‘How much smarter are you with a phone or computer than without? You’re vastly smarter, actually…it’s just that the data rate…it’s slow, very slow. It’s like a tiny straw of information flow between your biological self and your digital self. We need to make that tiny straw like a giant river, a huge, high-bandwidth – Elon Musk.
And that’s exactly what his company, Neuralink, is embarking on to do. However, it is noteworthy to understand that before Neuralink becomes advanced enough to heighten human sagacity; Neuralink first ventures to provide and restore function in people who have lost specific functions or have never experienced them since birth. Ambitiously, Neuralink staff declared that, in the future, their product has the potential to give a congenitally blind person the ability to see. Neuralink pledges a tomorrow where people with immobility have the capacity to interact with computers as efficiently as you and I can. Miguel Nicolelis, scientist, physician and Duke School of Medicine Professor in Neuroscience even disclosed; "Eventually, brain implants will become as common as heart implants. I have no doubt about that." The eminence and capacity of what Neuralink is at the forefront of doing is a rather exhilarating one to keep an eye out for.
The primary objective of Neuralink is to construct a whole-brain interface. Effectively creating an input, output device that can interact with all facets of the brain. They have been working to improve their current products; N1 and R1 - N1 is an implantable chip and R1 is a robot that performs the surgery to insert the chip into the brain.
Amidst declarations of animal abuse in Neuralink experiments; the show and tell demonstrated otherwise. During Elons introduction he made sure to preface the emphasis on the care and upkeep delivered to all the animals at Neuralink firms. The macaque monkeys were adoringly named and taken care of. Videos shown at the show and tell demonstrated that monkeys could merely 'walk away' from an experiment if they didn't feel like partaking, ascertaining that there is no 'intenseness' enforced into Neuralink ventures. Elon expressed that they aren't 'testing' on monkeys as the products are designed with intense measures of preciseness and have been tested on digital models of the brain. 'Testing' would suggest that the monkeys are at risk of fatal outcomes, which is shown not to be the matter.
Thus far, Engineers at Neuralink have only investigated what monkeys and pigs can do with N1, however, the first human clinical trials are currently in the process of obtaining clearance from the FDA. They are, however, confident that they will be able to conduct human clinical trials in the next 6 months; which is undoubtedly extremely thrilling.
This article will be split into 3 sections;
· N1 and R1
· Problems discussed during the show and tell
· Implications for the future
N1 and R1
N1 records the neural activity from the motor cortex of a monkey; from 1000 different channels. This activity is measured by 64 threads, each containing 16 electrodes, that are inserted into the monkey’s brain by R1; Neuralink’s surgical robot. The information, that is recorded as a monkey traces numbers digitally, is required to assemble a database that will become the foundation for predicting and deconstructing the number that the monkey is 'thinking' of selecting on a screen later on. This database is employed to train a neural network that can estimate the cursor velocity which is used to control the screen directly by the brain without any physical intervention.
How is the recorded data employed to construct a database? Once action potentials are collected, they are decoded to provide information about what certain patterns of activity, picked up by the 1000 channels, indicate. Essentially the action potentials for each different number demonstrate the action potentials that can be used to identify that particular number. This is the methodology that N1 utilizes. Instead of using a joystick or a mouse; the subject can use the screen with their mind instead. Additionally, when subjects use a joystick or mouse there is usually always an auditory/ visual stimulation that serves as the confirmation of having pressed on something; usually associated with the sound of a click. The chip can visually simulate this confirmation on a screen by changing the cursor color.
As the monkey traces different numbers, the N1 records different brain activity. To get a detailed database, engineers would need millions of database samples; this would really be impractical and inconvenient to collect. Therefore, initially, instead of decoding cursor velocity, engineers at Neuralink decode the digit that the subject is tracing on the screen. This method, again, proved to be inconvenient. To produce a suitable database using this methodology, would require too many digit inputs. Hence finally, they came up with a robust mechanism for the system - instead of decoding the entire trajectory of the number, they only decode the trajectory of the end of the trace. From the information collected, an MS data set is used to ‘predict’ the number that is being tracked. Therefore, in the future, when an individual thinks of selecting '8' on a screen, the chip can pick up this activity from N1 and do it automatically – it knows what the subject's brain activity looks like when it wants to press ‘8’. This is entirely prompted by the mere thought of doing so.
