reading" minds is that we still know so little about how language works.

At Building 8, they're approaching this in a quintessentially Facebook way: by throwing artificial intelligence at it. Dr. Chevillet says that if he and his team can get enough data from the right parts of the brain, they could train a machine-learning algorithm that correlates neural activity with language to extract words from our heads. The method would be similar to how scientists train computers to understand spoken language.

I explained Facebook's plan to Richard Barbour, managing vice president of NIRx Medical Technologies, a pioneer of a related (but distinct) light-based brain-imaging tech used in laboratories all over the world. His response was a long pause.

"I'm not saying it's impossible," he says, "but it wouldn't be my approach."

As risky as all this sounds, Facebook isn't just shooting in the dark. The research program at Building 8 is managed by Regina Dugan, who comes from the DOD's Defense Advanced Research Projects Agency by way of Google. The sizable project budgets come with strict two-year deadlines. Still, none of the tech giants who hire academics to research topics in their skunkworks have, to date, yielded much in the way of out-of-the-box breakthroughs.

"We accept that this problem is high risk, high reward," says Dr. Chevillet.

Write to Christopher Mims at christopher.mims@wsj.com

(END) Dow Jones Newswires

June 11, 2017 08:14 ET (12:14 GMT)