Andrew Huberman interviews Neuralink neurosurgeon Dr Matt MacDougall on brain-machine interfaces and neural augmentation
Andrew Huberman speaks with Dr Matt MacDougall, neurosurgeon at Neuralink, about the science and future of brain-machine interfaces.
Summary
Andrew Huberman interviews Dr Matt MacDougall, a neurosurgeon and key figure at Neuralink, in this Huberman Lab Essentials episode. MacDougall explains that Neuralink's immediate clinical goal is to restore digital agency to quadriplegic patients by reading motor intentions directly from the brain and translating them into computer cursor and keyboard control — not, as is commonly assumed, reconnecting the brain to the patient's own limbs, though that is on the longer-term roadmap. He describes the robotic insertion system that places electrodes finer than a human hair into the motor cortex with a precision no human surgeon could match, and argues that the device has already surpassed the safety thresholds of many standard FDA-approved neurosurgical procedures. MacDougall also discusses a personal RFID chip implanted in his own hand, his willingness to receive a Neuralink implant himself, and his longer-term vision for brain augmentation — including AI integration, high-bandwidth brain-to-brain communication, and neural interventions for addiction, depression, and obesity.
Key Takeaways
FULL TRANSCRIPT
Introduction
Andrew Huberman: Welcome to Huberman Lab Essentials, where we revisit past episodes for the most potent and actionable science-based tools for mental health, physical health, and performance. I'm Andrew Huberman and I'm a professor of neurobiology and ophthalmology at Stanford School of Medicine. And now for my discussion with Dr Matt MacDougall.
Dr. MacDougall, welcome.
Dr Matt MacDougall: Good to be here.
The Brain as an Organ — Failure Modes and Possibilities
Andrew Huberman: Just to kick things off — as a neurosurgeon and as a neuroscientist, could you share with us your vision of the brain as an organ as it relates to what's possible there? Surgeons tend to view the world a little differently than most because they like to fix and they like to mend, and in your case have the potential to add things into the brain that don't exist there already. So how do you think about and conceptualize the brain as an organ? And what do you think is really possible with the brain that most of us don't already think about?
Dr Matt MacDougall: The brain really is a collection of functional modules pinned together, duct-taped together in this bone box attached to your head. And sometimes you see very interesting failure modes. One of the most memorable patients I ever had was very early on in my training. I was down at UC San Diego and saw a very young guy who had just been in a car accident. We had operated on him and, as is so often the case in neurosurgery, we had saved his life potentially at the cost of quality of life. When he woke from surgery with bilateral frontal lobe damage, he had essentially no impulse control left. And so we were called back by his 80-year-old recovery room nurse saying, "You've got to come see your patient right away. Something's wrong." We walked in to see him, and he points at his elderly nurse and says, "She won't have sex with me." It was apparent at that moment his frontal lobes were gone, and that person is never going to have reasonable human behavior again. It's one of the most tragic ways to have a brain malfunction.
But anything a brain does — anything from control of hormone levels in your body, to vision, to sensation, to the most obvious thing which is muscle movement of any kind, from eye movement to moving your bicep — all that comes out of the brain. All of it can go wrong. Any part of it, or all of it. Working with the brain as a surgeon is very high stakes, but once in a while you get a chance to really help. You get a chance to fix something that seems unfixable, and you have Lazarus-like miracles not too uncommonly. It's extremely satisfying as a career.
Advances in Neurosurgical Technique
Andrew Huberman: Could you share with us one of the more satisfying experiences, or what qualifies as satisfying in neurosurgery?
Dr Matt MacDougall: One of the relatively newer techniques that we do is — if someone comes in with a reasonably small tumor somewhere deep in the brain that's hard to get to, the traditional approach to taking that out would involve cutting through a lot of good, normal brain and disrupting a lot of neurons, a lot of white matter, the wires connecting neurons. The modern approach involves a 2mm drill hole in the skull, down which you can pass a little fiber optic cannula, attach it to a laser, and just heat the tumor up deep inside the brain under direct MRI visualization in real time. The person is in the MRI scanner. You're taking pictures every second or so as the tumor heats up. You can monitor the temperature and get it exactly where you want it — hot enough to kill all those tumor cells but not harm hardly any of the surrounding brain.
Not uncommonly nowadays, we have someone come in with a tumor that previously would have been catastrophic to operate on, and we can eliminate that tumor leaving only a poke hole in their skin. Even in cases that previously would have been considered totally inoperable — say a tumor in the brain stem, or a tumor in primary motor cortex, or in primary verbal areas, Broca's area — where we would have expected to either not operate or do catastrophic damage, those people sometimes now come out unscathed.