Does the action potential that a brain produces when a subject thinks of the number ‘4’, not change daily? Is the brain activity ‘identical’ when a person thinks of ‘4’ every single time? No; in fact, this problem can produce a vector bias - a resistance posed to the cursor, preventing the cursor from moving in the direction that the subject wants. This is a problem that Neuralink is still actively trying to combat. Furthermore, there is also a phase lock embedded into this system. A phase lock automatically adjusts the phase of a locally generated signal to match the phase of an input signal – essentially ‘cleaning’ the signal received from the neurons. The system also utilizes spike detection – neuronal activity is only recorded when neurons spike. This helps to save power as there is no need to continuously measure and analyze any and all electrical activity produced by a set of neurons; Engineers only want the neurons that simulate a spiking neuron model. Moreover, the system also records the shape of the action potential which assists with the recognition of the timing of action potentials. The spike detector embedded into the system is a single-functioning unit with a buffering system called SRAM.
R1, the surgical robot, does the thread insertion part of the surgery after the neurosurgeon performs the incision, craniotomy, and removal of the meningeal layer dura as well as the remainder of the surgery. To make this process more affordable, however; Neuralink is working to remove neurosurgeons from the equation. Neurosurgeons are rare, take 10 years to train and most neurosurgeons are too occupied to partake in such an endeavor. Ultimately, the most effective way of doing this is to let neurosurgeons monitor multiple surgeries at one particular point in time. Neuralink employs femto laser ablation for the surgical procedure as well, however, the surgical insertion of the electrodes, as a whole, is notoriously challenging to conduct as it is described as "inserting a thread as thin as your hair into a jello covered in saran wrap; 64 times at a particular depth, angle and within a specific time" with a tungsten needle. Figure 1 demonstrates the 64 threads that have been implanted into the brain.
Problems and upgrades
Functional Goals
The time it takes for the neural activity to be translated or decoded into an inference must be shortened. Additionally, consistent measures are taken to enhance ASIC, doubling battery life, and reducing power consumption by 15%. Additionally, increasing channels is a major goal Neuralink is looking to achieve; 4096 channels are to be targeted, but the size of the chip is to remain unchanged.
Charging
The system must operate over a wide charging volume as it relies on magnets. This charging mechanism needs to be robust to disturbances and instantaneous. During charging, the temperature must not rise over 2 degrees for thermal safety purposes. Moreover, when the chip is brought too close to the charger, the highest charging frequency goes past the threshold and goes outside the ISM band; this problem is solved by introducing dynamic tuning.
Testing
For testing purposes, the in vivo chemistry of tissue should be simulated well. Neuralink is constantly trying to manufacture changes in this so they can minimize involving animals as much as they can. Moreover, engineers on site have to continuously measure the humidity of the brain.
Surgical Improvements
3 optical paths are operated for surgery; visible imaging, laser interferometry and OCT-optical coherence tomography. This gives the real-time positioning of the brain as the brain is not stationary. It does this by employing lighting and illumination so thread insertion can be performed smoothly and vessel avoidance is achieved. Additionally, the skull thickness and hardness are not uniform therefore ultrasonic cutters and oscillating cutters are utilized. Oscillating cutters have the benefit of not cutting through soft tissue but only hard bone. However, these have to generate a large amount of heat to cut at the ideal rate, therefore it needs improving.
Upgradeability
After implantation, over time all the space will fill with tissue. The tissue that forms above the implant makes it challenging for the implant to be removed. The dura must be kept in place as a solution to maintain the brain's natural protective layer. This is less invasive, however, the problem with this is that the dura is a tough and opaque membrane and this is hard to pass through to insert electrodes. The dura is also in the way therefore a new optical system is required; it uses a medical standard fluorescent dye to image vessels underneath the tissue. This will allow surgeons to avoid the vasculature underneath the dura.
Microfabrication
The biological environments the needles are exposed to when they are fully implanted into the brain need to be more extensively studied. This will be useful to produce synthetic biological material that is made so that the testing process can be optimized. The surgical robot is doubled up to pick up and sensitive information that is needed from the biological environment of the brain and fed into a system. After data is collected, it is fed into the improvement process. The current proxy is slightly more complex as they have upgraded to a composite hydrogel base brain proxy. Threads are also inserted into lab-grown cerebral organoids.
Further implications
Neuralink has the potential to accomplish the inevitable extrapolation of human intelligence through technological means. The civilization that the brain has formed will only beget problems that are increasingly more complex to unravel. The brain can not only comprehend the progression of tribulations that we will have to face but also recognise that its credentials won’t be enough. This, however, is not a shortcoming of the brain, because only upon this evaluation, can we harness the elements of developing intelligence far beyond human intelligence; synthetic brains, man-made intellect; something we call artificial intelligence. Initially, combining our brilliance with technology will be a small step for science - yet a monumental step for mankind.
Comments