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:
Since I've already written about two of Elon Musk's companies - Tesla and SpaceX - I think I understand his formula. It looks like this:
And his first thought for a new company starts on the right and goes all the way to the left.
He decides that some specific changes in the world will increase the likelihood that humanity will have a better future. He knows that large-scale world change happens the fastest when the whole world - the Human Colossus - is working on it. And he knows that the Human Colossus will strive to achieve a goal if and only if there is an economic driving force - if the very process of spending resources to achieve this goal is good business.
Often, before a booming industry picks up steam, it’s all like a stack of logs - all the ingredients for the fire are in place, everything is ready to go, but no match. There is some technological deficit that is preventing the entire industry from taking off.
So when Elon starts a company, her main strategy is usually to create a match that will ignite the industry and get the Human Colossus to work on it. This, in turn, Elon believes, will lead to events that will change the world in ways that increase the likelihood that humanity will have a better future. But you need to look at his companies from a bird's eye view to understand all this. Otherwise, you will mistakenly think of everything he does as business as usual - when in fact what looks like a business will be a mechanism to support the company innovating to create a big match.
When I was working on articles about Tesla and SpaceX, I asked Elon why he got into engineering and not science, and he explained that when it comes to progress, "engineering is the limiting factor." In other words, the progress of science, business and industry - all this happens with the permission of technological progress. And if you look at history, it makes sense - since every greatest revolution in human progress is a technical breakthrough. Match.
So, to understand Elon Musk's company, you need to think about the match he is trying to create - along with three other variables:
And when I started thinking about what Neuralink is, I knew which variables I needed to set. At that time, I had a very vague idea about one of the variables - that the company's goal is "to accelerate the emergence of a brain-wide neural interface." Or a magic hat, as I called it.
As I understand it, the common brain interface was supposed to represent a neurocomputer interface in an ideal world - a super-duper-advanced concept where all the neurons in your brain can communicate invisibly with the world outside. This concept was based on the sci-fi idea of "neural lace" from the Culture series by Ian Banks - a weightless, intangible whole-brain interface that can be teleported to the brain.
I had plenty of questions.
Luckily, I was on my way to San Francisco, where I had to sit down with half of the Neuralink founding team and be the dumbest person in the room.
A digression on why I am not exaggerating by calling myself the dumbest person in that room, just see for yourself.
I asked Elon how he put his team together. He replied that he had met with literally 1000 people to get this group together, and part of the task was a huge number of completely separate areas of expertise that needed to be sorted out: neurobiology, neurosurgery, microscopic electronics, clinical trials, etc. Since this is an interdisciplinary field, he looked for interdisciplinary experts. And this can be seen in their biographies - all members of the group have a unique combination of knowledge that intersects with the knowledge of other members of the group and together constitute, as it were, a mega-expert. Elon also wanted to find people who could look down on the mission - who were more focused on industrial results than papermaking. In general, it was not easy.
But now they were sitting at a round table and looked at me. I was a little shocked because I had to do a lot of research before coming here. I got the thesis out of myself, they picked it up and quadrupled it. And while the discussion continued, I began to gradually understand what was what.
Over the next few weeks, I met with other founders, playing the role of a fool every time. In these meetings, I focused on trying to get a comprehensive picture of the challenges ahead and what the path to the magic hat would look like. I wanted to understand these two boxes:
The first one was simple. The business part of Neuralink is a neurocomputer interface company. They want to create cutting edge NCIs - some of them will be "micron-sized devices." This process will support the company's growth and provide an excellent base for innovation (much like the way SpaceX uses its launches to sustain the company and experiment with cutting-edge engineering).
As for the interface they plan to work on, here's what Elon says:
The second box was more difficult. Today it seems clear to us that the use of steam engine technology for the power of fire had to begin for the industrial revolution to take place. But if you talked to someone in 1760 about this, there would be much less clarity - what obstacles need to be overcome, what innovations to implement, how long it will take. And here we are, trying to figure out what the match that will ignite the neurorevolution should look like, and how to create it.
The starting point for a discussion of innovation will be a discussion of barriers - why innovation is needed at all. In the case of Neuralink, the list will be long. But even with engineering as the main constraint, there are a few major challenges that are unlikely to be the main obstacle:
Public skepticism
Recently, a poll was conducted in which it turned out that Americans fear the future of biotechnology, in particular - NCI, more than gene editing.
