Part One: The Human Colossus
Part Two: The Brain
Part Three: Flying Over the Nest of Neurons
Part four: neurocomputer interfaces
Part Five: The Neuaralink Problem
Part Six: Age of Wizards 1
Part Six: Age of Wizards 2
Part Seven: The Great Fusion
Promotional video:
In 1969, a scientist named Eberhard Fetz connected one neuron in a monkey's brain to a dial in front of its face. The arrows had to move when the neuron fired. When the monkey thought so that the neuron was activated and the arrows shifted, she received a banana-flavored candy. Over time, the monkey began to improve in this game, because he wanted more delicious sweets. The monkey learned to activate a separate neuron and became the first character to receive a neurocomputer interface.
Over the next several decades, progress was rather slow, but by the mid-90s, the situation began to change and since then everything has accelerated.
Since our understanding of the brain and electrode equipment is rather primitive, our efforts tend to be directed towards creating simple interfaces that will be used in the areas of the brain that we understand best, such as the motor cortex and visual cortex.
And since human experimentation is only possible for people who are trying to use NCI to alleviate their suffering - and because market demand is focused on this - our efforts have been almost entirely devoted to restoring lost functions for people with disabilities.
The largest NCI industries of the future, which will provide people with magical superpowers and transform the world, are now in a state of embryo - and we have to be guided by them, as well as our own guesses, thinking about what the world might be like in 2040, 2060 or 2100.
Let's go through them.
This is a computer created by Alan Turing in 1950. It's called Pilot ACE. A masterpiece of its time.
Now look at this:
As you read the examples below, I want you to keep this analogy in front of your eyes -
Pilot ACE is the same for iPhone 7
than
each example below is for _
- and try to imagine what a dash should be in place. We will return to it later.
In any case, of everything I've read and discussed with people in the field, there are currently three major categories of neural computer interfaces in development:
First NCI type # 1: using the motor cortex as a remote control
In case you forgot, the motor cortex is this guy:
Many areas of the brain are incomprehensible to us, but the motor cortex is less incomprehensible to us than others. And more importantly, it is well mapped, its individual parts control individual parts of the body.
Importantly, this is one of the large brain regions that is responsible for our work. When a person does something, the motor cortex almost certainly pulls the strings (at least the physical side of the action). Therefore, the human brain does not need to learn to use the motor cortex as a remote control, because the brain already uses it as such.
Raise your hand. Now put it down. See? Your hand is like a small toy drone, and your brain simply uses the motor cortex as a remote control to take the drone off and back.
The purpose of an NCI based on a motor cortex is to connect to it, and then, when the remote control triggers a command, hear that command and send it to some device that can respond to it. For example, on hand. A bundle of nerves is the intermediary between your cortex and your hand. NCI is an intermediary between your motor cortex and your computer. It's simple.
One of these types of interfaces allows a person - usually a person paralyzed from the neck or with an amputated limb - to move the cursor on the screen with their minds.
It all starts with a 100-pin multi-electrode matrix that is implanted into the human motor cortex. The motor cortex in a paralyzed person works great - just the spinal cord, which served as an intermediary between the cortex and the body, stopped working. Thus, with the implanted electrode array, the researchers allowed the person to move their arm in different directions. Even if he cannot do it, the motor cortex functions normally, as if he could.
When someone moves their arm, their motor cortex explodes with activity - but each neuron is usually only interested in one type of movement. Therefore, one neuron can fire whenever a person moves his hand to the right, but will get bored when moving in other directions. Then only one of this neuron could determine when a person wants to move his hand to the right, and when not. But with an electrode array of 100 electrodes, each will listen to a separate neuron. Therefore, during tests, when a person is asked to move his hand to the right, for example, 38 out of 100 neurons record the activity of neurons. When a person wants to move their hand to the left, 41 others are activated. In the process of practicing movements in different directions and at different speeds,the computer receives data from the electrodes and synthesizes them into a general understanding of the pattern of neuronal activation, corresponding to the intentions to move along the XY axes.
Then, when they display this data on a computer screen, a person can, by the power of thought, "trying" to move the cursor, actually control the cursor. And it works. BrainGate enabled the boy to play a video game with just the power of thought, using NCIs linked to the motor cortex.
And if 100 neurons can tell you where they want to move the cursor, why can't they tell you when they want to pick up their coffee and take a sip? This is what this paralyzed woman did:
Another paralyzed woman managed to fly in an F-35 fighter simulator, and a monkey recently rode in a wheelchair with the help of his brain.
And why be limited to only hands? NKI Brazilian pioneer Miguel Nicolelis and his team built an entire exoskeleton that allowed a paralyzed person to take the opening kick at the World Cup.
These developments contain the seeds of other future revolutionary technologies, such as brain-to-brain interfaces.