Neuralink's Mission and Clinical Goals
Andrew Huberman: In your role at Neuralink and as a neurosurgeon in other clinical settings, surely you are using machines. It's an exciting time in which engineers are starting to tackle big problems, and obviously Elon is certainly among the best — if not the best — in terms of going really big. To the extent that you're allowed, could you share with us what the missions at Neuralink are and what the general scope of those missions is? And if possible, share some of the more specific goals? I can imagine basic goals of trying to understand the brain and augment the brain, and clinical goals of trying to repair things in humans who are suffering in some way.
Dr Matt MacDougall: Neuralink — and I think Tesla and SpaceX before it — end up being blank canvases that people project their hopes and fears onto. We experience a lot of upside in this. People assume that we have superpowers in our ability to alter the way brains work. And people have terrifying fears of the horrible things we're going to do. Those extremes are not true.
We are making a neural implant. We have a robotic insertion device that helps place tiny electrodes — smaller than the size of a human hair — throughout a small region of the brain. In the first indication that we're aiming at, we are hoping to implant a series of these electrodes into the brains of people who have had a bad spinal cord injury — people who are essentially quadriplegic. They have perfect brains, but they can't use them to move their body. They can't move their arms or legs. This pristine motor cortex up in their brain is completely capable of operating a human body; it's just not wired properly any longer to the arms or legs. Our goal is to place this implant into the motor cortex and have that person be able to control a computer — a mouse and a keyboard — as if they had their hands on a mouse and a keyboard, even though they aren't moving their hands. Their motor intentions are coming directly out of the brain into the device. They're able to regain their digital freedom and connect with the world through the internet.
Why Robotics for Electrode Insertion
Andrew Huberman: Why use robotics to insert these chips?
Dr Matt MacDougall: These electrodes are so tiny, and the blood vessels on the surface of the brain so numerous and densely packed, that a human physically can't do this. A human hand is not steady enough to grab this couple-micron-width loop at the end of our electrode thread and place it accurately — blindly, by the way — into the cortical surface at the right depth to get through all the cortical layers that we want to reach. I would love if human surgeons were essential to this process, but very soon humans run out of the motor skills sufficient to do this job. We are required in this case to lean on robots to do this incredibly precise, incredibly fast, incredibly numerous placement of electrodes into the right area of the brain. In some ways, Neuralink is pioneering the development of robotic surgeons as much as it's pioneering the exploration, augmentation, and treatment of human brain conditions.
As the device exists currently — as we're submitting it to the FDA — it is only the robot that handles the placement of the electrodes. I, or another neurosurgeon, still needs to do the more crude part of opening the skin and skull and presenting the robot a pristine brain surface to sew electrodes into.
Broadly speaking, the mission of Neuralink is to reduce human suffering, at least in the near term. There's hope that eventually there's a use case that makes sense for a brain interface to bring AI as a tool embedded in the brain that a human can use to augment their capabilities. I think that's pretty far down the road for us, but it's definitely on the desired roadmap. In the near term, we really are focused on people with terrible medical problems who have no options right now.
How the System Works — Bypassing the Spinal Cord Lesion
Andrew Huberman: So just to give us a fuller picture of this — you're talking about a patient who can't move their limbs because they have spinal cord damage.
Dr Matt MacDougall: Right.
Andrew Huberman: The motor cortex that controls movement is in theory fine.
Dr Matt MacDougall: Right.
Andrew Huberman: You make a small hole in the skull, and through that hole a robot is going to place electrodes into the motor cortex. But I still don't understand how the signals are going to get out of motor cortex, past the lesion, and out to the limbs — because the lesion hasn't been dealt with at all in this scenario.
Dr Matt MacDougall: Just to clarify — in the immediate future, we're not talking about reconnecting the brain to the patient's own limbs. That's on the roadmap, but it's way down the roadmap. What we're talking about in the immediate future is having the person be able to control electronic devices around them with their motor intentions alone.
Andrew Huberman: So a prosthetic hand and arm, or just a mouse and keys on a keyboard?