Flip Sabes doesn't share their concerns.
When a scientist thinks about changing the fundamental nature of life - about creating viruses, about eugenics, etc. - a spectrum is created that many biologists find quite alarming, but I know that when neuroscientists think about chips in the brain, they do not find it strange, because we already have chips in our brains. We have deep brain stimulation that relieves the symptoms of Parkinson's disease, we are conducting the first tests of chips to restore vision, we have a cochlear implant - it does not seem strange to us to put a device in the brain to read and write information.
And, having learned all about the chips in the brain, I agree - and when Americans learn everything about them, they will change their minds too.
History supports this prediction. People didn't get used to Lasik eye surgery very quickly when it first appeared - 20 years ago, only 20,000 people a year underwent surgery. Today this number is already 2,000,000. The same is true with pacemakers. And defibrillators. And organ transplants. But she once gave off Frankensteinism! Brain implants will be from the same opera.
Our misunderstanding of the brain
Remember, "if you think of a understood brain as one mile, we only walked three inches along it"? Flip thinks so too:
If we needed to understand the brain in order to interact with it in essence, we would have problems. But all these things in the brain can be deciphered without fully understanding the dynamics of computing in the brain. The ability to count all of this is a problem for engineers. The ability to understand the origin and organization of neurons in the smallest detail that would satisfy neuroscientists in full is a separate problem. And we don't need to solve all these scientific problems to make progress.
If we can simply make neurons talk to computers through technical methods, that will be enough and machine learning will take care of the rest. That is, it will teach us the science of the brain. As Flip notes:
The flip side of the phrase “we don't need to understand the brain to make progress” is that advances in engineering will almost certainly increase our scientific knowledge - much like Alpha Go will teach the world's best players the best strategies for playing Go. And this scientific progress will lead to even greater technological progress - engineering and science will push each other.
Evil giants
Tesla and SpaceX are both treading on very large tails (for example, the auto industry, oil and gas and military industrial complex). Big tails do not like to be stepped on, so they usually do everything possible to hinder the advance of the attacker. Fortunately, Neuralink doesn't have this problem. There is not a single major field of activity that Neuralink can destroy (at least in the foreseeable future - and there, a possible neurorevolution will disrupt almost every industry).
Neuralink obstacles are technological obstacles. There are many of them, but two of them stand alone, and if you overcome them, this may be enough for all the other walls to fall and completely change the trajectory of our future.
Big hurdle # 1: bandwidth
At the same time, the human brain has never had more than a couple of hundred electrodes. Compared to vision, this is equivalent to ultra-low resolution. Compared to the engine, these are the simplest commands with little control. Compared to thoughts, a few hundred electrodes will be enough just to convey a simple message.
We need higher bandwidth. Much taller.
Thinking about an interface that could change the world, the Neuralink team put the approximate number of "one million neurons read at the same time." They also say that 100,000 - this number will create many useful NCIs with various applications.
The first computers faced similar problems. Primitive transistors took up a lot of space and were difficult to scale. But in 1959 an integrated circuit appeared - a computer chip. Along with it, there was a way to increase the number of transistors and Moore's Law - the concept that the number of transistors that can fit on a computer chip doubles every 18 months.
Until the 90s, electrodes for NCI were made by hand. Then we started figuring out how to manufacture these tiny 100-electrode multi-electrode arrays using modern semiconductor technology. Ben Rapoport of Neuralink believes that "the shift from manual production to Utah Array electrodes was the first hint that Moore's Law could hold power in the NQI field."
This is a huge potential. Today, our maximum is a few hundred electrodes capable of measuring about 500 neurons simultaneously - this is far from a million, not even close. If we add 500 neurons every 18 months, we get to a million in 5017. If we double that number every 18 months, we get a million by 2034.
We are currently somewhere in between. Ian Stevenson and Konrad Kording published a paper in which they considered the maximum number of neurons that were read simultaneously at different times over the past 50 years (in any animals) and plotted the result on this graph:
This study, also called Stevenson's Law, suggests that the number of neurons that we can register at the same time appears to double every 7.4 years. If this indicator holds, by the end of this century we will be able to reach a million, and in 2225 - record every neuron in the brain and get our finished wizard hat.
