What’s the deal with Neuralink?
Launched Nearly 5 years ago with the help of Elon R Musk, Neuralink focused on groundbreaking interfaces that effortlessly listen to electrical signals in the brain and, at the same time, write electrical pulses onto the brain. But even by the norms of Silicon Valley, the organization has maintained a close watch on its development, doing all the engineering, analysis and animal testing in-house. The vision of marrying biological brains to artificial brains is not special to neuralink. Over the last decade, there has been an increase of brain machine interfaces or brain computer interfaces — some inserted into the brain, some into peripheral nerves. or others that rest like a helmet outside the skull. The core theory behind all the contraptions is simple: the brain is driven by electrical signals. If we can tap into these mysterious neural code — the brain’s secret language — we may actually become the architects of our own minds
Let the paralyzed people walk again? Check it out and done. Power the robotic arms with the mind? Yep. Rewriting brain impulses to combat depression? Right now, in people. “Recording the electrical behavior behind the basic memories and replaying it ? Human trials are underway. Linking human minds to a BrainNet to cooperate on a Tetris-like game over the internet? It is possible. With this context, maybe the most exciting aspect of the show is not the high profile projections of what brain machine interfaces or brain computer interfaces might actually achieve one day. In a way way we’re almost there . Rather, it was the revamped Link device itself that stood out. At Neuralink’s coming out party, the organization picture a wireless neural implant with a luxurious ivory processing device worn at the back of the neck.
The implant electrodes themselves are “sewn” onto the brain through advanced robotic surgery, focusing on the brain imaging procedures to prevent rupturing the blood vessels and brain bleeding. The problem with the design, Musk said “It had several parts and it was complicated. you also wouldn’t look absolutely human, since there’s something sticking out of your ear.” The prototype arrived with a very different physical shell. About the size of big coin, the unit replaces a little portion of your skull and lies flat with the surrounding skull. The electrodes, which are inserted within the brain, are attached to the topical device. If the hair is hidden, the implant is transparent.
The company envisions an outpatient therapy where a computer may simultaneously extract a portion of the skull, sew the electrodes in and cover the missing piece of the skull with the unit. According to the team, the relation has identical physical properties and thickness to the skull, rendering the substitution a kind of copy and paste. when implanted, the bond is sealed to the “superglue” skull.
“ I could have a neuralink right now, and you wouldn’t know,” Musk said. with such a compact unit, the team put a wonderful selection of features into it. The “Link” unit has over 1000 channels that can be individually enabled. This is on par with neuro pixel with 960 recording channels that is currently commonly used in testing, like the Allen institute of Brain science. compared to the utah array, the iconic implant device used for brain stimulation in humans with just 256 electrodes, the relation has a strong advantage in terms of pure electrode density. What is perhaps most remarkable, however, is the on board processing of neural spikes — the electrical pattern produced by neurons as they shoot. Electrical signals are fairly noisy in the brain, and filtering noise bursts, as well as sorting electrical stimulation trains into spikes, typically requires quite a bit of computing power. That is why neural spikes are typically reported off-line and processed using computers rather than on board electronics in the lab.
The dilemma is much more complex when contemplating the transmission of wireless data from the embedded device to the remote handset. Without effective and reliable compression of these neural results, the upload might be incredibly sluggish, run out of battery power, or heat up the computer itself — something you don’t want to do with a device trapped inside your brain. To solve these challenges, the team focused on algorithms that use “characteristic shapes” of electrical signals that appear like spikes to accurately recognize individual neural firings.
The data is stored on the chip inside the skull. Recordings from each channel are screened to eliminate apparent noise, and spikes are then observed in real time. Since various groups of neurons have their signature way of spiking — that is, the “shape” of their spikes is diverse — the chip may also be programmed to identify the exact spikes you’re searching for. This suggests that in principle, the chip will only be designed to record the kind of neuron behavior you’re interested in — for example, looking at inhibitory neurons in the cortex, and how they regulate the transmission of neuronal input.
This processed spike data are then sent to smartphones or other external devices via Bluetooth to allow wireless monitoring. The ability to do this effectively has been a stumbling block in wireless brain implants — raw neural recordings are too large for efficient delivery, and automatic data spike detection and compression is difficult, but a necessary step to enable neural interfaces to actually “cut the wire.” Connection has some other impressive features. For one, the battery life lasts all day, and the device can be charged at night by inductive charging. From my subsequent discussions with the team, it seems that alignment lights can help track whether the charger is aligned with the battery. Furthermore, the Connection itself also has an internal temperature sensor to check for overheating, which will immediately disconnect if the temperature increases above a certain threshold — a very important safety precaution that does not overheat the surrounding skull tissue.
From the outset of the demonstration, there was an undercurrent of friction between what is achievable in neuro engineering and what is needed to understand the brain. Since its inception, Neuralink has always been intrigued by electrode numbers: increasing the number of channels on its instruments and increasing the number of neurons that can be registered at the same time. At the case, Musk said that his goal was to raise the amount of neurons registered by a factor of “100, then 1,000, then 10,000.”
But here’s the point: as neuroscience gradually recognizes the neuronal code underlying our cognitive processes, it’s obvious that more electrodes or more activated neurons aren’t necessarily easier. Many neural pathways use what is termed “sparse coding,” so that only a few of neurons, activated in a way that mimics normal firing, will artificially activate visual or olfactory stimuli. Through optogenetics — a technique by activating the neurons with light — scientists already realize that it is possible to consume memories by manipulating only a few main neurons in the network. Sticking a lot of wires through the brain, which eventually causes scarring, and zapping hundreds of thousands of neurons, is not really going to help. Unlike engineering, the brain approach is not more networks or more implants. Rather, it deciphers the neural code — knowing what to induce, in what order, what action to produce. It is therefore telling that, amid promises of neural stimulation, the only evidence seen at the event were neurons firing from the mouse brain section — using two-photon microscopy for neural activation — after zapping brain tissue with an electrode. What material, if any, is actually “written” in the brain? Without an understanding of how neural pathways function and in what sequences, zapping the brain with electricity — no matter how cool the system itself is — is like pounding on all the keys of the piano at once, rather than writing a lovely melody.
Of necessity, the dilemma is much greater than Neuralink itself. Maybe it’s the next frontier to solve the brain’s puzzles. To their credit, the Neuralink team looked at possible brain injury from electrode placement. The biggest issue with new electrodes is that the brain can gradually activate non-neuronal cells to create an insulating sheath around the electrode, sealing it away from the neurons it needs to record from. According to some of the workers, scarring around electrodes is small, but there might be scar tissue buildup in the scalp in the long term.
This will make electrode threads difficult to remove — something that still needs to be refined. However, two months is just a fraction of what Musk proposes: a decade-long implant of hardware that can be upgraded. The team could also have a response to that. Rather than removing the whole implant, it might theoretically be helpful to leave the threads within the brain and remove only the top cap — the Link unit that houses the processor chip. The team is now carrying out the proposal, while considering the prospect of full-on removal and re-implantation.
As a demonstration of feasibility, the team brought out three pigs: one without an implant, one with a Link, and one with a Link implant and then removed. Gertrude, a pig currently inserted in locations connected to her snout, had her inner neural shots broadcast as a collection of electrical cracks as she roamed around her enclosure, stuck her snout in a variety of food and hay and ran into her handler. Pigs had arrived as a surprise. Many reporters, like me, were awaiting non-human primates. However, pigs appear to be a reasonable choice. For one, their skulls are similar in density and thickness to human skulls. In addition, they are clever sweets, which ensures that they can be taught to walk on a treadmill while the implant tracks the activity of each joint from their motor cortex. It is likely that pigs may be conditioned in more complex exercises and activities to prove that the implant influences their movements, tastes, or judgement. For the time being, the team does not yet have widely available evidence demonstrating that targeted activation of the pigs’ cortex — say, motor cortex — can force their muscles into motion. While it was mounted as a test, it is clear that the Relation remains experimental.
The team works closely with the FDA and was given a groundbreaking device certification in July 2020, which may pave the way for a human trial to treat patients with paraplegia and tetraplegia. However, it remains to be seen if the trials will take place by the end of 2021 (was promised that it would happen by Dec 2020). Rather than other brain-machine interface firms, which usually concentrate on brain diseases, it is clear that Musk sees Connection as something that can increase perfectly healthy humans. Given the need for surgical removal of part of your head, it’s hard to tell if it’s a compelling sell to the ordinary person, even with Musk’s star power and vision of the natural sight, memory retrieval, or a “third artificial layer” of the brain that connects us to AI. And when the team showed only an extremely condensed image of the pig’s neural fire — rather than real spike traces — hard it’s to gauge just how sensitive the electrodes actually are. Finally, right now, only the cortex — the outer layer of the brain — can record the electrodes. This leaves deeper brain circuits and processes, including memory, depression, emotion, and certain kinds of mental disorders, out of the table. Although the team is optimistic that the electrodes will be increased in length to reach the deepest areas of the brain, they are working for the future. Neuralink’s got a long way to go. All that said, having someone with Musk’s influence championing a fast-moving neurotechnology that could help people is priceless.
It’s possible to underestimate the excitement that neuroscience gives to people because you’ve been there for years or decades. It’s quick to roll our eyes to Neuralink’s data and say, “well, neuroscientists have been listening to live neurons firing inside animals and even humans for over a decade.” As much as I’m always cynical about how Connect relates to state-of-the-art neural probes built in academia, It is really amazing how much a comparatively small leadership team has achieved in the past 2 years. Neuralink is only off and aiming high. To quote Musk: “There is a huge amount of work to be done to move from here to a device that is widely available and affordable and reliable.”