- 4 months ago
Neurosurgeon and Engineer Dr. Ben Rapoport, co-founder of Precision Neuroscience, joins WIRED to answer the internet's burning questions about the emerging technology of brain implants and brain-computer interfaces. How does brain-computer interface technology work? What part of the brain is the BCI interfacing with? How do you get a chip into the brain in the first place? Will brain-computer interfaces one day allow us to replay our dreams? Answers to these questions and many more await on Brain-Computer Interface Support.
Director: Justin Wolfson
Director of Photography: Charlie Jordan
Editor: Richard Trammell
Expert: Ben Rapoport
Line Producer: Jamie Rasmussen
Associate Producer: Paul Gulyas; Brandon White
Production Manager: Peter Brunette
Production Coordinator: Rhyan Lark
Camera Operator: Chloe Ramos
Sound Mixer: Sean Paulsen
Post Production Supervisor: Christian Olguin
Supervising Editor: Eduardo Araujo
Assistant Editor: Justin Symonds
Director: Justin Wolfson
Director of Photography: Charlie Jordan
Editor: Richard Trammell
Expert: Ben Rapoport
Line Producer: Jamie Rasmussen
Associate Producer: Paul Gulyas; Brandon White
Production Manager: Peter Brunette
Production Coordinator: Rhyan Lark
Camera Operator: Chloe Ramos
Sound Mixer: Sean Paulsen
Post Production Supervisor: Christian Olguin
Supervising Editor: Eduardo Araujo
Assistant Editor: Justin Symonds
Category
🤖
TechTranscript
00:00I'm Ben Rappaport. I'm an electrical engineer and neurosurgeon and I'm here today to answer
00:04your questions from the internet. This is Brain Computer Interface Support.
00:12Whitney Cherian asks, how does brain computer interface technology work? Brain computer
00:15interface technology uses the fact that the brain communicates with itself and with the outside
00:20world using electrical signals. And so brain computer interfaces are implants that use tiny
00:25little electrodes that touch the brain and transform the electrical signals from the brain
00:30into ways of interacting with computers and external devices. That translation of electrical
00:35signals into useful means of communication with the outside world takes place using machine learning
00:40algorithms that transform the digitized bit streams from brain electrical data into means of communication
00:47with computers, smart devices, and in some cases robotics. InskaGator asks, could somebody, anybody,
00:55explain to me why a human being would even want a brain computer interface implanted in him slash her?
01:00I only ask because it's happened. Well, I think it's important to understand that one of the early
01:04uses of brain computer interface technology is especially for people with disorders of the brain
01:09and nervous system. And I'm talking about people with spinal cord injury, stroke, and some forms of
01:14neurodegenerative disease such as ALS, conditions that paralyze people and leave people with totally
01:19functioning mind unable to interact with the world in the ways that many of us take for granted.
01:24The first generation of brain computer interface technology is really geared towards enabling
01:29people with those kinds of conditions to interact with other people, with the outside world, return
01:34to work, and have a sense of dignity and independence that many of us take for granted. Govald asks,
01:39can you move a cursor in your mind or draw using a cursor? That's how BCIs for blind people work right
01:44now. Everything needs to be drawn, aka sequential, instead of all at once. Let me try to unpack this
01:49question. How do you move a cursor with your mind? Or how do people with brain computer interfaces learn
01:55to move a cursor using their mind? For each person who's had this experience, it's a little bit
01:59different. At the beginning, we usually provide an instruction. Think about moving a joystick or think
02:04about moving your hand or think about moving your arm. At first, it's very laborious. And eventually,
02:09the brain just connects to the cursor, like you learn to use a tool, a pencil, a baseball bat, or riding a
02:15bicycle. So that the brain computer interface is a tool like any other. And we're still really
02:20learning what that experience is like and how the brain accomplishes it. But that is the subjective
02:25experience that people who've used this technology explain, that it sort of clicks at some point and
02:30it starts to feel like magic. You know, when we interact with the real world, we don't realize it,
02:33but there's actually a delay between what our brain thinks and tells the body to do and when the body
02:39does it. In a brain computer interface where the interaction takes place directly between
02:44the brain and electrode, we can bring that latency down to individual milliseconds. That's why people
02:50feel like the experience is almost like the interface is predicting their thought. There's
02:54another question that's packed in here. Can brain computer interfaces manifest a fully formed thought?
02:59That gets to the question of how we subjectively feel like we're thinking. Very often, we imagine
03:05something or we have a feeling or a picture or a concept in our brains. And we don't right now have a
03:12way of expressing that fully formed thought other than through drawing a picture or speaking in
03:17paragraph form. But we do have this subjective sense that thoughts exist in a fully formed way.
03:22And there's this question that I think Govalt is asking, which is, will brain computer interfaces
03:26allow us to transmit thoughts in that fully formed way? I have a feeling that the technology will allow
03:32us to travel in that direction probably even faster than we can imagine, even if we can't say exactly how
03:37right now. Pesh909X asks, will BCIs ever be able to record our dreams and psychedelic trips to video?
03:43And the answer is yes. Brain computer interfaces already can see activity in the visual cortex,
03:49which is the part of the brain that processes visual information. And there has been some work
03:54showing that actually some of that information can be decoded to recreate the visual scene. There is
03:59some evidence that that kind of visual replay happens during our dreams. And so it may be possible in
04:04the future, just as some of these studies have begun to show that brain computer interfaces can record,
04:10replay, process the information that occurs during imagined visual activity and dreams.
04:16Steven Roto asks, the first human patient with a brain computer implant used the technology to
04:21successfully play Mario Kart. If that isn't the definition of a 90s kid, I don't know what is.
04:25It is true that the first Neuralink patient used his implant to play Mario Kart and seems to have had a
04:30great time doing it. That wasn't the first human patient with a brain computer implant,
04:34but he did use it to play Mario Kart. And I think that points to the fact that people are going to
04:38use brain computer interfaces to do all the sorts of things that we take for granted and know and
04:43love can be done in the digital world. Fashion Savage asks, will BCI lead to security issues of
04:49hacking or reprogramming people's brains? This is a really important question. It's one that many
04:53people have asked, which is given the sensitivity of neural information, will BCI technology lead to
04:59issues involving hacking or compromising the security of people's private thoughts. Certainly
05:04in the first generation of brain computer interfaces, we're really interacting with the parts of the
05:09conscious brain that move the body and move the hands, move the arms, move the face and the muscles
05:15that control speech. And these are not really areas of the brain that we consider private thought
05:20activity. And furthermore, patients, when you ask them, would you be concerned about a privacy issue,
05:26in this context, many of the people who stand to benefit from this technology would trade a little
05:31bit of privacy for the ability to interact smoothly with the outside world. But that doesn't minimize
05:37the possibility of real security and privacy issues when neural data is being transmitted wirelessly
05:43outside of the body. So we and others involved in the industry have taken real care to try to encrypt
05:50and secure any neural data streams that leave the body. Real Editor Six asks, I was wondering if ever
05:56a human brain merged with an LLM via a brain computer interface, what would the AI experience? What would
06:01that person experience? Is anyone connected already? The answer is a qualified yes, because people with
06:06brain computer interfaces now certainly have the ability using the same means that we do of textually
06:12querying an LLM. And so that interaction is definitely happening. I haven't asked the AI what the AI's experience
06:19is, but certainly we can do that. I look forward to that question. What would that person experience
06:25I think is actually very similar in the current form of the technology to what you and I experience when
06:29we query the AI. Those interactions right now are textual in nature. The AI's output is not fed back
06:36directly into the user's brain just yet. So the BCI users have an experience that's similar to what every
06:42other user of an LLM has. But I get the question is asking towards a future state in which there's a more
06:48symbiotic meld between artificial intelligence and humans through brain computer interfaces. And I
06:53think this is the direction of travel and it's hard to predict exactly what it's going to look like.
06:57James Rosen Birch asks, what part of the brain is the BCI interfacing with? Many brain computer
07:03interfaces to date spend a lot of time in and around the motor cortex, which is this area of the brain.
07:08This is the hand motor area. This is the leg motor area. This is the face motor area involved in speech.
