This video is sponsored by CuriosityStream. Get access to my streaming video service, Nebula, when you sign up for CuriosityStream using the link in the description. In the future we may be able to control machines with our minds… though that may open the door to others being able to use machines to control our minds. So today we’ll be looking at BCI, Brain Computer Interfaces, and updating on the progress that’s been made since we last looked at the topic in mind-machine interfaces a couple years back. And we’re going to dig into how BCI will work, beyond the usual assumption of us just jamming wires into someone’s head.
We’ll look at the impact BCIs are going to have as they rollout over the next couple decades. Make no mistake either, these are coming and sooner than later, your ability to think stuff directly into a computer and vice-versa, computers thinking stuff directly into you. Or more directly anyway. As I often mention in discussions of this topic, Technological Telepathy or Brain-to-Brain interfaces is something we already have.
You send a signal from your brain to your finger to hit a key which inputs into a computer that does something and displays that effect to a monitor which transmits to your eye, which your brain then sees and decodes and interprets. BCI is about cutting down on the steps or delays, and there are three primary ways that happens. The first is by cutting down on the number of middlemen, an example of which would be thinking letters or words onto a screen instead of sending a signal to your fingers and from them to their keyboard.
The second is cutting down on delays, most folks speak 4 to 5 times faster than they type, and that’s only including folks who type a lot. The record typing speed incidentally is 212 words a minute, set by Barabra Blackburn in 2005, which is faster than most folks talk, but the record speaking rate, held by Steven Woodmore, is still more than 3 times that, 637 words per minute. However speaking is still sending a signal to your mouth, tongue, throat, lips, etc, so we’ve got delays from these extra steps.
The third is all that encoding, decoding, and interpreting. You sending a signal to a hand and finger to find the right key, often including glancing at the keyboard to check position, is no easy thing compared to just thinking the letters A or B or C. Nor is it any small task to control your vocal cords or interpret sound vibrations or images as concepts. They only seem simple to us from a lifetime of practice.
There’s a lot of time and effort spent on that, and also a lot spent on conceptual compression, trying to express a concept accurately with as little time and effort as possible, which often results in confusion and misinterpretation by others, and requires much standardization of arbitrary words and phrases to minimize that. So BCI is about trying to cut down on all those steps, those middlemen, who introduce more time, more effort, and more confusion into our ability to interact with the Universe. Or with each other, and that’s another application of BCI, and probably a more valuable application than most, how it may speed up and improve human to human interaction, which if done with a computer as the lone middleman might be a Brain to Computer to Brain Interface, BCBI or Technological Telepathy though that probably doesn’t properly represent the notion. How does this get accomplished, this direct brain to machine interface? Well, at the moment it is indeed mostly done by jamming wires into someone’s brain, usually a rat. An image that always comes to mind for me is Neuralink’s rat with a chip on top of its skull that looks like a mohawk. We call this approach invasive BCI, and as we’ve discussed before, this offers the option of higher accuracy but I suspect it will not compete well with non-invasive options unless vastly superior.
We often see folks in science fiction with ‘neural jacks’, some port a wire can be stuck into to hook them up to their interface, but I suspect in the long run BCI will either be non-invasive scanners, or something surgical but akin to threading a detector mesh under the skull or throughout the brain so that there’s nothing sticking through the skin and no physical link. The big news for today, as I write this episode on April 12th, is that Elon Musk’s company Neuralink has gotten a monkey named Pager implanted with a chip, and as Musk tweeted “A monkey is literally playing a video game telepathically using a brain chip!!” On the one hand that really is impressive, on the other hand, that’s not exactly new. We’ve been doing monkey tests of computer cursor control for a couple decades now, but it’s a big step up. Pager had a chip implanted then they mapped his signals while he played games with a joystick and bribed him with banana smoothies, they eventually unplugged the joystick, which Pager used out of habit still but the motion was being mentally controlled, not by the joystick, then they had him playing Pong again.
