Tonight, Elon Musk has scheduled an event to discuss his plans for Neuralink, He announced a start-up company as early as 2017, and then fell silent. Now, if you go to the Neuralink website, you'll just find a vague description of the goal of developing an "extremely high-bandwidth brain-machine interface for connecting humans and computers." These interfaces have been under development for some time, typically under the pretense of brain-computer interfaces or BCIs. While there have been some notable achievements in the world of academic research, there is a remarkable shortage of products on the market.
The slow pace of progress is partly due to the fact that a successful BCI needs to tackle several difficult issues in part because regulatory and market conditions are challenging. Before today's announcement, we'll look at all these things and then see how Musk and the people who advise him decided to tackle them.
A Series of Problems
An effective BCI means figuring out how to get the nervous system to communicate with digital hardware. There are three issues that I will call reading, coding, and feedback. We will go over each of these points below.
The first step in a BCI is to find out what the brain is up to, which requires reading neuronal activity. While there have been some successes in doing so non-invasively using functional MRI, this is generally too blunt a tool. It does not have the resolution to figure out what small populations of cells do, and therefore can only provide a very approximate reading of the brain. For this reason, we are forced to opt for the alternative: invasive methods, especially the implantation of electrodes.
In the past, the electrodes were quite large compared to the cells they tracked, and therefore took inputs from a large number of population cell counterparts. The larger hardware meant that you could not put many electrodes on a single implant, and larger implants often came in contact with more than one region of the brain. To make matters worse, electrodes often develop scar tissue that interferes with our ability to read neuronal activity.
Our advanced manufacturing capabilities have gradually taken care of it. We can now make electrodes from biocompatible materials that limit scarring. The electrodes are finer and thus contact far fewer cells. And finally, the small size means that we can target the region of the brain we are interested in more closely.
But the challenge may be finding out which part of the brain we want to target. The anatomically visible features are often large and fulfill several functions. While we have a good understanding of the role of areas that handle things like motor control and visual input, there is much less information about what other areas, such as memory, actually do.
Plus brain activity is driven by the interaction of different regions. One might think that something like Parkinson's disease, which is characterized by tremor, comes from the area that controls muscle activity. They would be wrong, as the diagram in this article shows: The cells that die in Parkinson's patients are in an area involved in a complicated communication loop with at least six other areas of the brain.
In some cases There are other options. Amputees could work with an interface that reads the nerves immediately in front of the amputation site. The spinal cord is another potential site where information can be read. Of course, this has great advantages when it comes to avoiding the risk and complexity of introducing electrodes into the brain. But they are also a subset of people who could be helped with BCI technology.
Decrypt the Brain
Once we get on our nerves, we have to figure out what they are saying. Digital systems expect their data to be in an orderly series of voltage changes. That's not how the nerves work. Instead, they send a series of impulses; Information is encoded in a very analogous manner in the frequency, intensity and duration of these pulse trains. While this may seem manageable, there is no single code for the entire brain. A series of impulses from the visual centers is fundamentally different from the impulses that the hippocampus sends when recalling a memory.
So we need to find translations for each brain region we want to interact with. And, to a degree, even that will be variable, since individual cells within a particular region perform special functions and we can not say in advance which cells we will end up listening to.
There are possible ways to do it a bit easier. Neural networks are great for locating patterns in noisy data and can help us avoid having to understand specific code. And it would be easier if the process we want to hear could be controlled by a conscious patient, such as: B. the movements of the limbs. And again it should be a bit easier if we can intervene outside the brain. We have a good idea of which nerves control which limb muscles, for example, and possibly read from there.
One possible help with all this is that we do not necessarily have to retrieve things exactly right. The brain is a remarkably flexible organ that can re-learn how to control muscles after it's been damaged by things like a stroke. We may just have to tune the encoding to some extent, and then the brain adapts to give the BCI the input it needs to perform a task.
However, this requires some feedback. The brain needs to be aware of what it does right and what makes it bad when it wants to improve certain activities. Again, the example of limb movement is simple; The subject can easily observe what his thoughts are doing while controlling a prosthesis or robotic arm. But it is less true of many other things we may want to intervene with. Could a person who has always been blind understand how well she perceives visual input?
When we control motions, there are other levels of feedback as well. Let's say you want to get a cup of coffee. Normally you only have to look at it for a moment and then you can do the movements without watching it. That's because our body has a system that tracks where it's likely to have all of its parts (a sense called proprioception). We may also want to know if the cup feels hot when we grab it, and make sure that we only have enough strength to hold it, rather than crush it.
A truly effective BCI is not a one-man cup. But includes a series of bi-directional communications between the brain and the hardware with which it operates. And all these conversations will face the problems of reading and coding codes described above.
Almost ready for the market?
While all this makes progress seem like an impossibility, bear in mind that BCIs did some amazing things. as the following video shows. And implants that directly stimulate the nerves involved in hearing are now commonplace. There have also been successful demonstrations of retinal implants that stimulate the optic nerve to restore limited vision, promising work with movable prostheses and some deep brain stimulation applications in diseases such as Parkinson's and clinical depression. To a certain extent, the BCI already exists.
But it's also worth looking at the date in this video: 201
In other words, if Musk talks about what Neuralink will do in fifteen years, it should be treated The skepticism with which most people view their original timeline for landing on Mars.
But Neuralink is not just facing scientific hurdles. It is also expected to be a profitable company. And here are a number of additional hurdles.
Some of them are required by law. Such a thing would clearly bring FDA approval as an effective medical device. And legal risks for everything that goes along with brain surgery also pose a significant risk to a business. There is also the problem of market size. The number of people who are at least partially paralyzed due to spinal injuries, strokes and multiple sclerosis is significant in the US. However, serious problems would have to occur before a brain implant is a reasonable solution, and not everyone will be a good candidate. It's worth watching to see which products are probably the first ones from Neuralink as these will likely help determine if they survive long enough to enter the field of science fiction familiar territory for musk. The technology for the things shown in the video is clearly approaching the readiness for broader use, much as it was for reusable rockets and electric vehicles. There are also established players in the market for medical devices who would like to steal his lunch. There is currently no way to decide whether Neuralink is approaching SpaceX or Musk's efforts to enter the solar market.