In general, there is no ICI equivalent yet, because 7.4 years is too long to start a revolution. The breakthrough will come not from a device that can record a million neurons, but with a paradigm shift that will make this graph more like Moore's law and less like Stevenson's. Once that happens, millions of neurons will follow.
Big obstacle # 2: implantation
NCIs will not be able to take over the world if every time they have to open the skull.
This is an important topic at Neuralink. I think the word “non-invasive” or “non-invasive” was spoken about forty times during my conversations with the team.
In addition to being a major barrier to entry and a major safety issue, invasive brain surgery is costly and demanding. Elon said the final NCI implantation process should be automated. “A machine that can do this would have to be something like Lasik, an automated process - because otherwise you would be limited by the number of neurosurgeons and the costs would be too high. You need a Lasik-type machine to scale this process."
The development of high-throughput NKIs would be a breakthrough in itself, not to mention the development of non-invasive implants. But doing both will start a revolution.
Other obstacles
Today's NCI patients walk with a wire sticking out of their head. This certainly won't take off in the future. Neuralink plans to work on devices that will be wireless. But this is also fraught with problems. You need a device that can wirelessly transmit and receive tons of data. This means that it has to take care of such things as signal amplification, analog-to-digital conversion, and data compression on its own. And all this must also work on induction current.
Another big problem is biocompatibility. Sensitive electronics generally don't go well in a jelly ball. And the human body does not accept foreign objects into itself. But the brain interfaces of the future will have to work forever and without interruption. Consequently, the device will be hermetically sealed and secure enough to survive decades of buzzing and shifting neurons around. And the brain - which treats modern devices as intruders and covers them with scar tissue - will have to somehow be tricked into thinking that this device is a normal part of the brain.
There is also a problem with space. Where exactly will you place your device that will be able to interact with a million neurons in the skull, which already divides space into 100 billion neurons? A million electrodes using modern multi-electrode arrays will be the size of a baseball. Therefore, further miniaturization is another ongoing innovation to add to the list.
There is also the fact that modern electrodes are mostly optimized for simple electrical recording or simple electrical stimulation. If we really want an efficient interface, we need something other than single-function rigid electrodes - something with the mechanical complexity of neural circuits that can record and stimulate, and can also interact with neurons chemically, mechanically, and electrically.
And let's just put it all together perfectly - broadband, long-term, biocompatible, bi-directional, communicative, non-invasive implantable device. Now we can conduct a dialogue with a million neurons at the same time. Except … we don't know how to talk to neurons. It is not so easy to decipher the static flashes of hundreds of neurons, but we are, in fact, trying to study a set of specific flashes that respond to certain simple commands. It won't work with a million signals. An ordinary translator, in fact, uses two dictionaries, substituting words from one for another - but this does not mean understanding the language. We need to make a big leap in machine learning before the computer can learn a language, and even more leaps need to be made.to understand the language of the brain - because humans certainly won't learn to decode the code of a million simultaneously firing neurons.
Colonization of Mars seems simple now.
But I bet that the phone, the car, and the moon landing would have seemed like insurmountable technological challenges to humans decades earlier. And I'm willing to bet that it is -
- will seem completely insoluble for people of this time:
And yes, it's in your pocket. If the past has taught us anything, it is that there will always be technologies of the future, unthinkable for people of the past. We do not know which technologies, which seem completely impossible to us, will become ubiquitous in the future, but there will be such. People always underestimate the Human Colossus.
If everyone you know has electronics in their skulls by age 40, it will be thanks to a paradigm shift that has caused a fundamental shift in this entire industry. This shift is exactly what the Neuralink team is trying to organize. Other teams are working on this too, and some cool ideas have already begun to emerge:
Relevant innovations in the field of NCI
A group from the University of Illinois is developing a silk interface:
Silk can be folded into a thin bundle and inserted into the brain relatively non-invasively. There, it will theoretically expand and settle in the contours, like a shrink film. The silk will have flexible silicon transistor arrays.
In his TEDx Talk, Hong Yeo demonstrated an array of electrodes applied to his skin like a temporary tattoo, and scientists believe the technique could potentially be used in the brain:
Another group is working on a kind of nanoscale electrode neural mesh so tiny it can be injected into the brain with a syringe:
For comparison, this red tube on the right is the tip of a syringe.
