The first step is “filtering for quasi-ballistic photons,” declared Regina Dugan, the head of Facebook’s secretive Building 8 development team, at the annual Facebook F8 developer conference in San Francisco.
“We have a goal of creating a system capable of typing at 100 words per minute, five times faster than you can type on your smartphone, straight from your brain,” said Dugan during her presentation.
Achieving that ambitious goal requires inventing an optical imaging technology that will allow Facebook to see the actions of individual neurons inside the brain.
This relies on something called quasi-ballistic photons, or parts of light that are only deflected a little by going through a material. When you point a laser through your finger, it glows because the photons in the laser are deflected away from their original path. Ballistic photons are the rare few that get through your finger and exit where you originally pointed the beam. Quasi-ballistic photons are only deflected a little bit, so if you look around the edge of where the beam was supposed to exit, you can find them. If you can use quasi-ballistic photons to look at brain activity, instead of just ballistic photons, your chances get a lot better at seeing what is happening.
While the details of how Facebook is planning to create this so-called “Silent Speech System” haven’t been revealed, Dugan’s tantalizing hint about the use of quasi-ballistic photons suggests a few options. With Facebook’s resources, the possibility for success is real, although maybe not within the tight two-year window.
“They Have to Upgrade the Physics”
“The brain imaging is actually kind of wild,” Junjie Yao, a biomedical engineer at Duke University, tells Inverse. “If they’re trying to use the photons to do the job, they have to upgrade the physics.”
This is not to say that Dugan — who left her position as director of DARPA to work at Facebook — recognized that what she’s presenting isn’t remarkably ambitious.
“Now no such technology exists today, we’ll need to develop one, and we think optical imaging is the best place to start,” Dugan said. “So we’ll need new non-invasive sensors; sensors that can measure brain activity hundreds of times a second and precise to millimeters, and without signal distortions even as they read through hair and skin and skull.”
Facebook wants to be able to demonstrate this by 2019.
“It is an ambitious but achievable goal,” a Facebook spokesperson told Inverse after the presentation.
Yao, an expert in neural imaging, and quasi-ballistic photons in particular, agrees on it being both ambitious and potentially achievable, but he’s skeptical Facebook can do it in two years. In most cases, when a photon travels through something like tissue, it bounces off atoms within the material and is scattered from its original direction. Very few photons, called ballistic photons, will make it through tissue without hitting anything, retaining their original direction. Some photons are only scattered a few times, and come out pretty close to its original destination. These are quasi-ballistic photons.
In an ideal world, you could use ballistic photons to tell you about what’s inside the brain. But there just aren’t enough of them. So Facebook says it’s adding in the quasi-ballistic photons, too. “And if we get the trade-off just right, we can get the spacial resolution we need and have enough photons to measure,” Dugan said.
“If we’re talking about current technology, I don’t know if Facebook can do this because physics tells us it’s just too challenging to acquire the quasi-ballistic photons from the human brain with a centimeter or so of penetration,” Yao says.
Currently, quasi-ballistic photons are used in imaging peanut-sized mouse brains that have been plucked out of their skulls. For the resolution Facebook is talking about, ballistic photons can make it through about a human hair’s worth of brain tissue, Yao says. Quasi-ballistic photons can be measured after about 10 hair’s worth.
All that said, Yao says that with Facebook’s resources, it’s actually possible that it could make this a reality.
“Statistically speaking, for photons to survive a centimeter is just so small, but it doesn’t mean that it’s zero,” Yao says. “Potentially, if Facebook can develop some super-sensitive photon detector they can probably overcome the limitation.”
What About Wavefront Engineering?
The other way Facebook could create a non-invasive brain imaging method using photons takes advantage of a technique called wavefront engineering, says Yao. In this method, instead of sending all the photons into the brain as something like a laser-beam of light, each photon is pre-calibrated to enter the tissue with a unique direction. So if a researcher can tell that one photon is getting scattered from its original destination by 40 degrees, the photon can be sent into the tissue next time, rotated so that the 40 degree scatter makes it exit where you want it.
“You overcompensate with the photons so you can overcome the scattering effect of the tissue,” Yao explains.
This technique is extremely slow, says Yao. It can take minutes to adjust a single photon’s path, and each path has to constantly be adjusted because brain tissue is dynamic and constantly changing. “I think it’s not totally impossible if Facebook has the right technology,” says Yao. “If this is doable, it’s going to take a lot of people out of business — maybe my job as well — but I hope that they can succeed.”
If Facebook succeeds at creating a non-invasive, real-time, neuron imaging technology, it would be revolutionary, Yao says. The technology would be groundbreaking in many areas of research, particularly in medicine, and totally change how we understand how the brain functions.
“If Facebook can overcome the limitations, we’re talking about optical imaging for the human brain, or a full-body human imaging, then that is going to be a gold mine. That would not just be for fun, for Facebook to play with, but very impactful,” Yao says.
“There are so many things we can do with this technology, it could change the world.”