Dr Matt MacDougall: Mouse and keys on a keyboard for starters. You wouldn't see anything in the world move as they have an intention. The patient might imagine, say, flexing their fist or moving their wrist, and what would happen on the screen is the mouse would move down and left and click on an icon and bring up their word processor. Then a keyboard at the bottom of the screen would allow them to select letters in sequence and they could type. We're using smarter algorithms to decode the motor intentions out of the brain. We have been able to do this in monkeys really well. We have a small army of monkeys playing video games for smoothie rewards, and they do really well. We actually have the world record for bit rate of information coming out of a monkey's brain to intelligently control a cursor on a screen. We're doing that better than anyone else.
Would MacDougall Get the Implant Himself?
Andrew Huberman: Would you be willing — or are you willing, if allowed — to have these electrodes implanted into your motor cortex?
Dr Matt MacDougall: Yeah, absolutely. I would be excited to do that. For the first iteration of the device, it wouldn't be very useful because I can still move my limbs, and our first outputs from this are things I can do just as easily with my hands — moving a mouse, typing on a keyboard. We are necessarily making this as a medical device for people with bad medical problems and no good options. It wouldn't really make sense for an able-bodied person to get one in the near term.
As the technology develops and we make devices specifically designed to perform functions that can't be done even by an able-bodied person — say, eventually refining the technique to the point where you can type faster with your mind than you can with speech-to-text or your fingers — that's a use case that makes sense for someone like me to get. It doesn't really make sense for me to get one when it allows me to use a mouse slightly worse than I can with my hand currently.
That said, the safety of the device I would absolutely vouch for. From the hundreds of surgeries that I've personally done, I think it's much safer than many of the industry-standard, FDA-approved surgeries that I routinely do on patients that no one even thinks twice about as standard of care. Neuralink has already reached, in my mind, a safety threshold that is far beyond a commonly accepted safety threshold.
The RFID Chip in MacDougall's Hand
Andrew Huberman: Along the lines of augmenting one's biological function — I think now's the appropriate time to talk about the small lump present in the top of your hand. For those listening and not watching, there's a small lump between Dr. MacDougall's index finger and thumb, placed under the skin on the top of his hand. You've had this for some years now. What is that lump, and why did you put it in there?
Dr Matt MacDougall: It's a small writable RFID tag. It's a very small implantable chip that wireless devices can temporarily power. If you approach an antenna, they can power it and send a small amount of data back and forth. Most phones have the capability of reading and writing to this chip. For years it let me into my house — it unlocked a deadbolt on my front door. For some years it also unlocked the doors at Neuralink and let me through the various locked doors inside the building.
Andrew Huberman: What other sorts of things could one put into these RFIDs in theory? And how long can they stay in before you need to take them out and recharge them or replace them?
Dr Matt MacDougall: These are passive. They're coated in biocompatible glass. I'm a rock climber, and I was worried about that glass shattering during climbing, so I additionally coated them in a ring of silicone before implanting. They're passive — there's no battery, no active electronics in them — so they could last the rest of my life. I don't think I'd ever have to remove it for any reason. At some point the technology is always improving, so I might remove it and upgrade it. Already there are 10x more storage versions available that could be a drop-in replacement. It has a small niche use case, and it's an interesting proof of concept — tiptoeing toward the idea that you have to be willing to go through the things you're suggesting to your patients in order to say with a straight face that you think it's a reasonable thing to do.
Reconnecting the Brain to the Body — The Spinal Cord Bypass
Andrew Huberman: I'd like to go back to brain augmentation. You've made very clear that one of the first goals for Neuralink is to get quadriplegics — well, just to be very clear, the first step is not reconnecting the patient's own muscle system to their motor cortex, but allowing them agency over the movement of things in the world.
Dr Matt MacDougall: Yes. And eventually their body. We would love to do that. We've done a lot of work on developing a system for stimulating the spinal cord itself. That gets to the question you asked a few minutes ago — how do you reconnect a motor cortex to the rest of the body? If you can bypass the damaged area of the spinal cord and have an implant in the spinal cord itself connected to an implant in the brain, and have them talking to each other, you can take the perfectly intact motor signals out of the motor cortex and send them to the spinal cord. Most of the wiring should be intact in the spinal cord below the level of the injury — caused by a car accident, motorcycle accident, gunshot wound, or whatever. It should be possible to reconnect the brain to the body in that way. So it's not out of the realm of possibility that in some small number of years, Neuralink will be able to reconnect somebody's own body to their brain.
Brain-to-Brain Communication
Andrew Huberman: Along the lines of augmentation — early on in some of the public discussions about Neuralink that I overheard between Elon and various podcast hosts, there was this idea thrown out that you would have some chips in your brain and I would have some chips in my brain, and we could just sit here looking at each other and essentially hear each other's thoughts. Is that possible? And how would we restrict what the other person could hear?
