Ars chats with Precision, the brain-chip maker taking the road less invasive

Brain-chip buzz —

Precision tested its BCI on 14 people so far. Two more are scheduled this month.

Precision’s Layer 7 Cortical Interface array.

Enlarge / Precision’s Layer 7 Cortical Interface array.

Work toward brain-computer interfaces has never been more charged. Though neuroscientists have toiled for decades to tap directly into human thoughts, recent advances have the field buzzing with anticipation—and the involvement of one polarizing billionaire has drawn a new level of attention.

With competition amping up in this space, Ars spoke with Ben Rapoport, who is a neurosurgeon, electrical engineer, and co-founder of the brain-computer interface (BCI) company Precision Neuroscience. Precision is at the forefront of the field, having placed its BCI on the brains of 14 human patients so far, with two more scheduled this month. Rapoport says he hopes to at least double that number of human participants by the end of this year. In fact, the 3-year-old company expects to have its first BCI on the market next year.

In addition to the swift progress, Precision is notable for its divergence from its competitor’s strategies, namely Neuralink, the most high-profile BCI company and headed by Elon Musk. In 2016, Rapoport co-founded Neuralink alongside Musk and other scientists. But he didn’t stay long and went on to co-found Precision in 2021. In previous interviews, Rapoport suggested his split from Neuralink related to the issues of safety and invasiveness of the BCI design. While Neuralink’s device is going deeper into the brain—trying to eavesdrop on neuron signals with electrodes at close range to decode thoughts and intended motions and speech—Precision is staying at the surface, where there is little to no risk of damaging brain tissue.

Shallow signals

“It used to be thought that you needed to put needle-like electrodes into the brain surface in order to listen to signals of adequate quality,” Rapoport told Ars. Early BCIs developed decades ago used electrode arrays with tiny needles that sink up to 1.5 millimeters into brain tissue. Competitors such as Blackrock Neurotech and Paradromics are still developing such designs. (Another competitor, Synchron, is developing a stent-like device threaded into a major blood vessel in the brain.) Meanwhile, Neuralink is going deeper, using a robot to surgically implant electrodes into brain tissue, reportedly between 3 mm and 8 mm deep.

However, Rapoport eschews this approach. Anytime something essentially cuts into the brain, there’s damage, he notes. Scar tissue and fibrous tissue can form—which is bad for the patient and the BCI’s functioning. “So, there’s not infinite scalability [to such designs],” Rapoport notes, “because when you try to scale that up to making lots of little penetrations into the brain, at some point you can run into a limitation to how many times you can penetrate the brain without causing irreversible and undetectable damage.”

Further, he says, penetrating the brain is just unnecessary. Rapoport says there is no fundamental data that suggests that penetration is necessary for BCIs advances. Rather, the idea was based on the state of knowledge and technology from decades ago. “It was just that it was an accident that that’s how the field got started,” he said. But, since the 1970s, when centimeter-scale electrodes were first being used to capture brain activity, the technology has advanced from the macroscopic to microscopic range, creating more powerful devices.

“All of conscious thought—movement, sensation, intention, vision, etc.—all of that is coordinated at the level of the neocortex, which is the outermost two millimeters of the brain,” Rapoport said. “So, everything, all of the signals of interest—the cognitive processing signals that are interesting to the brain-computer interface world—that’s all happening within millimeters of the brain surface … we’re talking about very small spatial scales.” With the more potent technology of today, Precision thinks it can collect the data it needs without physically traversing those tiny distances.