07:13Those motor areas are really important for current generation brain computer interfaces because they
07:18perform the computations in the brain that allow us to interact with the world physically. When we
07:23think about things like typing or moving our hands or walking or speaking, those are the parts of the
07:28brain that serve those functions most directly. But there certainly are future directions of brain
07:33computer interface technology. We can think about connecting to and interfacing with other areas of
07:37the brain and other areas of the nervous system. So areas that control sensation, decision-making,
07:43memory, even parts of the brainstem and spinal cord that are involved in other types of neurological
07:47disease. Our friends at GTech Medical Engineering, how many electrodes are needed to run the brain
07:52computer interface? The answer to that is in the hundreds or thousands. And that just gets us off the
07:58ground. I think it's hard to say what the upper limit is to functionality. So many of us are familiar in the
08:04world of communication that basically the higher the speed of your connection, the more sophisticated
08:09the applications you can run. Or in the world of images, the more pixels or megapixels you have
08:14in your display, the higher fidelity the graphics you can render. And the same is true in brain
08:19computer interfaces. The more detailed and the higher resolution the picture of brain activity you can
08:24generate, the more smooth and sophisticated the real-time interaction you can have with the brain.
08:28Right now we're seeing that brain computer interfaces with around a thousand electrodes get us off the
08:33ground to incredibly high levels of functionality that include things like controlling a cursor,
08:38performing the sorts of tasks that we take for granted in everyday interactions with computers.
08:42But we can see a path towards many thousands of electrodes and even orders of magnitude higher
08:47in which smooth intuitive connections between the brain and the outside world will be even higher
08:52performance. OkHunter8210 asks, considering recent developments in brain computer interfaces,
08:58I'd love to hear from experts or enthusiasts about potential applications in assisting
09:02individuals with severe paralysis or ALS. Have we made sufficient strides toward leveraging BCI
09:07technology for rehabilitation purposes? Yes. Severe paralysis and ALS are really the first
09:12conditions that have received a tremendous amount of attention for brain computer interface technology.
09:17There's quite a few examples of patients in clinical studies who have had tremendous benefit from
09:22their implants quite apart from being part of the clinical studies. And it's exactly these individuals with
09:27severe paralysis from spinal cord injury, certain forms of stroke, and ALS will be among the first to
09:33benefit from the technology. A Reddit user in the Singularity thread asks, is there a concrete pathway to
09:39non-invasive BCI technology? Mostly no in the way that we think of, but also yes to the implied question of is
09:46there a use for non-invasive neural interfaces. So when I think about brain computer interfaces, I think about
09:53systems that are being used to drive real-time interactions between the brain and the outside
09:57world. And that kind of high bandwidth, sophisticated, smooth, high-speed interactions between the brain
10:03and the outside world that really happen at the speed of thought, that requires implanted technology.
10:08There's no way that we know of around the need to actually be touching the brain in some way to get
10:14that kind of high bandwidth, high-speed interaction. Non-invasive techniques and technologies have
10:18captured scientists and neuroscientists imagination for a long time. And there definitely is a use for
10:24non-invasive technologies. They can detect brain state, they can manipulate brain state, and they can
10:29be used to treat certain forms of disease, but not to manipulate at high fidelity and high resolution
10:35in real time. At Howard G9263 asks, brain computer interfaces sound cool, but kind of scary too,
10:41like what if it malfunctions or something? So many important technologies have these sorts of
10:46questions attached to them, whether it's cars or airplanes or gene sequencing or artificial
10:51intelligence. There's always this question of, it's magical when it works well, and what if it
10:56doesn't work well, or what if it malfunctions? It is important, as in the development of all kinds
10:59of technology, to have ways of fixing things when they go wrong. You know, Murphy's Law is real.
11:04Anything that can go wrong will happen in some way, and so we need to try to anticipate failure modes
11:09and plan for how to fix them. We've developed electrodes that interact with the brain in a way that
11:14doesn't damage the brain, and so these electrodes can be moved, removed, replaced, and upgraded
11:19as necessary in the future. And other components of the system can be modularly changed out over
11:24time. For example, the battery or the wireless or certain forms of tunneled connectors that are
11:28implanted under the skin. So understanding the way things obsolesce or change or may need to be
11:33switched out in the future is important to ensuring that we can safely repair a device and plan for all
11:39eventualities. Queen Guinevere asks, imagine what happens when an implant becomes obsolete,
11:43is no longer supported, and becomes increasingly vulnerable to hacking the longer it remains
11:47installed, and you can't uninstall it without surgery. Like, imagine if you had the equivalent
11:51of flash implanted in your brain. This is actually a really deep question, and it gets to the notion
11:56that all technologies have a cycle, and you want to be able to plan for upgrades and plan against
12:01obsolescence. At Precision, one of the ways that we've thought about this question of obsolescence and the
12:06potential need for replacement is to develop an electrode technology that's based on thin films.