Dr Matt MacDougall: Absolutely. Think about the fact that we could do this right now — if you pulled out your phone and started texting me and I looked down and started texting you, we would be communicating without looking at each other or talking. Shifting that function from a phone to an implanted device requires no magic advance, no leap forward. It's technology we already know how to do.
If we put in a device that allows you to control a keyboard and a mouse — which is our stated intention for our first human clinical trial — then texting each other with our brains is maybe an uninspiring rendition of this, but it's not very difficult to imagine the implementation of the same device in a more verbally focused area of the brain that allows you to more naturally speak the thoughts that you're thinking and have me hear them rendered into speech, maybe via a bone-conducting implant, so silently.
What attracts a lot of us to a technology like Neuralink is that the ceiling is incredibly high. There's no obvious reason that you can't interface with millions of neurons as this technology is refined and developed further. That's the kind of wide-band, high-bandwidth brain interface you want to develop if you're talking about a semantic prosthetic, an AI assistant to your cognitive abilities — the more sci-fi things we think about in the coming decades. It's an important caveat when evaluating these technologies: you really want it to be something you can expand off into the sci-fi.
How Brain-Machine Interfaces Learn and Adapt
Andrew Huberman: To what extent can these brain-machine interfaces learn and get better and better, because it's also listening to the nervous system that it's trying to support?
Dr Matt MacDougall: It goes both ways. Basically, what you're doing is taking a very crude software intelligence — not exactly a full-blown AI, but some well-designed software that can adapt to changes in firing of the brain — and coupling it with another form of intelligence, a human intelligence, and allowing the two to learn each other. Undoubtedly, the human who has a Neuralink device will get better at using it over time. Undoubtedly the software that the Neuralink engineers have written will adapt to the firing patterns that the device is able to record and over time focus in on meaningful signals toward movement.
If a neuron fires at a high rate when you intend to move the mouse cursor up and to the right, the software doesn't know that when it starts. When you first put this in, it's just a random series of signals as far as the chip knows. But you start correlating it with what the person wants to do, as expressed in a series of games. You assume that the person wants to move the mouse on the screen to the target that's shown, because you tell them that's the goal. So you start correlating the activity you record when they're moving toward an up-and-right target with that firing pattern, and similarly for up-and-left, down-and-left, down-and-right. You develop a model semi-intelligently in the software for what the person is intending to do, and let the person run wild with it for a while. They start to get better at using the model presented to them by the software, as expressed by the mouse moving or not moving properly on the screen.
Imagine a scenario where you're asking somebody to play piano, but the sound that comes out of each key randomly shifts over time. That's a very difficult problem, but a human brain is good enough — with the aid of software — to solve that problem and map well enough to a semi-stable state that they're going to know how to use that mouse even when they turn the device off for the night, come back to it the next day, and some of the signals have shifted.
Animal Research — Pigs and Monkeys
Andrew Huberman: Let's talk about pigs.
Dr Matt MacDougall: Sure.
Andrew Huberman: Elon and others at Neuralink have shared some of the progress they've made in experimental animals, which understandably evokes strong emotions in people. What was the rationale for using pigs?
Dr Matt MacDougall: Let me say first — Neuralink is almost entirely composed of animal-loving people. If there was any possible way to help people the way we want to help people without using animals in our research, we would do it. It's just not known how to do that right now. We are completely restricted to making advances and getting device approval through the FDA by first showing that it's incredibly safe in animals, as is the case for any medical advancement.
I think we would probably be having a very different conversation around animal research if we as a species, as a culture, weren't just casually slaughtering millions of animals to eat them every single day. That is the background against which the relatively minuscule number of animals used in research becomes almost impossible to understand — why someone would point to that ridiculously small number of animals used in research when the vast majority of animals that humans use and end the lives of are used for food. In that context, we do animal research because we have to. There's no other way around it. If tomorrow laws were changed and the FDA said, "Okay, you can do some of this early experimentation in willing human participants," that would be a very interesting option. I think there would be a lot of people who would step up and say, "Yes, I'm willing to participate in early-stage clinical research."
Andrew Huberman: You already volunteered.
Dr Matt MacDougall: Yeah. And I wouldn't be alone.