It's amazing overall progress, though maybe a little overhyped, again we’ve had basic cursor control for years and it's not really superior to options like following someone’s eyes or just holding a mouse or joystick, making it principally of interest for folks with disabilities, though that is obviously still of amazing value. So while maybe a bit overhyped it is still progress of an impressive caliber. A lot of that work is also about making sure nothing sticks through the skin, which is your big infection risk, and they’ve had good luck implanting chips into pig’s brains with nothing needing to poke through to outside. So how was it done? It varies by experiment and approach but the one being used by Neuralink is to stick wires thinner than a hair into various areas of the brain that control movement. These areas are used because the test subjects are animals, and while you can’t ask animals questions about what they’re thinking or visualizing -- or at least can’t expect clear answers -- you can see the animals move. Our brains – and our animal cousins’ brains – work on sending electric signals around so the wire can pick up a change to that in that portion of the brain.
You put a bunch of these in throughout the motor control area and you record which are changing by how much as the test subject, be it monkey, rat, or pig, moves around or performs some task. There are applications for humans of course, like human prosthetics, but much of this is testing out the basic interface and its insertion and maintenance. When they’re good enough that we feel comfortable putting them into a healthy human skull simply for experimentation, then we can start doing things like asking abstract questions or holding conversations and seeing what bits of the brain light up. Neuralink isn’t the only group looking at these things, nor the first, but let’s continue discussing their tech. The link is a sealed, implanted device with many “neural threads” coming out of it.
The implants process, stimulate, and transmit neural signals, and they are inserted in various regions of the brain. They are far too numerous and delicate for a human surgeon to do the work, so a lot of the research has been on making a robotic surgeon that can handle the implantation. In theory stimulating the brain by running current through the wires will allow feedback. To insert or effect the brain rather than read it, which is the second half of the task, computer to brain interfacing. The near term goal for the link is one with 1024 threads or electrodes.
Since you’re a human you could learn how to push each mentally, and be able to start recognizing and quickly mentally stabbing those in two or three key combinations to produce desired and programmable effects. If that sounds hard keep in mind that it’s no different, and indeed probably easier, than learning to control your fingers or type on a keyboard or use any of those devices we all have that took a few days to get used to but afterward through practice became second nature. In that same way, most of us by adulthood can glance at a big block of text and rapidly decipher it into words with little effort or attention, and decipher sound or video into concepts or words far younger. So while the long term goal of BCI, from an input standpoint, would be to funnel an image right into your brain, not via your eyes, or a word into your brain, not via your ears or eyes, you could probably learn pretty quickly and intuitively how to feel those inputs on those same wires you were using to send signals out to the link. Some options might lead to more intuitive controls than arbitrary assigned buttons though, like a remote.
Learning to stimulate wires 3, 72, and 997 to turn on a light and 3, 42, and 654 to turn one off is definitely a practice and memorization thing, though again probably a lot easier than learning to recognize a light switch and building up the habit of physically searching that spot near the door they tend to be at when hunting for one in a dark room. However if you have a lot of light switches in a room, or lights, you might have the default on/off light wire combo that combines with something more intuitive like tracking where you’re looking for the link or interface to figure out which light you’re trying to turn on. Interface options that were fairly set, on/off, brighter or louder vs quieter or dimmer, next and back, left or right mouse button, combined with a cursor control or something monitoring where your eyes are looking at might become a standardized approach. Now truth be told turning lights on or off as you enter a room by mental command is just one step up from voice command options, which are nice and handy but honestly not that handy in day to day life compared to remotes or manual control, and we are talking about jamming wires into your skull, which is a pretty big investment financially and emotionally for folks for a minor convenience, even is jamming wires into your skull is rather hyperbolic. Alternatively learning to type 5 times faster mentally might be well worth it for a lot of folks, for instance, and the financial and emotional investment is just the current or near-term entry fee.
Down the road it is likely to be pretty cheap and easy and routine. To get there though we have a lot of barriers, and let’s discuss those. This sort of interface, threading some wires into the brain to a receiver and transmitter, is conceptually easy enough. Our brains send signals around in the microvolt range and by itself this is easy to read and making wires that small is all stuff we could have done at any point in the last century, and indeed frequently did.
But to do it right, to have long term subtle implants, it means getting our wires compatible with the tissues near them, namely neurons. We want electrodes with the scale and flexibility of those neighboring neurons, protected from corrosion of the surrounding fluids or from doing any damage to those neurons. We need something that takes a long time to break down, which is tricky for a wire thinner than a human hair, and which decays into something harmless, or which doesn’t breakdown and is easily removable if damaged.. We also need a safe way to power the device. Hypothetically the electric signals in your brain can power it, but probably not strong enough to transmit wireless through the skull, and probably not to allow any feedback, the wires sending signals down those parts of the brain.
So we need to decrease power demands or find a way to generate power locally. This of course is part of the appeal to a physical link in the back of the skull, so you can run power in that way, and we probably can handle the infection concern of implants piercing the skin someday. We examined some of those options in our recent Biotech episode. We also can transmit power wirelessly, not just information, and the low wattage need of the brain implants might make a wireless charger in your shirt’s collar a perfectly viable power supply option. Still, our brain gets power, indeed it’s the most energy hungry organ of our body, so we could probably hijack some of that fuel to run the device. Our brain runs on the equivalent of 20 watts, devices like this would be expected to run on an order of magnitude less power, so hijacking some of the fuel going up to the brain, so long as we stimulated the body to send more fuel along, is one possible way to power such implants and one we often consider for other implants outside the brain too.
Adenosine Triphosphate, or ATP, the powerhouse of the biological cell, is potentially usable to run mechanical and electronic devices too. We also use a lot of ATP, a single cortical neuron burns through nearly 300 billion ATP molecules a minute in your brain. There’s potentially a lot of advantages to being able to use ATP to run internal devices, and to being able to increase ATP production in the body, for that matter. Ultimately one of our goals will probably be to grow electrodes very similar to neurons, potentially even artificial neurons. Those come in a big array of sizes and shapes but the four basic types are Unipolar, Bipolar, Multipolar, and Pseudo-unipolar.
Most sensory cells are pseudo-unipolar. These are up on the screen for those watching rather than listening, but a unipolar neuron looks a lot like a small tree in a pot, waiting transplant, with the pot being the Soma, the main chunk of the cell containing the nucleus, and the trunk coming out of the pot is the Axon, or nerve fiber, and the branches sprouting out are the dendrites. A bipolar neuron is the same but has two trunks or axons sticking out of the soma or pot, each with dendrites, while a multipolar is more like bush, lots of small little trunks, axons, each with dendrites. Pseudo-unipolar neurons have one trunk that splits into two extensions, each with their own dendrites, rather than two sticking out from the soma itself, and again are the main sensory neurons. One of the sets of dendrites is receiving sensory information and the other transmitting it, and they would be of particular interest to us further ahead in BCI technology when we get to the stage we can play with individual neurons and want to be sending and transmitting high resolution sensory data. You could run a wire to each side of each sensory pseudo-unipolar neuron for instance, and that is physically possible and doable with even fairly modest nanotechnology and nano-robotics, but we are not anywhere near that good yet.
That is the level of technology though that allows someone to exist in either a virtual reality or augmented reality while receiving vision, hearing, taste, smell, and touch that’s indistinguishable from reality. A little more on neurons. Talking about putting a wire down to each one of them starts sounding like some serious micro-sizing down to the atomic scale, there are a hundred billion neurons in the typical brain after all, but as is often the case with scale in science and futurism topics, that can be misleading. Your brain has a hundred billion neurons, but each composed of something like a million-billion atoms itself.
Neurons are on the big side for biological cells, and as I often say in regard to cells, the difference between a cell and an atom is the difference between a city and the individual bricks composing its many buildings. Neurons are gigantic, some can be nearly a millimeter long, which is huge for a cell, but that’s actually on the tiny side of neurons, some of which are meters long. And they are quite tangled up, so if you’re trying to identify them individually or trace them, you need some very skinny and flexible wires. Of course this allows another potential aspect of BCI, which is neuron replacement. If you’re tracing out and connecting up to individual neurons – which again is way down the road but still on the horizon – then you could repair, augment, or replace them with either cloned, grown, or entirely synthetic versions.
Fundamentally though for all of this, your three big hurdles are first: making the devices sturdy enough, small enough, and compatible enough we can use them in the brain, second getting them power and replacing and maintaining them, and third making everyone comfortable with the idea. The entry level of this probably isn’t some mental cursor or keyboard for ordinary people looking to speed up their work or easy of use with devices, but for folks with a handicap, so we don’t need to sell the idea to folks in the context of augmentation of the ordinary to do our R&D and prototyping. We focus on folks with improving prosthetics, motor control impairments, or sensory impairments like being deaf or blind. From there as the technology improves and gets to be mundane, then folks can contemplate replacing their tons of remotes or Alexa devices with a brain chip. That probably handles most of the comfort issue with folks using it, that for most people it would be a fairly mundane part of the technological landscape like pacemakers or hearing aids are. The security angle though is more worrisome because if you can run inputs into and out of the brain then neuro-hacking becomes an option.
Probably not much of one though, at least in terms of doing it covertly. There wouldn’t be much defense against someone grabbing you, strapping you down and physically taking over your implants but then they could, in that situation, just give you implants if you didn’t already have them too. They could also use classic means of coercion, so new avenue, old problem, in that particular case, and one might say the silver lining of neuro-hacking is that it’s probably a lot less destructive than torture or drugging someone and probably more effective. I mention that since it's very likely that psychologists and law enforcement and intelligence agencies, all having a vested interest in such technology, would develop fairly sophisticated ways of hacking or using them.
Even if we gave them the benefit of the doubt and assumed that was always open, benevolent, and ethical use, it leaves the technology out there for folks with less altruistic or sinister motives to employ it. I don’t want to focus overly much on this, but basic security-wise, your wireless implant can employ a lot of cybersecurity options to be essentially unhackable. Hollywood and human laziness tend to lead most folks to think anything can be hacked, that’s just not so, and even most of the time you hear about some government agency getting hacked its something trivial like their facebook page not their data and records, but that does happen and it's typically either bad security habits, like leaving your password as 1234 or writing it on a post-it note and pinning it to your console. The other option is inside jobs, and presumably wouldn’t apply to you and your own brain, but there’s plenty of folks who might need or be given access to your implants and for that matter once we start talking neural augmentation we need to consider weird options like being hacked by your own split personality or sentient AI virtual assistant. Short form, there are genuine security concerns but they’re not really a barrier to employing the technology. Again if you can start doing this cheap and easy enough most folks can afford a brain implant it starts implying you can hack someone’s brain by sneaking a little robot into their bedroom and implanting them when they’re asleep too.
On the alternative side, just hacking someone’s encoded link by typing in a username and password is probably not going to be a concern. So how about applications? Near-term the big one’s are allowing those sensory hook ups in a general sort of way, not individual electrodes to some millions of pseudo-unipolar neurons but the basic that allows us to give basic sensory inputs for lost limbs or damaged sensory organs via prosthetics. The sorts of controls that allow that next step up represent a simple cursor, but keep in mind that even just a mouse and keyboard in your brain is a very handy thing and that is now technology we have, and have without needing to plug you in to something immobile to use it. We also do have non-invasive options and I’d not be too surprised if we started getting something like a hair net close to the scalp that was taking non-invasive reading to permit that sort of simple keyboard and cursor, and which might become as ubiquitous and mundane in a decade or two as smart phones and fitbits or watches are. Longer term it's everything from being able to watch TV in your brain with no screen nearby, or total sensory immersion in virtual reality, to adding the machine to your mind. Not brain-computer interfaces but computer peripherals or add-ons to the brain.
Imagine not even needing to think ‘Calculator, what is 987x251?’ but just to be able to look at a field and know instantly and without effort that it was 987 meters long by 251 meters wide, and thus 247,737 square meters, or 24.8 hectares, 61.3 acres, and that even as casually as you know some of the objects on it were trees or some wheat, your brain augments instantly and seamlessly informed you what the various obscure plants were and many other bits of data. See our episodes on Mind Augmentation, Cyborgs, or Superpowers for more ideas of all the neat things such seamless brain-computer interfacing might permit. One of those options is brain to brain too, potentially allowing concepts like Hive Minds to exist, but also to allow you to talk to someone else as seamlessly and easily and quickly as you can think of concepts, just relaying images instead of describing them for instance. That’s a ways off but make no mistake, BCI isn’t the technology of the future, it’s the technology of the present, and I would not be surprised if it was to the next generation what smartphones are to the current one and personal computers were to the last.
For better or for worse, 30 years from now keyboard and mice will probably be about as rare as they were 30 years ago, when many but not most homes and workplaces had computers. This might be invasive or non-invasive, or rely on secondary options like blinking or moving your eyes instead of your hands and fingers, and indeed it's likely to depend on the individual’s preferences. The final big consideration then might be how we handled all the potential clutter issues, of things like instant seamless mental access to every encyclopedia ever composed and personal recordings from your own eyeballs or cameras of any moment of your life. That seems such an incredible boon and incredible burden all at the same time.
It is an interesting question for discussion then, assuming it was safe medically and in terms of security, and affordable, would you be okay with an invasive brain implant that gave you a BCI? If yes, what feature was most attractive to you about it, if no, why not? Feel free to type your thoughts on the matter in the comments below, using the primitive keyboard BCI on your computer or smartphone. I mentioned in today’s episode the idea that you might get hacked by your own split personality or virtual assistant, and that is an example of a growing number of biotechnology ethical concerns we’re encountering that are discussed in “The Ethics of Biologically Inspired Technology” over on Curiositystream. I also thought we should spend a little time discussing this notion of BCI or Mind Augmentation-based split personalities, so we’ll be doing that in an extended edition of today’s episode over on Nebula.
Nebula is our rapidly-growing streaming service where you can see all of our episodes ad and sponsor free and a couple days early, as well as some bonus content like our extended editions or Nebula Exclusives like our Coexistence with Alien series. You can also see tons of content from many other amazing creators and help support this show while you’re at it. Now you can subscribe to Nebula all by itself but we have partnered up with CuriosityStream, the home of thousands of great educational videos, to offer Nebula for free as a bonus if you sign up for CuriosityStream using the link in our episode description. That lets you see content like “The Ethics of Biologically Inspired Technology”, and watch all the other amazing content on Curiositystream, and also all the great content over on Nebula from myself and many others. And you can get all that for less than $15 by using the link in the episode’s description. So that will wrap us up for today but not for the Month, as this Sunday we will be having our end of the month livestream Q&A.
July will be a busy month, starting July 1st with a long-requested return to our Faster Than Light series for Cheating Reality. From there will move on to Strip Mining the Galaxy, our second Galactic Domination episode on July 8th, then our scifi Sunday mid-month episode, Annoying Aliens. Then we’ll discuss if we should go to the Moon or Mars First, what the End of the Earth might actually be like, and what if someone turned the entire galaxy into their personal laboratory. If you want alerts when those and other episodes come out, make sure to subscribe to the channel, and if you’d like to help support future episodes, you can donate to us on Patreon, or our website, IsaacArthur.net, which are linked in the episode description below, along with all of our various social media forums where you can get updates and chat with others about the concepts in the episodes and many other futuristic ideas. You can also follow us iTunes, Soundcloud, or Spotify to get our audio-only versions of the show.
Until next time, thanks for watching, and have a great week!
2021-06-27