Elon Musk is set to make an announcement about Neuralink, a company designing brain-computer interface technology, on Friday, Aug. 28. It sounds like science fiction, but research in this area has progressed rapidly in recent years, though we’re still far from being able to send emails with our minds. Unlike Musk’s other famous ventures, SpaceX and Tesla, however, Neuralink will be vastly more limited in terms of how fast it can innovate and push out consumer products. Here’s what you should know about the project, including what’s theoretically possible, how skeptical you should be, and who else is designing brain-computer interfaces.
Announced by Elon Musk in 2017, Neuralink will attempt to use “ultra-high bandwidth brain-machine interfaces to connect humans and computers,” or more simply, to connect human brains with computers via implantable brain chips.
At first, Neuralink’s brain-machine interfaces could be used to treat brain disorders, such as Parkinson’s disease, epilepsy, and depression. They could also be used in conjunction with advanced assistive devices, in which a person’s thoughts could control artificial limbs or other prosthetics. Should Musk’s ultimate vision be achieved, however, this technology would take on a more transhumanistic complexion, allowing future humans to control external devices with their minds, transmit thoughts directly to another person’s brain, and even augment cognitive capacities, such as increased intelligence and memory.
More conceptually, Musk has positioned Neuralink as a potential way for humanity to prevent an AI apocalypse, saying the technology could help us “achieve a sort of symbiosis with artificial intelligence,” as he said when the project was launched three years ago. By boosting our puny brains, he argued, we will stand toe-to-toe with our advanced technologies, in a kind of “can’t beat ‘em, join ‘em” solution to the pending problem, which I critiqued back in 2017.
These ideas are nothing new, of course. Science fiction has been on top of this for decades, whether it be William Gibson’s cranial jacks, Iain Banks’s neural lace, The Matrix’s brain plug, or any speculative vision in which human minds commune directly with the digital realm.
So this all sounds very fascinating—and it is—but here’s the requisite bucket of cold water: Unlike electric cars or rockets, brain-computer interfaces are considered medical devices, which means the company will have to go through the appropriate regulatory channels to get its experiments and products approved for use in humans, including consent from the U.S. Food and Drug Administration.
Like other drug and medical device developers, whether public or private, Neuralink will have to demonstrate the safety and efficacy of its products, typically through meticulous and time-consuming clinical trials. Given that the company wants to implant chips into people’s brains—including the brains of perfectly healthy people—this will present some unique challenges, involving timeframes that may be measured in decades. Neuralink will also be hampered by the fact that some of its more futuristic offerings will be considered an enhancement, not a therapy, which will undoubtedly further complicate regulatory approvals.
Despite these challenges, scientists have made great strides over the years as they try to turn science fiction into reality. Elon Musk might get the most media attention, but not-so-famous researchers have been making breathtaking progress in this area, giving us a sneak preview of what might actually be possible.
Last year, a team of neuroscientists from Columbia University translated brain waves into recognizable speech, while a team from the University of California San Francisco built a virtual vocal tract capable of simulating the mechanical aspects of verbal communication by tapping directly into the human brain. In 2016, a brain implant allowed an amputee to use their thoughts to move the individual fingers of a prosthetic hand. Brain-machine interfaces have also been used to create mind-controlled robotic exoskeletons and restore a sense of touch and partial motor function in people with spinal injuries. Some interesting work has also been done to mediate brain-to-brain communication in humans, though it’s still early days.
Work in nonhuman animals has also yielded good results. Notable examples include a wireless brain-machine interface that allowed a monkey to control a wheelchair with its mind and a brain implant that enabled monkeys to type at 12 words a minute using only their thoughts.
Musk’s foray into this world is hardly groundbreaking, at least for now. What’s potentially different is the scale, funding, and intent of the Neuralink mission, not to mention the charismatic nature of Elon Musk himself. That said, there are some rival projects outside of academia, including similar endeavors launched by Facebook (which recently purchased neural interface startup Ctrl-labs in a deal worth somewhere between $500 million and $1 billion); Kernel (a $100 million project launched by Braintree founder Bryan Johnson); and the U.S. government’s DARPA, which has devoted $65 million to its effort. Musk, it would seem, is hardly the only person throwing big money at this sort of initiatives, and it remains to be seen if Neuralink will succeed in what appears to be an increasingly competitive space.
The Neuralink system will employ “neural lace” technologies (an apparent hat tip to author Iain Banks)—presumably a method for using brain implants, or a kind of implantable mesh, to connect brains with an external computer using a “direct cortical interface,” as the Wall Street Journal reported back in 2017.