12:11This is the precision electrode array. If I turn on its side in this way, you can see it's incredibly
12:15thin. This film coats the brain surface with electrodes. Rather than penetrating the brain,
12:20the idea is that those electrodes could be removed or upgraded over time, and that other components of
12:25the implant also can be removed or swapped out. This notion that something can't be uninstalled without
12:31surgery is an important notion, but I would also just point out that surgery itself is
12:36not necessarily such a huge barrier. Actually, almost everyone over the course of their life in
12:40the United States will have a surgery, if not more than one surgery. So really, it's a question of
12:45making sure that it can be done in a safe manner rather than the concept of a surgery itself.
12:50DCkill97 asks, why can brain-computer interface technology only read from the brain and not write to it?
12:56The short answer is that brain-computer interface technology can both read from and write to the brain.
13:01Reading really means recording neural activity and decoding it, recording the electrical signals
13:06that the brain uses to communicate, and transforming those into command and control signals and ways of
13:11interaction with digital technology, computers, and the outside world. So that's really a translation
13:16problem. The other side of the coin is what people sometimes call writing into the brain, and that
13:21really means stimulating the brain in some way. In the context of current generation brain-computer
13:25interface is that stimulation takes place using electrical pulses, electrical stimulation. Those
13:31kinds of stimulation have been used to restore a sense of sensation, touch, vision, and also to
13:38stimulate the brain in other ways. Most of current generation brain-computer interface technology
13:42really is focused on the read and decode side of things. Reading and writing, or recording and
13:48stimulating, are actually quite different problems. You can't just sort of reverse the decoding into an
13:53encoder that writes information into the brain. It's not that simple. The field of genomics is
13:57one good example of this, in which the early 2000s saw an explosion in our ability to record or read
14:03and decode the human genome, and it wasn't until years later really using almost a completely different
14:09form of technology. Things like CRISPR for gene editing allowed gene modification in a scalable and
14:15programmatic way. UlteriorKid324 asks, how do you get a brain chip into the brain? Is the surgery long?
14:21The surgery is a relatively short surgery. General anesthesia will not always be required. It involves
14:27making a small incision in the scalp and an incision in the bone, but not necessarily removing a
14:32significant amount of bone. It will be an incision right about here. The electrode array itself, you can
14:37see it's very thin, and it gets slipped through an incision in the bone onto the surface of the brain.
14:42Right now that whole process takes about an hour or two. In the future it probably will be a little bit
14:47quicker. We foresee a future in which this can be done as a same day surgery, much as a lot of
14:52surgery across the country is done. Piyush K this side asks, think of a future where thoughts and
14:58feelings can be completely shared through a brain computer interface surpassing the current limits
15:02of language. Will such a future be better or worse? I think we will eventually get to that future,
15:06and I think it will be a better future. I don't exactly know how we'll get there, and I'm sure there'll be
15:10trade-offs. But just like many forms of technology where you can't exactly predict what's going to happen,
15:14I think that's the case with brain computer interfaces. Oli Driver asks, is anybody using AI
15:19to create a better brain computer interface? Definitely yes. Actually all brain computer
15:24interfaces today use forms of artificial intelligence as a core part of how the interface works. It's
15:30important to understand that basically the problem of translating neural data into actions and
15:36interactions with digital technology, that is a translation problem. And that translation between
15:41neural code and digital code requires artificial intelligence and machine learning. But there's
15:47so many other applications of how artificial intelligence is involved in BCI, it's an extremely
15:52important and exciting area. Chrome Plated asks, how can I get involved in brain computer interfaces
15:58and neurotechnology? There are a number of ways to get involved, but especially if you're a talented
16:02engineer, especially if you're a software engineer or have experience in modern machine learning
16:06techniques, I would invite you to apply to work with any of the companies that are working in the space
16:11today. Many of them are hiring. All right, that's it. That's all the questions. Hope you learned something.
16:15Until next time.
Comments