On that note, to whatever extent possible, I think Neuralink goes really, really far — much farther than anyone I've ever heard of, any organization I've ever seen — to give the animals agency in every aspect of the research. We have just an incredible team of people looking out for the animals and trying to design the experiments such that they're as purely opt-in as humanly possible. No animal is ever compelled to participate in experiments beyond the surgery itself. If on a given day our star monkey, Pager, doesn't want to play video games for smoothie, no one forces them to — ever. They have free and full access to food the entire time. They aren't hungry, they aren't thirsty. The only thing that would motivate them is if they want a treat extra to their normal rations. There's never any deprivation, never any adverse negative stimuli that pushes them to do anything. It's one of the underappreciated innovations out of Neuralink — how far the animal care team has been able to move in the direction of humane treatment.
Andrew Huberman: Why pigs specifically?
Dr Matt MacDougall: Pigs are actually fairly commonly used in medical device research — more so in the cardiac area, where their hearts are somewhat similar to human hearts. There are a lot of different kinds of pigs that we use in different contexts when we're trying to optimize a certain characteristic. We don't necessarily need them to be smart or task performers, although occasionally we have trained them to walk on a treadmill when we're studying how their limbs move for some of our spinal cord research. But we're not recording interesting cognitive data out of their minds. They're really just a biological platform with a skull that's close enough in size and shape to humans to be a valid platform to study the safety of the device. The pigs are really just a safety platform — to confirm the device is safe to implant, that it doesn't break down or cause any kind of toxic reaction.
The monkeys are where we are really doing our heavy lifting in terms of ensuring that we're getting good signals out of the device, and that what we expect to see in humans is validated on the functional level in monkeys first.
The Long-Term Vision for Brain Augmentation
Andrew Huberman: What is your wish for brain-machine interface and brain augmentation? Let's assume the clinical stuff can be worked out — or maybe you have a pet clinical condition you're just yearning to see resolved, and that would be fine too. But in addition to that, let's expand out. Let's say we can extend your life 200 years, or we're thinking about the kind of world your children and their grandchildren will live in. What do you think is really possible with brain augmentation and brain-machine interface? Please feel no bias whatsoever to answer in a way that reveals your empathy and consideration of clinical conditions — that's how you spend your days, fixing patients and helping their lives be better. But for the sake of fun and delight, and for the sake of really getting the audience to understand what's possible here, please feel no shackles.
Dr Matt MacDougall: I love the idea — down the road, and we're talking a 10-, maybe 20-year time frame — of humans just getting control over some of the horrible ways that their brains go wrong. I think everybody at this point has either known someone, or second-order known someone, who has been touched by addiction, depression, suicide, obesity. These functions of the brain — or malfunctions of the brain — are what drives me. These are the things I want to tackle in my career.
In terms of my kids' lifetime, I'm thinking full human expansion of cognition into AI, full immersion in the internet of your cognitive abilities — having no limitation for what you think as bottlenecked by needing to read the Wikipedia article first to have the data to inform your thoughts. Having communication with anyone you want, unrestricted by this flapping air past meat on your face. It's a means of communication that's ridiculously prone to being misunderstood, and also a tiny, narrow bottleneck. We're trying to send messages back and forth through a tiny straw. There's no reason that needs to necessarily be true. It's the way things have always been, but it isn't the way things are going to be in the future.
I think there are a million very sci-fi possibilities in terms of banding human minds together to be even more potent as a multi-unit organism — as an opt-in multi-brain. These are things so far down the road I can't even directly see how they would be implemented. But the technology we're working on is a little crack in the door that allows some of this stuff to even be thought about in a realistic way.
To that point, I encourage anyone who is excited about things like that — especially mechanical engineers, software engineers, robotics engineers — to come to the Neuralink website and look at the jobs we've got. We need the brightest people on the planet working on the hardest problems in the world, in my opinion. If you want to work on this stuff, come help us.
Andrew Huberman: I really want to thank you, first of all, for coming here today and taking time out of your important schedule of seeing patients and doing brain surgery — literally time away from your family and time away from your mission at Neuralink — to share with people what you guys are doing. And I really just want to thank you also for being you, which perhaps sounds like an odd thing to hear, but as made apparent by the device implanted in your hand, you don't just do this for a job. You live and breathe and truly embody this stuff around the nervous system — trying to figure out how to fix it, how to make it better. I know your deep love for it. So I want to thank you not just for the brains and the energy you put into it, but also for the heart that you put into it.
Dr Matt MacDougall: Thanks for that, Andrew. I appreciate it. We just want to help people. We want to make things better.