September 16, 2026 • 39 mins
Controlling a computer with your thoughts may sound like science fiction, but it’s closer to reality than you might think. Just ask Precision Neuroscience. The company developed a non-damaging brain-computer interface that’s undergoing testing with nearly 100 participants across 18 major academic medical centers.
Oz sits down with Michael Mager, the CEO and Co-Founder of Precision, to discuss how the technology works, why his company is taking a different approach from Neuralink, and why the brain-computer interface could be groundbreaking for neurological conditions far beyond paralysis.
See omnystudio.com/listener for privacy information.
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:14):
Welcome to Tech Stuff. I'm Oswald Loshin. Back in 2024,
a 71-year-old Parkinson's patient named Jeffrey Kiefer underwent brain surgery.
While his skull was open, surgeons temporarily placed an experimental
brain-computer interface directly onto his brain to record neural activity,
a key step towards technology that could one day allow
(00:35):
paralyzed individuals to control computers using only their thoughts. The
following year, Precision Neuroscience, the startup behind that device, reached
a big milestone, FDA clearance for 30-day human trials. As
of this year, testing has expanded to nearly 100 participants
across 18 major U.S. academic medical centers. Joining me today
(00:57):
to discuss how this technology works and where it's heading
is Michael Maker, CEO of Precision Neuroscience. Michael, welcome to
Tech Stuff.
Speaker 2 (01:06):
Thanks so much for having me.
Speaker 1 (01:06):
What are you building?
Speaker 2 (01:08):
We're building, as you mentioned, a brain-computer interface. That is
a system that I think is best thought of as
a communication tool. It creates a communication link between the
brain and computers. The first use of that link is
to enable people to control computers with their thoughts, people
who have lost ability to move their arms and hands
(01:29):
as a result of, you know, paralysis due to spinal
cord injury and certain diseases. But we think the applications
of this communication link effectively between human and artificial intelligence
go way beyond paralysis.
Speaker 1 (01:41):
So they said everyone might have one?
Speaker 2 (01:43):
I think the future is hard to predict. We see
a path to having major impact clinically and potentially eventually
beyond that. But we're really focused today on making a
meaningful positive difference medically and clinically.
Speaker 1 (01:57):
Talk about that journey from working with Jeffrey Kiefer to
the FDA clearance to the 100 participants? What does that
mean that you get to work with 100 people now?
Speaker 2 (02:06):
Taking a step back, I think that it's important to
understand medical technology requires people like Jeffrey Kiefer, who are
really selfless in offering themselves to the advancement of medical technology.
The reason that he was able to get an implant
of this kind is that our system is is designed
(02:27):
to be totally safe, it is reversible, and it is non-damaging.
And because of that, we were able to get into
the clinic and start implanting patients within two years, which
is a very, very expedited path. And our first FDA clearance,
as you mentioned, within about 40 years of founding. That
is really a function of the fact that our system
(02:48):
is able to be explanted safely without having done damage
to the brain.
Speaker 1 (02:53):
Explanting means you can take it out.
Speaker 2 (02:54):
That means you can take it out. Yeah. And I
think that that's an incredibly important feature. Really, when it
comes to patient adoption, you know, people want a system
that they know if they, for whatever reason, don't want
can be removed safely. And also for upgrades, you know,
this is a technology that is in the early stages
of its development. I think people rightfully want a system
(03:15):
that's going to be able to be upgraded over time
as functionality improves.
Speaker 1 (03:20):
And who founded the company? What was their vision?
Speaker 2 (03:22):
So I co-founded the company with Ben Rappaport. Ben is
really ideally suited to pioneer this industry. He is a neurosurgeon.
He also has a PhD from MIT in electrical engineering.
That is not by accident, you know, getting both a
medical degree and a PhD in electrical engineering. The brain
is an electrical system. So every thought you've ever had,
(03:45):
every idea, every memory, every intention, there is a physical
representation of of those thoughts.
Speaker 1 (03:52):
By the way, it's a remarkable coincidence that both computers
and the brain work on electricity. When did people start
to understand what that could mean?
Speaker 2 (04:00):
A hundred years ago, I think, is when we started
to understand the electrical nature of the brain. I think
that for a long time, these two fields have been
quite disparate. I think we're actually entering a period, and
I'm sure we're going to talk about this, where there
is really a convergence of neuroscience and artificial intelligence that
has the potential to be impactful really on both sides
(04:21):
of the coin.
Speaker 1 (04:22):
But your co-founder, in a sense, this is a substantiation
or a harvest of decades of academic research and vision.
Speaker 2 (04:30):
That's exactly right. I mean, he really was there at
the birth of this industry, which initially happened like so
much in terms of sort of breakthrough advances in academia initially.
where the first proofs of concept that you could record
and decode electrical activity in order to drive a function
on a computer actually happened in 2004, so over 20
(04:53):
years ago. And the industry has been developing really in
academic settings with prototype devices. Ben wanted to be involved
in that from the very beginning. He studied electrical engineering
and became a neurosurgeon because that's the perfect combination of
skills to pioneer this industry. He was also one of
the co-founders of Neuralink with Elon Musk. And after a
(05:15):
few years at Neuralink, he left because of the direction
that they went in. And he and I co-founded Precision
in 2021. So what was the direction they went in
that he objected to? So the initial work in the space, again,
going back to sort of the academic proofs of concept,
involved electrodes to record and stimulate the brain that penetrate
(05:37):
into the brain.
Speaker 1 (05:38):
Right.
Speaker 2 (05:39):
And when you penetrate into the brain... a few things happen.
First of all, you kill neurons. So there is damage
that is irreversible. There's also a response from the brain.
The body doesn't like it when you put a foreign
object into it. And so there's something that's called gliosis,
which is effectively sort of scarring around the foreign body.
Over time, if you think about creating an interface between
(06:03):
the brain and technology and compute, you really want a
system that's going to be able to scale. that's going
to be able to start in a small discrete area
of the brain, but over time expand because different areas
of the brain are responsible for different functions. And so
to unlock greater and greater performance, you need a system
that's going to be able to expand to multiple regions
(06:24):
of the brain at once. If you have to poke
holes in the brain to do that, that just, there's a,
there's a ceiling to the scalability of such a system.
There's also a patient adoption and surgeon adoption issue. You know,
When you're talking about any surgical intervention, there is a
necessary weighing of risk and benefit. And you want that
(06:47):
trade-off to be as favorable as possible.
Speaker 1 (06:50):
When you're talking about this idea of reaching all the
regions of the brain, what came into my brain was
the image of a metal helmet underneath my skull. Is
that a crazy image?
Speaker 2 (07:00):
Metal helmet is not how we would describe it. So
what we've developed is an ultra-thin film. Think about it
like a piece of saran wrap. in your language, maybe sellotape?
Speaker 1 (07:11):
Yeah, I know, cling film.
Speaker 2 (07:12):
Cling film. Cling film.
Speaker 1 (07:13):
Sellotape is something, yeah.
Speaker 2 (07:16):
That basically covers the brain in a super, super gentle,
thin film with thousands and eventually hundreds of thousands of
electrodes on it. It does so in a way that
is fundamentally sort of non-damaging.
Speaker 1 (07:32):
What are the sort of planes of competition, let's say?
Because obviously Neuralink you know, because of Elon, probably because
of first mover is the most famous player in the space.
There's you, there's maybe two other kind of household names
of people who know the industry, right? Like, are you
all chasing exactly the same holy grail or are there
different planes of competition, so to speak?
Speaker 2 (07:53):
I think there are absolutely different planes of competition that
will emerge over time. So for the time being, you're right.
Everybody is focused on this initial question really around the
hardware of how do you get electrodes into or onto
the brain in a way that's safe and scalable. And
our system, unlike Neuralink's, which is penetrating into the brain
because they've built on that academic work that sort of
(08:15):
founded the industry, ours creates a non-damaging thin film. Others
go through the vasculature. Some folks are trying to use ultrasound.
I think that there will emerge clear areas where certain
approaches are better than others, and companies will necessarily focus
where their systems are sort of optimized. That said, We've
(08:38):
picked the approach that we've picked because we think it
is really the only one that achieves ultra high performance
for all of the functions of BCI that exist today,
but without some of the drawbacks around penetration and damage
that some of the competing products entail.
