Neuralink, the secretive company bankrolled by Elon Musk to develop brain-machine interfaces, has finally revealed what it’s been working on over the last two years. But despite promises of human trials, the technology needs to progress a long way before it can make an impact.
In a presentation to the California Academy of Sciences on Tuesday evening, Neuralink presented a medical device capable of reading information from 1,500 flexible electrodes connected to a laboratory rat – 15 times faster than current systems embedded in humans. The goal is to eventually implant it in people with paralysis or other medical conditions that will let them control computers with their minds – and the company has ambitious plans to begin human trials as soon as next year.
So how does it work? Neuralink says surgeons would have to drill holes through the skull to insert flexible electrodes. But in the future, they hope to use a laser to pierce tiny holes in the skull. “One of the big bottlenecks is that a mechanical drill couples vibration through the skull, which is unpleasant, whereas a laser drill, you wouldn’t feel,” Max Hodak, Neuralink’s president told the New York Times. The threads would be considerably thinner than a human hair, some 4 to 6 μm in width.
If fully functional, Neuralink’s flexible threads may have a substantial advantage over older technology as they are less likely to damage the brain. “What we currently find is, if we put [stiff] electrodes into a brain, something like scar tissue starts to build around them after a few months,” says Konrad Kording of the University of Pennsylvania, who is an expert in computational neuroscience. He adds that the quality of electrodes degrades rapidly as the brain moves.
Any threads that are placed into the brain will need to be durable and stable. “If we put tech into people, then it must stay there for life. We can’t take stuff in and out of brains in arbitrary intervals,” Kording adds. “It always makes damage”.
The flexible cellophane-like conductive wires that Neuralink has been developing is a concept that has had a lot of interest in the academic field, Kording says. Recent technology that has been tested by the international consortium called BrainGate has allowed people to control a robotic arm to drink from a can and to type, using only their thoughts. But it relies on a series of stiff needles with up to 128 electrode channels, which can be problematic in the long-term as the brain moves but the needles don’t.
Neuralink’s polymers may solve that problem, but neurosurgeons will still require a needle-like tool to insert the soft wires, Kording says. Enter: the “sewing machine”. Musk’s startup, which has received £120 million in funding and hired a team of 90 people, has developed ”a neurosurgical robot capable of inserting six threads (192 electrodes) per minute [automatically],” according to a technical white paper published alongside its presentation. The robot, which looks something like a cross between a microscope and a sewing machine would use a stiff needle to insert the threads and avoid blood vessels in doing so, reducing the risk of an inflammatory response in the brain.
However, there is still a risk for infection with soft wires penetrating through skin layers, says Chad Bouton, director of the Center for Bioelectronic Medicine at the Feinstein Institute for Medical Research in New York.
His team currently uses stereo-electroencephalography (EEG) in patients suffering with paralysis. Neuralink may have found a way to fabricate and connect the electrodes, but Bouton says a major challenge will be to get the information out of the brain. Neuralink’s microchip, the N1 sensor, would be wired into the skull.
It currently transmits data via USB-C wired connection, although the team is working on a wireless option. “There have been some advances in wireless telemetry but there are still challenges around powering those devices implanted in the brain without generating too much heat,” Bouton says, adding that achieving the bandwidth that Neuralink strives for still raises questions. The more bandwidth and the more electrodes are implanted, the more data they will transfer, which will require more power.
Overall, the use of flexible and soft threads may seems to be the next step in the field of brain–computer interfaces. However, Musk’s plan to trial the first implants on humans in 2020 seems wildly optimistic, according to Ana Matran-Fernandez, an artificial intelligence industry fellow at the University of Essex.
The approval processes from the US Food and Drug Administration (FDA) can be slow and often take several attempts to be approved. And recruiting human test subjects might prove even more difficult within such a short timescale.
Matran-Fernandez’ team, which is currently working on a project involving transradial amputees, took more than a year to find a single volunteer to try a technology that is much less invasive than Neuralink’s. “If you are already have something that is working as is the case with many amputees, you might be reluctant to try something new.” Someone with an intact brain, she thinks, would be even less likely to risk invasive surgery.
Bouton adds that patients might be willing to experiment with new technologies as long as these are done in the most efficient and safest manner. “It comes back to making sure that the technology will be effective and will actually have a positive impact on their daily lives,” he says. The medical applications of brain-computer interfaces, such as restoring hand movement, should be one of the top priorities, Bouton says.
Some 50 million people worldwide live with some sort of paralysis today, and at least 250,000 suffer a spinal cord injury every year. “I’m happy to see companies investing in the brain-computer interface area because of the important [medical] applications it could be used for,” Bouton adds. He believes laying out plans with specific goals could accelerate the sector. “The question is what you are moving towards. If you don’t have a well-defined endpoint, you can pour in as much money as you want into it, you will still find yourself going into different tangents,” he says. Focusing on a specific medical application would allow the company to work backwards and establish what the unknown challenges and risks are associated with introducing this sort of new technology.
What Neuralink’s “sewing machine” will be really for is unclear at this stage. Parkinson’s disease, for instance, is typically treated in a small number of parts in the centre of the brain – the subthalamic nucleus and the thalamus – using a stimulator with four to six stiff electrodes at different depths.
“Now, when they talk of their own device it’s very vague. The 96 threads they talk of, would you be able to push them down into that region?,” questions Kevin Warwick, a cybernetics professor at Coventry University. Current stimulators used to treat Parkinson’s disease are effective, he says, and wouldn’t require thousands of connections. “In a way we’ve had technology that could have taken things further in recent years such as BrainGate. If they now have threads with 1,000 plus connections, they have a lot more flexibility but they need to do the experiments,” he says, raising questions about what those experiments would entail and how Neuralink would take the technology further than therapy.
Neuralink seems to have the right people on board and the necessary resources and technology, but “what’s the first thing that they are going to have a go at and do experiments on?,” Warwick asks.
While Musk’s vision of a mind-reading computer may still be a long way off, focusing on the technology’s medical use will be necessary to receive backing from the FDA and start trials on humans in the coming years. Musk’s team said it will be working with neurosurgeons such as Jaimie Henderson from Stanford University, who is an expert in the treatment of epilepsy and an adviser to Neuralink, to take the next step in creating the clinical device.
The company’s current focus may lie on bringing a therapeutic device into market that is able to treat medical conditions such as paralysis or Parkinson’s disease, but Musk seems to have bigger plans. Speaking last year on The Joe Rogan Experience podcast, he said the ultimate technology would allow humans to “effectively merge with AI”.
Warwick thinks this idea is not completely far-fetched. “I’m 100 per cent with him on that. That’s the way to go and it’s incredibly exciting,” he says, adding that the potential of upgrading humans is enormous. “But I challenge him to actually have a go himself. He talks the talk but he hasn’t done anything [experiments] himself.”
This article was originally published by WIRED UK