Nicolelis conducted an experiment in which the motor cortex of one rat in Brazil, which pressed one of two levers in a cage - one of which the rat knew it would enjoy - was connected via the Internet to the motor cortex of another rat in the United States. A rat in the United States was in a similar cage, except that, unlike a rat in Brazil, she had no information about which of her two levers would please her - other than the signals she received from the Brazilian rat. In the course of the experiment, if the American rat correctly selected the lever, the same one pulled by the rat in Brazil, both rats received a reward. If they pulled the wrong one, they didn't get it. Interestingly, over time, the rats got better and better, worked together, like one nervous system - although they had no idea about the existence of each other. The success of the American rat without information was 50%. With signals coming from the brain of the Brazilian rat, the success rate rose to 64%. Here's a video.
In part, it worked in humans as well. Two people in different buildings worked together while playing a video game. One saw the game, the other was holding a controller. Using simple EEG headsets, the player who saw the game could, without moving his arms, think of moving his hand to “shoot” the controller - and since their brains were communicating with each other, the player with the controller felt the signal in his finger and pressed the button.
First NCI type # 2: artificial ears and eyes
There are several reasons why giving sight to the blind and sound to the deaf are among the most accessible categories of neurocomputer interfaces.
First, like the motor cortex, the sensory cortex are parts of the brain that we understand quite well, in part because they tend to map well.
Second, among many of the first approaches, we did not need to deal with the brain - we could interact with the places where the ears and eyes connect to the brain, because this is where the disorders were most common.
And while the activity of the brain's motor cortex was primarily about reading neurons to extract information from the brain, artificial senses work differently - by stimulating neurons to send information inward.
Over the past decades, we have seen incredible development of cochlear implants.
A cochlear implant is a small computer that has a microphone at one end (which sits on your ear) and a wire at the other that connects to an array of electrodes lining the cochlea.
The sound enters the microphone (the little hook at the top of the ear) and goes into the brown thing, which processes the sound to filter out less useful frequencies. The brown thing then transmits the information through the skin, through electrical induction, to another component of the computer, which converts the information into electrical impulses and sends it to the cochlea. The electrodes filter impulses in frequency like a cochlea and stimulate the auditory nerve like the hairs in a cochlea. This is how it looks from the outside:
In other words, the artificial ear performs the same function of converting sound into impulses and transmitting it to the auditory nerve as the normal ear.
But this is not ideal. Why? Because in order to send sound to the brain with the same quality as a normal ear, you need 3500 electrodes. Most cochlear implants contain only 16. Rough.
But we are in the era of Pilot ACE - of course, rude.
Nevertheless, today's cochlear implant allows people to hear speech and speak, which is already good.
Many parents of deaf children get cochlear implants when they are one year old.
In the world of blindness, a similar revolution is taking place in the form of a retinal implant.
Blindness is often the result of retinal disease. In this case, the implant can perform a similar function for vision as a cochlear implant for hearing (though not so directly). It does the same thing as the normal eye, transmitting information to the nerves in the form of electrical impulses, just as the eyes do.
A more complex interface than a cochlear implant, the first retinal implant was approved by the FDA in 2011 - the Argus II implant made by Second Sight. The retinal implant looks like this:
And it works like this:
The retinal implant has 60 sensors. There are about a million neurons in the retina. Rough. But seeing blurry edges, shapes, play of light and darkness is much better than not seeing anything at all. What is especially interesting is that a million sensors are not needed to achieve good vision at all - modeling has suggested that 600-1000 electrodes will be enough for face recognition and reading.
First NCI type # 3: deep brain stimulation
Since the late 1980s, deep brain stimulation has become another crude tool that is still life-changing for many people.
It is also a category of NCIs that are not related to the outside world - this is the use of neurocomputer interfaces to heal or improve oneself by changing something inside.
What happens here is one or two electrode wires, usually with four separate electrode sites, that enter the brain and often end up somewhere in the limbic system. A small pacemaker is then implanted into the upper chest and connected to the electrodes. Like this:
The electrodes can then deliver a small charge as needed, which is useful for many important things. For instance:
- reduction of tremor in people with Parkinson's disease
- reducing the severity of attacks
- reduction of obsessive-compulsive disorder
Through experiments (that is, so far without FDA approval), scientists have been able to alleviate certain types of chronic pain, such as migraines or phantom pain in the limbs, cure anxiety or depression in PTSD, or, in combination with muscle stimulation, restore certain disturbed brain circuits that have broken down after stroke or neurological disease.
* * *
This is the state of the still underdeveloped area of the NCI. And at this moment Elon Musk enters it. For him and for Neuralink, the modern NCI industry is point A. While we have been studying the past throughout these articles to get to the present moment. Now it's time to look into the future - to find out what point B is and how we can get to it.
ILYA KHEL
Part One: The Human Colossus
Part Two: The Brain
Part Three: Flying Over the Nest of Neurons
Part four: neurocomputer interfaces
Part Five: The Neuaralink Problem
Part Six: Age of Wizards 1
Part Six: Age of Wizards 2
Part Seven: The Great Fusion