As of 2019, $158 million in funding has been channeled to the project, including $100 million from Mr. Musk himself, reports the New York Times. The company already employs 90 people and has plans to include neurosurgeons from Stanford University and possibly elsewhere.
Neuralink will take an incremental approach to the problem, starting with the treatment of brain disorders and then scaling up to more enhancement-minded applications. Boosting the bandwidth of information coming out from the brain will be critical to any progress. This will almost certainly have to involve wireless brain implants (as opposed to non-invasive techniques such as EEG), requiring surgery and flexible, durable, biocompatible components.
Musk disclosed further details of the Neuralink approach back in July 2019 during a live-streamed presentation at the California Academy of Science. Their solution, as also elucidated in the company’s whitepaper, would involve a sewing machine-like robot, which a surgeon would use to implant ultra-thin threads, or probes, into a person’s brain. At just 5 to 6 nanometers wide, these threads would be thinner than human hair.
These threads would connect to chips embedded in the skull, like strings of pearls. As noted in the whitepaper, the machine would be capable of implanting six threads, or 192 electrodes, per minute. The Neuralink team has already demonstrated “the rapid implantation of 96 polymer threads, each thread with 32 electrodes for a total of 3,072 electrodes,” according to the paper. Brain surgery would still be required, but Neuralink president Max Hodak envisions the same task being accomplished by lasers, as a way of avoiding mechanical drilling, according to the New York Times.
Neuralink has already demonstrated a system capable of reading information from 1,500 implanted electrodes, though in rats. Still, this is 15 times better than current systems used on humans.
“It’s impressive to see how quickly they have got to this point, and will be interesting to see how far they get,” wrote Andrew Jackson, professor of neural interfaces at Newcastle University, in an email. “Theirs is one of a number of efforts to ‘read’ the electrical activity of large numbers of brain cells. The Neuralink approach is to insert many flexible polymer threads into the brain using a sort of sewing machine. The threads attach to an electronic package implanted under the skin.”
Jackson described other notable approaches in which electronics are incorporated onto small silicon needles, including the Neuropixels probe developed by Tim Harris from the Naelia Research Campus at the Howard Hughes Medical Institute. The $5.5 million collaboration has already produced probes capable of recording more than 700 neurons simultaneously. Jackson also pointed to a concept called “neural dust,” in which many small, wireless implants are distributed throughout the brain.
“Only time will tell whether Elon has backed the right horse,” said Jackson. “One thing this does demonstrate is the potential for commercial investment to advance the field of neural interfaces. Until recently, neuroscientists were using some pretty old-fashioned equipment to record from the brain, so it’s great to see this kind of interest and investment from Silicon Valley.”
Neuralink was supposed to have started tests on human subjects by now, but it hasn’t. It’s possible the company was overly ambitious with its timelines, or it was denied the requisite approval from the U.S. Food and Drug Administration, but we really don’t know. The company has expressed interest in opening its own animal testing facility in San Francisco, highlighting its need for ongoing experimentation with animal models.
Kevin Warwick, an expert on brain-machine interfaces from Coventry and Reading Universities, likes that Neuralink is using thin polymer probes, and not just because they’re flexible.
“This is great because different patterns of multiple electrodes can be fabricated. It should also help with regard to mechanical issues, as they’re unlikely to break,” he explained in an email. “The problem is how to insert it into the brain, for which they have designed a robot.”
Warwick says the whitepaper includes a “hand waving” description of the robot, which is regrettable, as it’s “critical to the whole method,” he said. “If the robot can do what they say it can, then we would be able to have many electrodes in multiple sites. But for me, this is the part that needs to be proved—can they insert such polymer probes reliably, safely, accurately into the brain, and show that the robot works on the human brain?”
Looking ahead, the Neuralink team—and anyone else working on neural interface devices—will need to overcome several major challenges, including the invasive nature of the technology, developing a universal way to map brain signals (each system will have to learn the idiosyncrasies of each person’s brain), and scaling the required testing (both in nonhuman animals and people) in safe, ethical, and effective ways.
They’ll also have to deal with potential unforeseen problems, such as excessive heat generated by the implants or rapid obsolescence of implanted devices. Importantly, the researchers will have to determine if all that data being transferred out of the brain can actually be applied to something useful and in a way that attracts commercial interest. There’s still lots we don’t know about the human brain and how it works, so it may be a stretch to assume these current strategies will work as intended.
We’re interested to hear Musk’s announcement on August 28 and to learn what progress has been made in the past year. But we’re not ready to hype up this project just yet, as we should expect slow and incremental updates, given the complex nature of the endeavor.