Other non-invasive methods include vein and artery entry. Elon mentioned the following: “The least invasive method would be something like a solid stent that enters through the femoral artery and unfolds in the circulatory system to interact with neurons. Neurons use a lot of energy, so this is essentially a road grid to each neuron."
DARPA, the technology innovation arm of the US military, through the recently funded BRAIN program, is developing tiny “looped” neural implants that could replace drugs.
A second DARPA project aims to fit a million electrodes into a coin-sized device.
Another idea that is being worked on is transcranial magnetic stimulation (TMS), in which a magnetic coil outside the head can create electrical impulses inside the brain.
These impulses can stimulate targeted areas of neurons, providing a completely non-invasive type of deep brain stimulation.
One of the co-founders of Neuralink, DJ Seo, has put in an effort to develop an even cooler interface called neural dust. Neural dust is tiny silicon sensors measuring 100 microns (roughly the width of a hair) that must be injected directly into the cortex. Nearby, above the dura mater, there will be a 3-millimeter device that can interact with sensors in the dust using ultrasound.
This is yet another example of the innovative benefits derived from a multidisciplinary team. DeJ explained to me that "there are technologies that are not thought of at all in this area, but we can bring some of the principles of their work into it." He says the neural dust was inspired by the principles of microchip and RFID technologies. You can easily see how the cross-effects of different fields work:
Others are working on even more incredible ideas, such as optogenetics (where you inject a virus that attaches to a brain cell, causing it to be stimulated by light afterwards) or the use of carbon nanotubes, a million of which can be linked together and sent to the brain through the bloodstream.
These people are working on innovation in the company.
This is a relatively small group now, but when the breakthrough really starts to make itself felt, that will quickly change. Events will begin to develop rapidly. The brain bandwidth will get better and better as implantation procedures become easier and cheaper. Public interest will arise. And when public interest picks up speed, the Human Colossus will notice an opportunity - and then the speed of development will jump to the skies. In the same way that breakthroughs in computer hardware led to the development of software, so large industries will get involved in the development of smart applications and advanced machines that work in conjunction with neurocomputer interfaces. Sometime in 2052, you’ll tell some kid about how it all started and they’ll get bored.
I was trying to get the Neuralink team to talk about 2052 with me. I wanted to know what would happen when this all came true. I wanted to know what they themselves would like to put in place a dash. But it was not easy - after all, this team was created specifically to focus on concrete results, and not on empty words.
But I kept asking, gritting my teeth, until they expressed their thoughts for the future. I also spent most of my distant future conversations with Elon and Moran Cerf, a neuroscientist who works on NCI and thinks a lot about the long-term implications of their work. Finally, a member of the Neuralink team told me that of course he and his colleagues have a lot of dreams - otherwise they wouldn't do what they do - and that many things in their field were influenced by science fiction. He recommended that I talk to Ramez Naam, author of the famous Nexus trilogy. So I asked 435 questions to Ramez to get a complete picture.
As a result of this conversation, I left completely dead. I once wrote what it would be like if we went back to 1750 - when there was still no electricity, motors or telecommunications - pulling out George Washington and showing him our modern world. He will be so shocked that he will die. Then I started thinking about the concept of how many years into the future you need to go to experience the fatal shock of progress. I named it the Point of Death Progress (TPP).
Since the birth of the Human Colossus, our world has acquired a strange property - over time it becomes more and more magical. This is how merchant works. And as development generates even more rapid development, the trend is that over time, the TSP is getting closer and shorter. For George Washington, the TSP was several hundred years old, which is not so much in the general scheme of human history. But now we live in a time when everything is flying so fast that we may experience several TSPs in our lifetime. The volume of everything that happened from 1750 to 2017 can be repeated already during your life, and more than once. This is a magical time to live - and it is difficult to understand, difficult to notice, because the life we live we live from within.
Anyway, I think a lot about TSP and always wonder what it would be like if we went ahead in a time machine and experienced what George would experience here. What should the future be like for me to die of shock? You can talk about things like artificial intelligence and gene editing, and I have no doubt that progress in these areas could lead to my death from shock. But the phrase "who knows how it will be" has never been descriptive.
I believe I may finally have a descriptive picture of our shocking future. Let me outline it for you.
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