Speaker 1 (08:55):
So you mentioned saran wrap. What does this thing physically
look like? How big is it? And how do you
get it into somebody's head?
Speaker 2 (09:02):
We manufacture the system that is about 20 microns thick.
So that is about a fifth the width of a
human hair. So super thin. And it's about one and
a half square centimeters in terms of surface area. So
think about it like a postage size stamp. We deposit
1,024 tiny platinum electrodes on that. These electrodes are 50
(09:24):
microns in diameter. That's the size of a neuron. So
we're really speaking to the brain at its own natural scale.
The only way to manufacture at that scale is to
do so using manufacturing techniques borrowed from semiconductor manufacturing. So
we use photolithography to make these devices. We're the first
company with an FDA-cleared device that is photolithographic.
Speaker 1 (09:47):
Is ASML the same providers as the chip industry? We're not.
Speaker 2 (09:52):
So those guys are at the nanoscale. We're at the
micron scale. But the concept is the same, where basically
to create... a breakthrough, in this case in medical technology,
you really have to think about it in a different
way and take techniques that have been pioneered elsewhere. In
our case, that's required actually buying and retrofitting a fab,
(10:13):
which we own in Texas. We have 20 people on
the ground who do.
Speaker 1 (10:17):
Which used to be used to build chips is now
being used to build your.
Speaker 2 (10:19):
That's right. And we have a level five clean room.
All of this is necessary. It's capital intensive. It's resource intensive.
But there is no supply chain that exists to do
this kind of work because it's so pioneering.
Speaker 1 (10:32):
How do you put it inside somebody's head? And is
the vision to make them much bigger ultimately? Yeah. And
how do you know where to put it?
Speaker 2 (10:39):
First of all, I don't put them in anybody's brain.
I will assure you. But we're now working at 18
of the best hospitals in the United States, which by
the way, is multiples of the number of any other company.
Part of that is enabled by our FDA clearance, which
also is a first. And we're working with neurosurgeons at
these institutions who are absolutely best in class. They are
(11:01):
using the system in different ways. And part of the
benefit of a system that does not penetrate into the
brain and does not damage the brain is that you
can explore. You can put it. in the area that
controls the motor part of the brain. So the part
of the brain that controls the arms and the fingers
and the hands and use that to control a computer
(11:22):
cursor just with thoughts, which we've done. But you can
also put it in areas of the brain that are
much less well characterized. We have gone to somatosensory cortex
and prefrontal cortex. We've gone to the brainstem. We've actually
even gone to the spinal cord. That is the only
way that you can explore in that manner and you
can enable academics and clinicians to explore in that manner
(11:45):
is if your system is fundamentally non-damaging and reversible. So
we think this is a huge advantage for precision, but
also for the field, because I think it's going to
unlock understanding that previously has just not been possible.
Speaker 1 (11:58):
What are you finding today by going to those places?
Speaker 2 (12:01):
I think we're at the beginning of this journey. I
think a lot of our efforts up until now have
been focused on that motor part of the brain. So
you asked the earlier question, is everyone focused on the
same thing? And today, the answer is largely yes, where
we're really focused all on paralysis, enabling people whose brain-body
connection has been disrupted by a disease or an injury.
(12:22):
So think about someone who's in their 20s, they're in
a catastrophic car accident, their brains function as well as
yours or mine probably better, but they have lost the
ability to control their arms and hands. And so they
can't use technology in the way that we all take
for granted. So you can't use a phone, you can't
use a tablet, you can't use a computer. Our goal
is to give people in that situation control again over
(12:46):
a digital ecosystem to enable them to have higher quality
of life and eventually to go back to work if
they want. The FDA is effectively pushing us and others
to addressing that initial use case first, because we know
it works. The risk benefit is just very obviously favorable,
(13:07):
and it's a way to get this technology out into
the world in a way that's going to make a
real positive difference.
Speaker 1 (13:12):
Pushing by providing grant funding or pushing by not delaying
things too much.
Speaker 2 (13:16):
I'd say encouraging. I'd say the FDA has been extremely
constructive But it's clear that we all should start in
this place. That said, as this technology gets out in
the world and as it advances, as discussed earlier, I
think different companies are going to end up focusing in
different areas. We are focused initially on the motor part
(13:37):
of the brain because that is our first clinical application.
And so decoding intended movements to drive a digital keyboard
and digital mouse effectively is where we've started. But Now
that we have established that functionality, we have the opportunity
to look beyond it, and that's something that we're in
the early days of doing.
Speaker 1 (13:58):
There was a New Yorker story, I guess probably in 2018,
about the scientist at Carnegie Mellon who was leading. You
know his name, I guess, but I can't remember it
off the top of my head.
Speaker 2 (14:07):
There are a few fantastic ones there.
Speaker 1 (14:09):
I met him, and it was a New Yorker story
about how two young people had become paralyzed, both had
the ability with this electronic array placed into their brain
to control artificial limbs and maybe it was University of
Pittsburgh actually but the piece had this kind of lyrical
but ultimately quite tragic quality which was that it stopped
working because whatever the word you use for the scar
(14:31):
tissue in the brain would essentially grow over the arrays
and like this miraculous experience like had its limit and
that was a second sadness essentially in that but what
you're saying is we may be on the cusp of
that not being true.
Speaker 2 (14:46):
I think What we and a few others are doing
is taking that amazing proof of concept and productizing it
because that was achieved. And the other experiments like it,
places like Carnegie Mellon and University of Pittsburgh and a
few other centers of excellence, using prototype devices that involved
(15:06):
parts of the system actually protruding from the head, wires
often physically connecting the brain to external compute. There's all
sorts of issues and risks associated with that, with infection
being sort of the most obvious one. And the systems
were nowhere near robust enough to go through the FDA
regulatory process. So the positive sort of take on this
(15:32):
is this technology works and is going to make a
huge difference to a group of people who right now
have nothing. The negative is it's been validated in academia
for two decades, and about 100 people have been implanted.
I mean, incredibly poor real-world impact. And so I think
it's our job to make sure that the next 20
years are not like the previous 20.
Speaker 1 (15:53):
How many of these implants do you think you may
do in the next 20 years?
Speaker 2 (15:57):
The path for precision is initially to start with severe paralysis.
There are 350 or so thousand people in the United
States and several million globally who are completely paralyzed. So
unable to use arms and hands at all, but still
sound of mind. I think the next expansion from that
(16:20):
is towards partial paralysis. So think about, for example, someone
who's had a stroke. So they can use one side
of their body, but not the other. Again, that means
severe limitations in terms of an ability to control a keyboard,
a mouse, a touchscreen. And that is a population that
is five to 10 times larger than the initial severe
(16:40):
paralysis population. So just in those two groups, which are
our initial two areas of focus, you're talking about millions
of people in the United States and tens of millions
of people globally. I'm very excited over time to address
some of the neurological illnesses that we either have no
treatments for now or that they're very, very poor. So
(17:03):
two examples I'll give are one is refractory depression. So
depression that doesn't respond well medication. The idea of using
electricity to disrupt neural circuits that are not operating properly
is extremely well established. It's electroconvulsive therapy. You're talking about
putting electrodes on people's skulls. It's very coarse. There are
(17:25):
significant off-target effects, which is sort of a euphemism for reality.
And yet, it's still used as standard of care because
it works.
Speaker 1 (17:33):
So what you've been talking up until now is output,
decoding the thoughts from the brain to create action in
the world. What you're now talking about is input.
Speaker 2 (17:40):
That's exactly right. I think the way that a really
high-performance depression-focused system must work is bidirectional. So you need
to be able to, instead of just introducing current to
the brain, irrespective of brain state, you really want to
understand when the brain needs intervention. And then when it does,
(18:02):
do so as precisely as possible to limit the sort
of the side effects of that intervention. I think that
is only possible with a high-performance, high-bandwidth brain-computer interface.
Speaker 1 (18:14):
And when might UI want to get one of these
voluntarily and why?
Speaker 2 (18:17):
I think the short answer is there's already people who
want one. That said, I think that that's, you know,
Neuralink has publicly said they have a 10,000-person wait list
for their system. I guarantee not all those people are
severely paralyzed. we all get inquiries from folks who are
interested in controlling computers in a more seamless way. I
(18:38):
think the kind of maybe superficial way of describing it
would be to say, well, we have always evolved the
interface between computers and human beings, and that's unlocked new function.
So like mainframe computers, we had punch cards. With the
desktop and the laptop, we have a keyboard and a mouse. Increasingly,
(19:00):
we control computers with smartphones by tapping on a little
black screen. And all of those shifts have created whole
new areas of function, new applications. I think the end
state is direct neural control. That said, would I or
would I recommend my family get an implant simply to
(19:20):
be able to control a computer cursor a little bit faster,
a little bit more efficiently? No, I think most people
would not. I think where the really interesting and exciting
developments lie is around being able to interact with technology
in ways that are just not possible right now. So
I'll give you an example. We all use AI in
(19:43):
one way, shape, or form, largely through text prompts. You
either speak a prompt or you... type it. That involves
taking an idea and basically condensing it into words and
outputting those words to a prompt. And what I think
everybody experiences is that the output is magical in some ways,
but often off.
Speaker 1 (20:04):
Lost in translation.
Speaker 2 (20:05):
A little bit. And you got to keep going back
and forth and Eventually, it usually gets it. But the
reason for that is at least partly because the AI
doesn't have all of the context that you're bringing to
this request. It doesn't have your memories. It doesn't have
your intentions. It doesn't have your perspective. It can't understand
(20:28):
all of that. And it's very lossy to take an
idea and condense it into a few sentences and a
text prompt. I think where... we are headed is a
system that can decode and interpret higher level mental processes.
You know, we're starting with more mechanistic sort of motor control,
but if you can decode the motor part of the brain,
(20:49):
you can also decode some of these more sophisticated mental
processes over time. And I think that that will enable
us to interact with AI in ways that just aren't
possible without a system like this.
Speaker 1 (21:02):
But so, I mean, talk about that because the, I mean,
It's obviously one of the hardest science challenges humanity has
ever solved to decode the brainwaves into electricity and turn
it into output information to move a mechanical limb. I mean,
no one would take away from how extraordinary that is.
But ultimately, like, I know I want to raise my
hand is different from, I know I want to sit
(21:23):
here with Michael and continue the conversation, but I'm also hungry. Like,
you know, like, should I go for lunch or should
I stay? I'm going to stay because I have a
higher order brain function that makes the social norm is
that we continue this interview even though I'm hungry. First,
what's been done now and what hasn't been done? But secondly,
how do you begin to bridge that gap?
Speaker 2 (21:40):
It's a fantastic question. And I only have partial answers,
but I'll sort of offer you what I have. That
is absolutely true. It's part of the complexity of the
brain and interfacing with the brain. And it's actually present
even in the work that we're doing now. So we
brought on, as an advisor, we announced this a couple
months ago, John Donahue. He was the pioneer. John is
(22:03):
the godfather of BCI. He did that sort of initial
2004 implant and most of the other ones. He's seen
it and done it all. And he joined Precision for
a variety of reasons, which I can get into. But
talking to him about some of the experiences, you know,
what they found, especially in their earlier systems, was if someone...
(22:23):
was watching television. They had, you know, one of the
participants that he worked with had a favorite soap opera.
And every time the soap opera would come on, it
would be on, you know, sort of in the corner
of this person's eye, but he'd end up watching it
unbeknownst to the researchers who were working with him. And
the decoding of his motor performance would degrade. So, you know,
(22:46):
that is absolutely a real phenomenon that we all have
to deal with. That said, The systems that we have
now on the algorithmic side are just orders of magnitude
more powerful than even the ones we had two years ago,
let alone the ones we had 20 years ago. That's right.
And that's what's so interesting that I think artificial intelligence
(23:07):
is enabling us to understand aspects of human intelligence that
just weren't possible even a few years ago. But this
is a question of more and more data applied to
cutting edge algorithms to help us untangle these different mental
processes and have systems that work to our benefit.
Speaker 1 (23:49):
I read somewhere, and this may be wrong, but that
Neuralink had a kind of proprietary data set of chimpanzee
brain research that you guys don't have, and therefore that
gives them an edge. Like, how does the AI with
reinforcement learning, et cetera, how do you teach it what
the signals mean?
Speaker 2 (24:03):
Yeah. Neuralink made a choice to work with non-human primates
and monkeys. We have not. Our animal model is pig.
And part of the reason that we didn't work with
monkeys is because we were able to get into humans
within two years. It took them, you know, almost triple
the time. And that's, again, just the function of the
fact that our system is non-damaging. And so the barrier
(24:25):
to doing a first implant was much lower. We have
really built on that fundamental advantage. And I would say
that our clinical ground game is by far the best
in the industry. So our velocity of implants is rapidly increasing. That, again,
on top of the fact that we have a vertically
(24:45):
integrated supply chain, This whole clinical operation is resource intensive.
It's capital intensive. It's really hard. You know, companies that
come from more of sort of an engineering or tech background,
I think really struggle to understand how to work in
hospitals in ways that aren't disruptive. You know, you have
to get invited into these systems in order to work.
(25:08):
And I think having been as, you know, the chief
science officer and co-founder of the company lends a huge
amount of credibility within this community to the work that
Precision is doing. So I think the answer to your
question is, yeah, non-human primate data is great, but what
you really want is human neural data. And I think
(25:29):
we have access to that at scale and with diversity
across patients that has just never before been done in
human history. And again, this is all focused today on
motor decoding, and it's all based on patients have to opt-in.
So it's a consent-based system. It operates all under the
(25:49):
review boards of the institutions where we're working. And it's
all de-identified data that complies with HIPAA. So I want
to say that because I know that this stuff can
get understandably concerning when people hear about neural data. But yeah,
what we're talking about is really sort of correlating what
these electrical signals in the brain actually mean and the
(26:10):
extent to which they are representative across different people, which
it turns out for motor systems, they really are.
Speaker 1 (26:18):
I should have asked you this earlier, but the array,
how does it communicate with the outside world from inside
the human body?
Speaker 2 (26:25):
The people that we're implanting today are undergoing a neurosurgical
procedure anyway. So they're in the hospital. For example, they're
having put out a demo a couple months ago of
someone who was undergoing a surgery because he had a
benign brain tumor. And so while he was in the hospital,
The surgeon said, listen, we're working with this company, Precision Neuroscience.
They have an FDA clearance. Would you like to, while
(26:47):
you're in the hospital anyway, be implanted and learn how
to control a computer with your thoughts? And about 50%
of people say, that sounds amazing. I'd love to. And
about 50% say, no, thank you. The system that we're
implanting in those cases is wired. So the array is
bonded to a little black package that sits outside the
head that has a wire that goes to a computer.
(27:11):
It is the same signal chain and the same basic
architecture as the fully implanted system. But the fully implanted
system needs to be hermetic and it needs to stay
in the body forever. So you have a different level
of sort of testing rigor that you have to go
through in order to get a fully implanted system approved
for human use. We're in the process of doing that.
(27:32):
You know, we're going to have more news on that soon.
But what we're able to do uniquely is basically demonstrate
that our system is able to perform at the highest
level today and not have to wait for, both on
the performance validation or on the data collection side, the
time that the permanent implant is ready.
Speaker 1 (27:51):
I was at a talk that was an interview with
Demis Hassabis and Reid Al-Brigotti from Semaphore a couple of
months ago. And Demis was talking about decoding an image
a person is thinking about or dreaming of. and then
using a model to recreate the visuals, and then asking
somebody in the scanner what they're thinking about, and it
is indeed the thing that the machine saw. Kind of mind-blowing. Yeah.
Speaker 2 (28:13):
I mean, it turns out that some of these things
are a lot easier than others. It turns out that
sort of, you know, mind's eye representations are quite prominent
signals in the brain, whereas other mental processes are more subtle.
But yeah, it is. It's amazing. And I think it's
exciting for a variety of reasons, but also something that
(28:33):
we need to be very careful about.
Speaker 1 (28:35):
Are you building cyborgs?
Speaker 2 (28:38):
I would say cyborgs have existed for 70 years. Anyone
who has a pacemaker is in some ways a cyborg
or a spinal cord stimulator or a deep brain stimulator.
I think the idea of technology addressing... human health challenges
is not a new one. This is, I think, an
(28:58):
evolution in that sense, not a revolution.
Speaker 1 (29:01):
But you alluded to people's concerns. What are your fears?
How could this, if not properly controlled, what are some
of the negative scenarios?
Speaker 2 (29:12):
I mean, I think what we're doing is, and not
just us, the other implantable BCI companies, I think is
an unambiguous positive. I think offering a tool to people
who right now have no therapy available to them is terrific.
But like any tool, it can be used, a hammer
(29:34):
can build a house or it can you know, harm
someone else. And so I think that there are for
sure legitimate concerns around the evolution of this technology, both
in the initial sort of read-only setting, but I think
as you alluded to, when it comes to introducing current
to the brain, so write versus read, I think that
(29:55):
opens up a whole nother set of concerns. I think
it's important to remember that this is a technology, though,
that people have to opt into. No one's going to
be forced to have a brain-computer interface. And we as
a company won't succeed, nor would we ever want to
if we were doing things that were not ultimately in
the people who adopt the technology's best interests. That said,
(30:18):
I don't think this is going to be a case
where companies self-regulate. There are already a number of different
groups that are coming together to help ensure that this
technology is developed in the right way. One of those
is the FDA helps convene something called a collaborative community,
which is basically all the different stakeholders that are important
(30:39):
to a new area of technology get together. And ethics
and data security is one of the most prominent work
streams of that collaborative community. And I think there's some
incredibly thoughtful work that's coming out of that. Similarly, I'm
part of a group at the World Economic Forum that
is also thinking through and going to soon propose guidelines
around neurotechnology, data privacy, and data security. So there's a
(31:03):
sense that we got social media and a lot of
other areas of technology wrong. The horse is bolted, and
we need to make sure that we do this one right,
which I think is the right instinct.
Speaker 1 (31:14):
You talked about read, and the current paradigm for read
is like motor. If you're paralyzed, you have this amazing
opportunity to take action in the world with your thoughts.
Maybe the next generation of read is what Demis is
talking about, where your visual imagination can be directly interpreted.
So for example, making a movie rather than decoding into
(31:34):
words and then going back from words to images, which
happens with a director, you could just download directly from
your imagination potentially.
Speaker 2 (31:41):
Totally. I mean, I think some of the people who
have described what they believe the future of BCI entails
is along the lines of what you described, which is,
you know, right now, I described how we communicate with AI,
but it's actually how we communicate with each other also, right?
We have ideas. We have a picture in our mind's eye.
We have a complicated concept. And in order for me
(32:01):
to communicate that to you, we have to condense that
thought or idea or image into words. And then we
either like flap our mouths or we tap our fingers
in order to communicate to somebody else. Imagine being able
to project an image or an idea. I think that,
you know, that has the potential to unlock. I think
(32:24):
it could be incredibly positive for humanity. It's kind of
scary because no one has ever done that before and
it's new. But I think that that potential exists.
Speaker 1 (32:32):
So what's the far horizon of read as it pertains
to decoding the brain? What's sci-fi today that might be
science fact five years or ten years from now?
Speaker 2 (32:39):
I mean, I think about that, honestly, more clinically than
I do in any other way. you know, the brain today,
neurological illness is the largest source of sort of economic
loss in the world. You know, the WHO estimates one
in three people at one point or another have a
serious neurological illness in their lives. And, you know, the
(33:03):
brain is still like basically offline and analog. And when
it goes wrong, we have either no interventions or very,
very sort of coarse ones. If you think about medicine
more generally, and especially it feels like it's accelerating in
an exciting way, we're making such rapid advances. And they're
really driven by digitizing biology and then applying compute in
(33:26):
a way that unlocks breakthroughs. So mRNA, the incredible performance
of the potential cancer therapeutics and the whole genomics revolution
is a great example of this. There's so many areas
of medicine that I think are being transformed effectively by compute,
not the brain. the brain remains impervious to commute. It's
hard to get to. So it's like physically very well
(33:47):
protected by the skull. And then once you get to it,
it's delicate. And so we have not up until now
had a tool to digitize the brain so that we
can apply compute to it and understand what's going on.
And I believe intervene in much more precise ways when
(34:08):
it's not behaving properly. So my vision for the next
five years, what I hope that we can head towards
is a world where neurological illness generally is much less debilitating.
Speaker 1 (34:22):
And is organoids like a parallel research path? I mean,
is that like there's the organoid crowd and there's the
BCI crowd or like what's the.
Speaker 2 (34:29):
You know, I'm not an expert on organoids. I think
they certainly have a place in the research. I think
that they are proxies in some ways, but maybe not
as good proxies as we had hoped for for the
full complexity and dynamism of all sorts of different organs,
including the brain. I hope that that also is an
area that produces breakthroughs, but I think it's not an
(34:53):
area that I'm expert in.
Speaker 1 (34:55):
The Wall Street Journal wrote, and to be fair, this
was back in 2024, other than perhaps the quest for
human-level AI, well, a show that is a bit dated,
or colonization of other planets, Not coincidentally, two other areas
where Musk is a big proponent. Few fields exhibit such
a wide gulf between technology's potential and its near-term prospects. Obviously,
(35:18):
written in 2024 and showed his age with the first
part of the quote, but what about the second part?
Speaker 2 (35:23):
I think going back to what we described earlier, two
decades of academic proofs of concept, and yet fewer than
100 people walking around with a BCI. Yeah, I think
that that's incredibly disappointing. I think that it's our job
to make sure that we don't look back five years
hence and think that that comment remains accurate.
Speaker 1 (35:42):
And yet other people have said 2026 will maybe be
the chat GPT moment for BCIs. Why do they say that?
I don't know.
Speaker 2 (35:49):
I don't say that. But I think that a chat
GPT moment is likely to come. I think if you
take a step back, so I think one of the
most important and valuable assets that we have as a
company is this pipeline to collect data at scale. This
incredibly valuable, first of its kind on earth, high resolution
(36:12):
neural data. And we've used that to implant nearly a
hundred patients. And, you know, that's multiples of what anyone
else says, but a hundred patients is a hundred patients.
It's not, I mean, We need to get to thousands
and tens of thousands and hundreds of thousands over time.
I think that is exactly what we're set up to do.
(36:33):
But I think in the context of big data, you're
looking at further scaling up. And we don't know exactly
where the numbers start to tip and the breakthroughs really
start to come. I think we're excited to find out, though.
Speaker 1 (36:46):
What does a world in which you deliver on your
mission look like?
Speaker 2 (36:51):
In the first instance, and I think about this in stages,
in the first instance, some of the illnesses that we're
going after. So if you think about paralysis and other
BCI companies are going after blindness, these are like biblical afflictions.
These are things that have been debilitating and irreversible since
human beings existed. And I think in the next five years,
(37:15):
we have the potential to really change what it means
to be disabled. And I think that's incredibly exciting. I
think that should be doable based on the technology that
we have today and the trajectory that we're on. I
think 10 to 15 years is more along the lines
of what we were discussing earlier around taking neurological illnesses
(37:39):
that maybe have poor therapies today or no therapies that
are debilitating in different ways, but which we have very
little understanding of and very little ability to intervene in.
And really changing those and making those sorts of neurological
illnesses addressable. I think we have the potential to reduce
(37:59):
suffering enormously. Over time, do I think that this will
be a system or an approach that gets adopted by,
you know, for general use? Probably. I think so. I
think we're going to need to see continued performance breakthroughs
to justify, you But I think that those are quite likely.
(38:20):
I do think in that sort of future state, having
a system that is fundamentally safe and non-damaging and reversible
are going to be table stakes for a system to
achieve adoption.
Speaker 1 (38:33):
Michael, thank you.
Speaker 2 (38:34):
Thanks so much for having me.
Speaker 1 (38:52):
For Tech Stuff, I'm Oz Veloshin. This episode was produced
by Liza Dennis and Tyler Hill. It was executive produced
by me and Julian Nutter for Kaleidoscope and Katrina Norvell
for iHeart Podcasts. Jack Inslee mixed this episode and Kyle
Murdoch wrote our theme song. A special thank you to
all our listeners. Please rate, review and reach out to
us at techstuffpodcast at gmail.com. We love hearing from you.
Speaker 2 (39:20):
Thank you.
Welcome to Tech Stuff. I'm Oswald Loshin. Back in 2024,
a 71-year-old Parkinson's patient named Jeffrey Kiefer underwent brain surgery.
While his skull was open, surgeons temporarily placed an experimental
brain-computer interface directly onto his brain to record neural activity,
a key step towards technology that could one day allow
(00:35):
paralyzed individuals to control computers using only their thoughts. The
following year, Precision Neuroscience, the startup behind that device, reached
a big milestone, FDA clearance for 30-day human trials. As
of this year, testing has expanded to nearly 100 participants
across 18 major U.S. academic medical centers. Joining me today
(00:57):
to discuss how this technology works and where it's heading
is Michael Maker, CEO of Precision Neuroscience. Michael, welcome to
Tech Stuff.
Speaker 2 (01:06):
Thanks so much for having me.
Speaker 1 (01:06):
What are you building?
Speaker 2 (01:08):
We're building, as you mentioned, a brain-computer interface. That is
a system that I think is best thought of as
a communication tool. It creates a communication link between the
brain and computers. The first use of that link is
to enable people to control computers with their thoughts, people
who have lost ability to move their arms and hands
(01:29):
as a result of, you know, paralysis due to spinal
cord injury and certain diseases. But we think the applications
of this communication link effectively between human and artificial intelligence
go way beyond paralysis.
Speaker 1 (01:41):
So they said everyone might have one?
Speaker 2 (01:43):
I think the future is hard to predict. We see
a path to having major impact clinically and potentially eventually
beyond that. But we're really focused today on making a
meaningful positive difference medically and clinically.
Speaker 1 (01:57):
Talk about that journey from working with Jeffrey Kiefer to
the FDA clearance to the 100 participants? What does that
mean that you get to work with 100 people now?
Speaker 2 (02:06):
Taking a step back, I think that it's important to
understand medical technology requires people like Jeffrey Kiefer, who are
really selfless in offering themselves to the advancement of medical technology.
The reason that he was able to get an implant
of this kind is that our system is is designed
(02:27):
to be totally safe, it is reversible, and it is non-damaging.
And because of that, we were able to get into
the clinic and start implanting patients within two years, which
is a very, very expedited path. And our first FDA clearance,
as you mentioned, within about 40 years of founding. That
is really a function of the fact that our system
(02:48):
is able to be explanted safely without having done damage
to the brain.
Speaker 1 (02:53):
Explanting means you can take it out.
Speaker 2 (02:54):
That means you can take it out. Yeah. And I
think that that's an incredibly important feature. Really, when it
comes to patient adoption, you know, people want a system
that they know if they, for whatever reason, don't want
can be removed safely. And also for upgrades, you know,
this is a technology that is in the early stages
of its development. I think people rightfully want a system
(03:15):
that's going to be able to be upgraded over time
as functionality improves.
Speaker 1 (03:20):
And who founded the company? What was their vision?
Speaker 2 (03:22):
So I co-founded the company with Ben Rappaport. Ben is
really ideally suited to pioneer this industry. He is a neurosurgeon.
He also has a PhD from MIT in electrical engineering.
That is not by accident, you know, getting both a
medical degree and a PhD in electrical engineering. The brain
is an electrical system. So every thought you've ever had,
(03:45):
every idea, every memory, every intention, there is a physical
representation of of those thoughts.
Speaker 1 (03:52):
By the way, it's a remarkable coincidence that both computers
and the brain work on electricity. When did people start
to understand what that could mean?
Speaker 2 (04:00):
A hundred years ago, I think, is when we started
to understand the electrical nature of the brain. I think
that for a long time, these two fields have been
quite disparate. I think we're actually entering a period, and
I'm sure we're going to talk about this, where there
is really a convergence of neuroscience and artificial intelligence that
has the potential to be impactful really on both sides
(04:21):
of the coin.
Speaker 1 (04:22):
But your co-founder, in a sense, this is a substantiation
or a harvest of decades of academic research and vision.
Speaker 2 (04:30):
That's exactly right. I mean, he really was there at
the birth of this industry, which initially happened like so
much in terms of sort of breakthrough advances in academia initially.
where the first proofs of concept that you could record
and decode electrical activity in order to drive a function
on a computer actually happened in 2004, so over 20
(04:53):
years ago. And the industry has been developing really in
academic settings with prototype devices. Ben wanted to be involved
in that from the very beginning. He studied electrical engineering
and became a neurosurgeon because that's the perfect combination of
skills to pioneer this industry. He was also one of
the co-founders of Neuralink with Elon Musk. And after a
(05:15):
few years at Neuralink, he left because of the direction
that they went in. And he and I co-founded Precision
in 2021. So what was the direction they went in
that he objected to? So the initial work in the space, again,
going back to sort of the academic proofs of concept,
involved electrodes to record and stimulate the brain that penetrate
(05:37):
into the brain.
Speaker 1 (05:38):
Right.
Speaker 2 (05:39):
And when you penetrate into the brain... a few things happen.
First of all, you kill neurons. So there is damage
that is irreversible. There's also a response from the brain.
The body doesn't like it when you put a foreign
object into it. And so there's something that's called gliosis,
which is effectively sort of scarring around the foreign body.
Over time, if you think about creating an interface between
(06:03):
the brain and technology and compute, you really want a
system that's going to be able to scale. that's going
to be able to start in a small discrete area
of the brain, but over time expand because different areas
of the brain are responsible for different functions. And so
to unlock greater and greater performance, you need a system
that's going to be able to expand to multiple regions
(06:24):
of the brain at once. If you have to poke
holes in the brain to do that, that just, there's a,
there's a ceiling to the scalability of such a system.
There's also a patient adoption and surgeon adoption issue. You know,
When you're talking about any surgical intervention, there is a
necessary weighing of risk and benefit. And you want that
(06:47):
trade-off to be as favorable as possible.
Speaker 1 (06:50):
When you're talking about this idea of reaching all the
regions of the brain, what came into my brain was
the image of a metal helmet underneath my skull. Is
that a crazy image?
Speaker 2 (07:00):
Metal helmet is not how we would describe it. So
what we've developed is an ultra-thin film. Think about it
like a piece of saran wrap. in your language, maybe sellotape?
Speaker 1 (07:11):
Yeah, I know, cling film.
Speaker 2 (07:12):
Cling film. Cling film.
Speaker 1 (07:13):
Sellotape is something, yeah.
Speaker 2 (07:16):
That basically covers the brain in a super, super gentle,
thin film with thousands and eventually hundreds of thousands of
electrodes on it. It does so in a way that
is fundamentally sort of non-damaging.
Speaker 1 (07:32):
What are the sort of planes of competition, let's say?
Because obviously Neuralink you know, because of Elon, probably because
of first mover is the most famous player in the space.
There's you, there's maybe two other kind of household names
of people who know the industry, right? Like, are you
all chasing exactly the same holy grail or are there
different planes of competition, so to speak?
Speaker 2 (07:53):
I think there are absolutely different planes of competition that
will emerge over time. So for the time being, you're right.
Everybody is focused on this initial question really around the
hardware of how do you get electrodes into or onto
the brain in a way that's safe and scalable. And
our system, unlike Neuralink's, which is penetrating into the brain
because they've built on that academic work that sort of
(08:15):
founded the industry, ours creates a non-damaging thin film. Others
go through the vasculature. Some folks are trying to use ultrasound.
I think that there will emerge clear areas where certain
approaches are better than others, and companies will necessarily focus
where their systems are sort of optimized. That said, We've
(08:38):
picked the approach that we've picked because we think it
is really the only one that achieves ultra high performance
for all of the functions of BCI that exist today,
but without some of the drawbacks around penetration and damage
that some of the competing products entail.
Speaker 1 (08:55):
So you mentioned saran wrap. What does this thing physically
look like? How big is it? And how do you
get it into somebody's head?
Speaker 2 (09:02):
We manufacture the system that is about 20 microns thick.
So that is about a fifth the width of a
human hair. So super thin. And it's about one and
a half square centimeters in terms of surface area. So
think about it like a postage size stamp. We deposit
1,024 tiny platinum electrodes on that. These electrodes are 50
(09:24):
microns in diameter. That's the size of a neuron. So
we're really speaking to the brain at its own natural scale.
The only way to manufacture at that scale is to
do so using manufacturing techniques borrowed from semiconductor manufacturing. So
we use photolithography to make these devices. We're the first
company with an FDA-cleared device that is photolithographic.
Speaker 1 (09:47):
Is ASML the same providers as the chip industry? We're not.
Speaker 2 (09:52):
So those guys are at the nanoscale. We're at the
micron scale. But the concept is the same, where basically
to create... a breakthrough, in this case in medical technology,
you really have to think about it in a different
way and take techniques that have been pioneered elsewhere. In
our case, that's required actually buying and retrofitting a fab,
(10:13):
which we own in Texas. We have 20 people on
the ground who do.
Speaker 1 (10:17):
Which used to be used to build chips is now
being used to build your.
Speaker 2 (10:19):
That's right. And we have a level five clean room.
All of this is necessary. It's capital intensive. It's resource intensive.
But there is no supply chain that exists to do
this kind of work because it's so pioneering.
Speaker 1 (10:32):
How do you put it inside somebody's head? And is
the vision to make them much bigger ultimately? Yeah. And
how do you know where to put it?
Speaker 2 (10:39):
First of all, I don't put them in anybody's brain.
I will assure you. But we're now working at 18
of the best hospitals in the United States, which by
the way, is multiples of the number of any other company.
Part of that is enabled by our FDA clearance, which
also is a first. And we're working with neurosurgeons at
these institutions who are absolutely best in class. They are
(11:01):
using the system in different ways. And part of the
benefit of a system that does not penetrate into the
brain and does not damage the brain is that you
can explore. You can put it. in the area that
controls the motor part of the brain. So the part
of the brain that controls the arms and the fingers
and the hands and use that to control a computer
(11:22):
cursor just with thoughts, which we've done. But you can
also put it in areas of the brain that are
much less well characterized. We have gone to somatosensory cortex
and prefrontal cortex. We've gone to the brainstem. We've actually
even gone to the spinal cord. That is the only
way that you can explore in that manner and you
can enable academics and clinicians to explore in that manner
(11:45):
is if your system is fundamentally non-damaging and reversible. So
we think this is a huge advantage for precision, but
also for the field, because I think it's going to
unlock understanding that previously has just not been possible.
Speaker 1 (11:58):
What are you finding today by going to those places?
Speaker 2 (12:01):
I think we're at the beginning of this journey. I
think a lot of our efforts up until now have
been focused on that motor part of the brain. So
you asked the earlier question, is everyone focused on the
same thing? And today, the answer is largely yes, where
we're really focused all on paralysis, enabling people whose brain-body
connection has been disrupted by a disease or an injury.
(12:22):
So think about someone who's in their 20s, they're in
a catastrophic car accident, their brains function as well as
yours or mine probably better, but they have lost the
ability to control their arms and hands. And so they
can't use technology in the way that we all take
for granted. So you can't use a phone, you can't
use a tablet, you can't use a computer. Our goal
is to give people in that situation control again over
(12:46):
a digital ecosystem to enable them to have higher quality
of life and eventually to go back to work if
they want. The FDA is effectively pushing us and others
to addressing that initial use case first, because we know
it works. The risk benefit is just very obviously favorable,
(13:07):
and it's a way to get this technology out into
the world in a way that's going to make a
real positive difference.
Speaker 1 (13:12):
Pushing by providing grant funding or pushing by not delaying
things too much.
Speaker 2 (13:16):
I'd say encouraging. I'd say the FDA has been extremely
constructive But it's clear that we all should start in
this place. That said, as this technology gets out in
the world and as it advances, as discussed earlier, I
think different companies are going to end up focusing in
different areas. We are focused initially on the motor part
(13:37):
of the brain because that is our first clinical application.
And so decoding intended movements to drive a digital keyboard
and digital mouse effectively is where we've started. But Now
that we have established that functionality, we have the opportunity
to look beyond it, and that's something that we're in
the early days of doing.
Speaker 1 (13:58):
There was a New Yorker story, I guess probably in 2018,
about the scientist at Carnegie Mellon who was leading. You
know his name, I guess, but I can't remember it
off the top of my head.
Speaker 2 (14:07):
There are a few fantastic ones there.
Speaker 1 (14:09):
I met him, and it was a New Yorker story
about how two young people had become paralyzed, both had
the ability with this electronic array placed into their brain
to control artificial limbs and maybe it was University of
Pittsburgh actually but the piece had this kind of lyrical
but ultimately quite tragic quality which was that it stopped
working because whatever the word you use for the scar
(14:31):
tissue in the brain would essentially grow over the arrays
and like this miraculous experience like had its limit and
that was a second sadness essentially in that but what
you're saying is we may be on the cusp of
that not being true.
Speaker 2 (14:46):
I think What we and a few others are doing
is taking that amazing proof of concept and productizing it
because that was achieved. And the other experiments like it,
places like Carnegie Mellon and University of Pittsburgh and a
few other centers of excellence, using prototype devices that involved
(15:06):
parts of the system actually protruding from the head, wires
often physically connecting the brain to external compute. There's all
sorts of issues and risks associated with that, with infection
being sort of the most obvious one. And the systems
were nowhere near robust enough to go through the FDA
regulatory process. So the positive sort of take on this
(15:32):
is this technology works and is going to make a
huge difference to a group of people who right now
have nothing. The negative is it's been validated in academia
for two decades, and about 100 people have been implanted.
I mean, incredibly poor real-world impact. And so I think
it's our job to make sure that the next 20
years are not like the previous 20.
Speaker 1 (15:53):
How many of these implants do you think you may
do in the next 20 years?
Speaker 2 (15:57):
The path for precision is initially to start with severe paralysis.
There are 350 or so thousand people in the United
States and several million globally who are completely paralyzed. So
unable to use arms and hands at all, but still
sound of mind. I think the next expansion from that
(16:20):
is towards partial paralysis. So think about, for example, someone
who's had a stroke. So they can use one side
of their body, but not the other. Again, that means
severe limitations in terms of an ability to control a keyboard,
a mouse, a touchscreen. And that is a population that
is five to 10 times larger than the initial severe
(16:40):
paralysis population. So just in those two groups, which are
our initial two areas of focus, you're talking about millions
of people in the United States and tens of millions
of people globally. I'm very excited over time to address
some of the neurological illnesses that we either have no
treatments for now or that they're very, very poor. So
(17:03):
two examples I'll give are one is refractory depression. So
depression that doesn't respond well medication. The idea of using
electricity to disrupt neural circuits that are not operating properly
is extremely well established. It's electroconvulsive therapy. You're talking about
putting electrodes on people's skulls. It's very coarse. There are
(17:25):
significant off-target effects, which is sort of a euphemism for reality.
And yet, it's still used as standard of care because
it works.
Speaker 1 (17:33):
So what you've been talking up until now is output,
decoding the thoughts from the brain to create action in
the world. What you're now talking about is input.
Speaker 2 (17:40):
That's exactly right. I think the way that a really
high-performance depression-focused system must work is bidirectional. So you need
to be able to, instead of just introducing current to
the brain, irrespective of brain state, you really want to
understand when the brain needs intervention. And then when it does,
(18:02):
do so as precisely as possible to limit the sort
of the side effects of that intervention. I think that
is only possible with a high-performance, high-bandwidth brain-computer interface.
Speaker 1 (18:14):
And when might UI want to get one of these
voluntarily and why?
Speaker 2 (18:17):
I think the short answer is there's already people who
want one. That said, I think that that's, you know,
Neuralink has publicly said they have a 10,000-person wait list
for their system. I guarantee not all those people are
severely paralyzed. we all get inquiries from folks who are
interested in controlling computers in a more seamless way. I
(18:38):
think the kind of maybe superficial way of describing it
would be to say, well, we have always evolved the
interface between computers and human beings, and that's unlocked new function.
So like mainframe computers, we had punch cards. With the
desktop and the laptop, we have a keyboard and a mouse. Increasingly,
(19:00):
we control computers with smartphones by tapping on a little
black screen. And all of those shifts have created whole
new areas of function, new applications. I think the end
state is direct neural control. That said, would I or
would I recommend my family get an implant simply to
(19:20):
be able to control a computer cursor a little bit faster,
a little bit more efficiently? No, I think most people
would not. I think where the really interesting and exciting
developments lie is around being able to interact with technology
in ways that are just not possible right now. So
I'll give you an example. We all use AI in
(19:43):
one way, shape, or form, largely through text prompts. You
either speak a prompt or you... type it. That involves
taking an idea and basically condensing it into words and
outputting those words to a prompt. And what I think
everybody experiences is that the output is magical in some ways,
but often off.
Speaker 1 (20:04):
Lost in translation.
Speaker 2 (20:05):
A little bit. And you got to keep going back
and forth and Eventually, it usually gets it. But the
reason for that is at least partly because the AI
doesn't have all of the context that you're bringing to
this request. It doesn't have your memories. It doesn't have
your intentions. It doesn't have your perspective. It can't understand
(20:28):
all of that. And it's very lossy to take an
idea and condense it into a few sentences and a
text prompt. I think where... we are headed is a
system that can decode and interpret higher level mental processes.
You know, we're starting with more mechanistic sort of motor control,
but if you can decode the motor part of the brain,
(20:49):
you can also decode some of these more sophisticated mental
processes over time. And I think that that will enable
us to interact with AI in ways that just aren't
possible without a system like this.
Speaker 1 (21:02):
But so, I mean, talk about that because the, I mean,
It's obviously one of the hardest science challenges humanity has
ever solved to decode the brainwaves into electricity and turn
it into output information to move a mechanical limb. I mean,
no one would take away from how extraordinary that is.
But ultimately, like, I know I want to raise my
hand is different from, I know I want to sit
(21:23):
here with Michael and continue the conversation, but I'm also hungry. Like,
you know, like, should I go for lunch or should
I stay? I'm going to stay because I have a
higher order brain function that makes the social norm is
that we continue this interview even though I'm hungry. First,
what's been done now and what hasn't been done? But secondly,
how do you begin to bridge that gap?
Speaker 2 (21:40):
It's a fantastic question. And I only have partial answers,
but I'll sort of offer you what I have. That
is absolutely true. It's part of the complexity of the
brain and interfacing with the brain. And it's actually present
even in the work that we're doing now. So we
brought on, as an advisor, we announced this a couple
months ago, John Donahue. He was the pioneer. John is
(22:03):
the godfather of BCI. He did that sort of initial
2004 implant and most of the other ones. He's seen
it and done it all. And he joined Precision for
a variety of reasons, which I can get into. But
talking to him about some of the experiences, you know,
what they found, especially in their earlier systems, was if someone...
(22:23):
was watching television. They had, you know, one of the
participants that he worked with had a favorite soap opera.
And every time the soap opera would come on, it
would be on, you know, sort of in the corner
of this person's eye, but he'd end up watching it
unbeknownst to the researchers who were working with him. And
the decoding of his motor performance would degrade. So, you know,
(22:46):
that is absolutely a real phenomenon that we all have
to deal with. That said, The systems that we have
now on the algorithmic side are just orders of magnitude
more powerful than even the ones we had two years ago,
let alone the ones we had 20 years ago. That's right.
And that's what's so interesting that I think artificial intelligence
(23:07):
is enabling us to understand aspects of human intelligence that
just weren't possible even a few years ago. But this
is a question of more and more data applied to
cutting edge algorithms to help us untangle these different mental
processes and have systems that work to our benefit.
Speaker 1 (23:49):
I read somewhere, and this may be wrong, but that
Neuralink had a kind of proprietary data set of chimpanzee
brain research that you guys don't have, and therefore that
gives them an edge. Like, how does the AI with
reinforcement learning, et cetera, how do you teach it what
the signals mean?
Speaker 2 (24:03):
Yeah. Neuralink made a choice to work with non-human primates
and monkeys. We have not. Our animal model is pig.
And part of the reason that we didn't work with
monkeys is because we were able to get into humans
within two years. It took them, you know, almost triple
the time. And that's, again, just the function of the
fact that our system is non-damaging. And so the barrier
(24:25):
to doing a first implant was much lower. We have
really built on that fundamental advantage. And I would say
that our clinical ground game is by far the best
in the industry. So our velocity of implants is rapidly increasing. That, again,
on top of the fact that we have a vertically
(24:45):
integrated supply chain, This whole clinical operation is resource intensive.
It's capital intensive. It's really hard. You know, companies that
come from more of sort of an engineering or tech background,
I think really struggle to understand how to work in
hospitals in ways that aren't disruptive. You know, you have
to get invited into these systems in order to work.
(25:08):
And I think having been as, you know, the chief
science officer and co-founder of the company lends a huge
amount of credibility within this community to the work that
Precision is doing. So I think the answer to your
question is, yeah, non-human primate data is great, but what
you really want is human neural data. And I think
(25:29):
we have access to that at scale and with diversity
across patients that has just never before been done in
human history. And again, this is all focused today on
motor decoding, and it's all based on patients have to opt-in.
So it's a consent-based system. It operates all under the
(25:49):
review boards of the institutions where we're working. And it's
all de-identified data that complies with HIPAA. So I want
to say that because I know that this stuff can
get understandably concerning when people hear about neural data. But yeah,
what we're talking about is really sort of correlating what
these electrical signals in the brain actually mean and the
(26:10):
extent to which they are representative across different people, which
it turns out for motor systems, they really are.
Speaker 1 (26:18):
I should have asked you this earlier, but the array,
how does it communicate with the outside world from inside
the human body?
Speaker 2 (26:25):
The people that we're implanting today are undergoing a neurosurgical
procedure anyway. So they're in the hospital. For example, they're
having put out a demo a couple months ago of
someone who was undergoing a surgery because he had a
benign brain tumor. And so while he was in the hospital,
The surgeon said, listen, we're working with this company, Precision Neuroscience.
They have an FDA clearance. Would you like to, while
(26:47):
you're in the hospital anyway, be implanted and learn how
to control a computer with your thoughts? And about 50%
of people say, that sounds amazing. I'd love to. And
about 50% say, no, thank you. The system that we're
implanting in those cases is wired. So the array is
bonded to a little black package that sits outside the
head that has a wire that goes to a computer.
(27:11):
It is the same signal chain and the same basic
architecture as the fully implanted system. But the fully implanted
system needs to be hermetic and it needs to stay
in the body forever. So you have a different level
of sort of testing rigor that you have to go
through in order to get a fully implanted system approved
for human use. We're in the process of doing that.
(27:32):
You know, we're going to have more news on that soon.
But what we're able to do uniquely is basically demonstrate
that our system is able to perform at the highest
level today and not have to wait for, both on
the performance validation or on the data collection side, the
time that the permanent implant is ready.
Speaker 1 (27:51):
I was at a talk that was an interview with
Demis Hassabis and Reid Al-Brigotti from Semaphore a couple of
months ago. And Demis was talking about decoding an image
a person is thinking about or dreaming of. and then
using a model to recreate the visuals, and then asking
somebody in the scanner what they're thinking about, and it
is indeed the thing that the machine saw. Kind of mind-blowing. Yeah.
Speaker 2 (28:13):
I mean, it turns out that some of these things
are a lot easier than others. It turns out that
sort of, you know, mind's eye representations are quite prominent
signals in the brain, whereas other mental processes are more subtle.
But yeah, it is. It's amazing. And I think it's
exciting for a variety of reasons, but also something that
(28:33):
we need to be very careful about.
Speaker 1 (28:35):
Are you building cyborgs?
Speaker 2 (28:38):
I would say cyborgs have existed for 70 years. Anyone
who has a pacemaker is in some ways a cyborg
or a spinal cord stimulator or a deep brain stimulator.
I think the idea of technology addressing... human health challenges
is not a new one. This is, I think, an
(28:58):
evolution in that sense, not a revolution.
Speaker 1 (29:01):
But you alluded to people's concerns. What are your fears?
How could this, if not properly controlled, what are some
of the negative scenarios?
Speaker 2 (29:12):
I mean, I think what we're doing is, and not
just us, the other implantable BCI companies, I think is
an unambiguous positive. I think offering a tool to people
who right now have no therapy available to them is terrific.
But like any tool, it can be used, a hammer
(29:34):
can build a house or it can you know, harm
someone else. And so I think that there are for
sure legitimate concerns around the evolution of this technology, both
in the initial sort of read-only setting, but I think
as you alluded to, when it comes to introducing current
to the brain, so write versus read, I think that
(29:55):
opens up a whole nother set of concerns. I think
it's important to remember that this is a technology, though,
that people have to opt into. No one's going to
be forced to have a brain-computer interface. And we as
a company won't succeed, nor would we ever want to
if we were doing things that were not ultimately in
the people who adopt the technology's best interests. That said,
(30:18):
I don't think this is going to be a case
where companies self-regulate. There are already a number of different
groups that are coming together to help ensure that this
technology is developed in the right way. One of those
is the FDA helps convene something called a collaborative community,
which is basically all the different stakeholders that are important
(30:39):
to a new area of technology get together. And ethics
and data security is one of the most prominent work
streams of that collaborative community. And I think there's some
incredibly thoughtful work that's coming out of that. Similarly, I'm
part of a group at the World Economic Forum that
is also thinking through and going to soon propose guidelines
around neurotechnology, data privacy, and data security. So there's a
(31:03):
sense that we got social media and a lot of
other areas of technology wrong. The horse is bolted, and
we need to make sure that we do this one right,
which I think is the right instinct.
Speaker 1 (31:14):
You talked about read, and the current paradigm for read
is like motor. If you're paralyzed, you have this amazing
opportunity to take action in the world with your thoughts.
Maybe the next generation of read is what Demis is
talking about, where your visual imagination can be directly interpreted.
So for example, making a movie rather than decoding into
(31:34):
words and then going back from words to images, which
happens with a director, you could just download directly from
your imagination potentially.
Speaker 2 (31:41):
Totally. I mean, I think some of the people who
have described what they believe the future of BCI entails
is along the lines of what you described, which is,
you know, right now, I described how we communicate with AI,
but it's actually how we communicate with each other also, right?
We have ideas. We have a picture in our mind's eye.
We have a complicated concept. And in order for me
(32:01):
to communicate that to you, we have to condense that
thought or idea or image into words. And then we
either like flap our mouths or we tap our fingers
in order to communicate to somebody else. Imagine being able
to project an image or an idea. I think that,
you know, that has the potential to unlock. I think
(32:24):
it could be incredibly positive for humanity. It's kind of
scary because no one has ever done that before and
it's new. But I think that that potential exists.
Speaker 1 (32:32):
So what's the far horizon of read as it pertains
to decoding the brain? What's sci-fi today that might be
science fact five years or ten years from now?
Speaker 2 (32:39):
I mean, I think about that, honestly, more clinically than
I do in any other way. you know, the brain today,
neurological illness is the largest source of sort of economic
loss in the world. You know, the WHO estimates one
in three people at one point or another have a
serious neurological illness in their lives. And, you know, the
(33:03):
brain is still like basically offline and analog. And when
it goes wrong, we have either no interventions or very,
very sort of coarse ones. If you think about medicine
more generally, and especially it feels like it's accelerating in
an exciting way, we're making such rapid advances. And they're
really driven by digitizing biology and then applying compute in
(33:26):
a way that unlocks breakthroughs. So mRNA, the incredible performance
of the potential cancer therapeutics and the whole genomics revolution
is a great example of this. There's so many areas
of medicine that I think are being transformed effectively by compute,
not the brain. the brain remains impervious to commute. It's
hard to get to. So it's like physically very well
(33:47):
protected by the skull. And then once you get to it,
it's delicate. And so we have not up until now
had a tool to digitize the brain so that we
can apply compute to it and understand what's going on.
And I believe intervene in much more precise ways when
(34:08):
it's not behaving properly. So my vision for the next
five years, what I hope that we can head towards
is a world where neurological illness generally is much less debilitating.
Speaker 1 (34:22):
And is organoids like a parallel research path? I mean,
is that like there's the organoid crowd and there's the
BCI crowd or like what's the.
Speaker 2 (34:29):
You know, I'm not an expert on organoids. I think
they certainly have a place in the research. I think
that they are proxies in some ways, but maybe not
as good proxies as we had hoped for for the
full complexity and dynamism of all sorts of different organs,
including the brain. I hope that that also is an
area that produces breakthroughs, but I think it's not an
(34:53):
area that I'm expert in.
Speaker 1 (34:55):
The Wall Street Journal wrote, and to be fair, this
was back in 2024, other than perhaps the quest for
human-level AI, well, a show that is a bit dated,
or colonization of other planets, Not coincidentally, two other areas
where Musk is a big proponent. Few fields exhibit such
a wide gulf between technology's potential and its near-term prospects. Obviously,
(35:18):
written in 2024 and showed his age with the first
part of the quote, but what about the second part?
Speaker 2 (35:23):
I think going back to what we described earlier, two
decades of academic proofs of concept, and yet fewer than
100 people walking around with a BCI. Yeah, I think
that that's incredibly disappointing. I think that it's our job
to make sure that we don't look back five years
hence and think that that comment remains accurate.
Speaker 1 (35:42):
And yet other people have said 2026 will maybe be
the chat GPT moment for BCIs. Why do they say that?
I don't know.
Speaker 2 (35:49):
I don't say that. But I think that a chat
GPT moment is likely to come. I think if you
take a step back, so I think one of the
most important and valuable assets that we have as a
company is this pipeline to collect data at scale. This
incredibly valuable, first of its kind on earth, high resolution
(36:12):
neural data. And we've used that to implant nearly a
hundred patients. And, you know, that's multiples of what anyone
else says, but a hundred patients is a hundred patients.
It's not, I mean, We need to get to thousands
and tens of thousands and hundreds of thousands over time.
I think that is exactly what we're set up to do.
(36:33):
But I think in the context of big data, you're
looking at further scaling up. And we don't know exactly
where the numbers start to tip and the breakthroughs really
start to come. I think we're excited to find out, though.
Speaker 1 (36:46):
What does a world in which you deliver on your
mission look like?
Speaker 2 (36:51):
In the first instance, and I think about this in stages,
in the first instance, some of the illnesses that we're
going after. So if you think about paralysis and other
BCI companies are going after blindness, these are like biblical afflictions.
These are things that have been debilitating and irreversible since
human beings existed. And I think in the next five years,
(37:15):
we have the potential to really change what it means
to be disabled. And I think that's incredibly exciting. I
think that should be doable based on the technology that
we have today and the trajectory that we're on. I
think 10 to 15 years is more along the lines
of what we were discussing earlier around taking neurological illnesses
(37:39):
that maybe have poor therapies today or no therapies that
are debilitating in different ways, but which we have very
little understanding of and very little ability to intervene in.
And really changing those and making those sorts of neurological
illnesses addressable. I think we have the potential to reduce
(37:59):
suffering enormously. Over time, do I think that this will
be a system or an approach that gets adopted by,
you know, for general use? Probably. I think so. I
think we're going to need to see continued performance breakthroughs
to justify, you But I think that those are quite likely.
(38:20):
I do think in that sort of future state, having
a system that is fundamentally safe and non-damaging and reversible
are going to be table stakes for a system to
achieve adoption.
Speaker 1 (38:33):
Michael, thank you.
Speaker 2 (38:34):
Thanks so much for having me.
Speaker 1 (38:52):
For Tech Stuff, I'm Oz Veloshin. This episode was produced
by Liza Dennis and Tyler Hill. It was executive produced
by me and Julian Nutter for Kaleidoscope and Katrina Norvell
for iHeart Podcasts. Jack Inslee mixed this episode and Kyle
Murdoch wrote our theme song. A special thank you to
all our listeners. Please rate, review and reach out to
us at techstuffpodcast at gmail.com. We love hearing from you.
Speaker 2 (39:20):
Thank you.
TechStuff News
Hosts And Creators
Oz Woloshyn
Karah Preiss
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.