Neuralink: The Future of BCIs
Imagine having thousands of tiny electrodes all held together by thin fibers implanted into your head, constantly reading the information coming from you brain. Sounds like something straight from a movie, right? Well that kind of technology may be closer than you think. Everyday, Tech companies are getting closer and closer to the melding of mind and machine, but the one who is on the edge of innovation is Neuralink. Using Brain Computer Interfaces (BCIs), Neuralink wants to combine brain and machine so that we can harness the power of advanced technology. However there is a lot of complexity when it comes to this topic, so I wanted to try and breakdown all of Neuralinks main advancements into a simplified explanation of what this company has to offer.
Brain Computer Interfaces
Before understanding the specifics of the BCIs used by Neuralink, we first should get a basic understanding of what a BCI really is. A BCI, or Brain Computer Interface, is a machine that uses some method to read our brain signals and then upload it to a computer. This data can then be processed and used for various different tasks, such as movement of a prosthetic limb or finding the correlation between some actions and signals.
There are various ways for these brain signals, which are usually electrical, to be recorded. One machine used is an EEG cap, a cap that has a bunch of electrodes on it so that it can read postsynaptic potentials from a cluster of neurons. This is an example of a non-invasive BCI, which basically means it doesn’t go inside the body. This does have its setbacks, as the machine cannot pinpoint single neurons firing, and it cannot read that deep into the brain.
This is where electrocorticography (ECoG) comes in, an invasive alternative to the EEG. Instead of electrodes being placed on the scalp of the patient, when using ECoG the electrodes are placed directly on the surface of the brain. This leads to a higher spatial resolution and higher signal to noise ratio than the EEG.
Overall, these BCIs have some crazy applications. They can be used to turn out basic commands in a video game, used to treat patients with epilepsy, and even control prosthetics with just the mind. It’s clear why Neuralink wanted to delve into the subject of BCIs and why they are trying to go further than ever before.
1000x the Electrodes
Let’s look at arguably the biggest advancement that Neuralink has made on it’s path of mind and machine unification. When it comes to common BCIs we usually see around 10–20 electrodes fitted into the machine. This is sufficient if all that you want to look at are clustered brain waves or small parts of the brain. However Neuralink wanted more, they decided to up the amount of electrodes by a factor of 1000!
But why so many? Well if you want to map out and record the signals coming from every neuron in your brain, you would need a lot of electrodes to be able to collect that data. This is the main reasoning behind the large number of electrodes, Neuralink wants to be able to gather all the data, not vague data or data that is limited to one area.
From the image above, it is clear that these electrodes are tiny, and no those threads aren’t each electrodes. Each thread actually hold 32 electrodes, crazy right? So why do these electrodes need to be so tiny? Well if you want to have these electrodes up close to your neurons you need them to be on the same scale. If these electrodes were any bigger they probably wouldn’t be able to fit into your brain and safely record data. So now we have a bunch of tiny electrodes ready for insertion into the brain, but since they’re so small we need a way to safely and precisely place them into our heads. That is where Neuralinks next innovation comes into play.
Since the threads of electrons that need to be inserted into the brain are on such a small scale, Neuralink needed to think of a more precise way of inserting all the threads. A normal human would not be able to make such precise incisions and insertions, since the threads are too flexible to properly insert with out damaging the inside of your head. This is where robotic insertion comes into play. Since robots can be programmed to find precise locations to insert small things, such as these electrodes, the safest approach for insertion would be using a robot to perform the operation. The actual needle of the robot is miniscule allowing it to be on the same level as the small threads.
The tip of the needle is designed in a way so that it can both hook onto the electrode threads and also pierce brain tissue. The reason the needle would need to hook onto the threads is so that it can move them around during insertion, and the reason it needs to pierce brain tissue is because it needs to properly insert the threads into the brain. Before insertion even happens, the robot first maps out the brain by using a coordinate frame and depth tracking to see where certain brain structures are. This is crucial, because the whole reason we need this precise approach is so that we do not affect the blood-brain barrier since that will lead to various harmful problems in your head.
Some people may be scared of the thought of leaving their lives in the “hands” of a surgical robot, and while the whole operation can be fully automated, there will always be a human surgeon monitoring the robot through an image stick that is with the needle. If need be, the surgeon can also manually make adjustments to insertions if he sees fit. Finally, we have got the threads into the brain, so now what? How are these electrodes supposed to record data and send it back to an external computer, well that is where the electronics developed by Neuralink come in.
Reading the Brain
Now that everything is set-up inside you brain, it is time for us to get the data from your neurons, but how do we do that? Let’s quickly look at the basics, your neurons fire signals all the time so that you brain can send out or receive information. These electrical signals are what hold the functions or information we are looking for, so using our large amount of electrodes, we can pick up these signals and record them.
Since there are usually not that many electrodes to record data the amount of channels that are needed for collecting the data are small in number as well. However, in this case where there are thousands of electrode sites, Neuralink will need as many channels so that the data can be streamed to an external computer. This leads to there needing to be upgrades with certain specs to make this all possible.
To accommodate for the large amount of channels in the chip, the signal amplifier and the digitization stack are both incorporated into the chip. This allows for most of the processing to be done in the chip so cables aren’t needed for all of the electrodes. The amplifier needs to amplify the super small neural signal while rejecting out-of-band noise, so that the signals can actually be processed. The digitization stack then samples and digitizes the amplified signals so that they can then be streamed out to the computer. This whole process is done on a very small scale to match the scale of the chip. All of these advancements allows for the chip to properly gather all the specific information from each neuron and process for external use.
Now that we have gone over Neuralinks more important advancements, let’s quickly recap each ones functions:
- 1000x Electrodes=More sites for data to be recorded, and more specific information coming from single neurons.
- Robot Surgeon=Allows for precise incisions and proper insertion of electrode threads.
- Electronics (Channels, Amplifiers, Digitizer)=Allows for the specific data from the large amount of electrode sites to be properly recorded and processed for external use.
These key innovations truly show why Neuralink is one of the companies that are on the top of their industry, and I hope that through this article you can understand what they are trying to create.
If you want the specifics of what Neuralink is building, check out this article: https://www.biorxiv.org/content/10.1101/703801v4.full
And to see Elon’s initial presentation of Neuralink, check out this video: