Latest / Elon Musk Thinking / Neuralink Presentation With Elon Musk!!!

Transcript

Neuralink Presentation With Elon Musk!!! · Elon Musk Thinking

0:00
The Monkey Order's room service. Yeah, we'll won't be all that long before
0:04
someone with a neuralink device can communicate faster than someone who has a fully functional
0:10
body. And the goal is to give people superpowers. We just want to
0:14
be clear if there's only one person with a neural link chip in their brain.
0:18
So for people out there who think we've put a chip in their brain,
0:21
would like to assure you for what it's worth, you probably won't believe
0:25
us, but we did not put each other in your brain. Okay,
0:29
welcome to Neuralink Live Update. We're going to tell you about the progress of
0:35
the first patient with a neuralink ioland and sort of your recap of the progress
0:42
there, then talk about what changes we were making for the second patient,
0:48
which we're hoping to do an implant in the next week or so. And
0:53
this is for our first product, which is called Lefathy, which enables you
0:56
to control a computer or a phone just by thinking. So let's in fact,
1:00
so we'll start off with just some introductions. DJ, you wants to
1:04
hi everyone. My name is d J. Shull' an electrical engineer and a
1:07
chip designer by training. I led the design of first several generations of the
1:14
Neuralink implant. Currently, I was on the founding team and currently a president.
1:19
I'm Matthew McDougall. I'm a practicing neurosurgeon and a head of neurosurgery at
1:23
Neuralink. Yeah, go ahead, a head of burning the first as applications,
1:30
and I'm blessed a software engineer at Nerlink trying to figure out how to
1:33
turn bern activity into cool stuff. All right, thank you, Well,
1:37
so let's see, so we'll just get going into the presentation. So our
1:42
first product is sort of like SAP we call a telepathy, which is enables
1:49
the person with the Neuralink implant to control their phone or computer just by thinking.
1:53
And once you can control your phone and computern essentially control almost anything just
1:59
and literally just thinking. So there's no eye tracking or anything. It is
2:02
purely purely your thoughts. So a this is a really quite quite a profound
2:12
device that can help a lot of people who have lost the connection between their
2:15
brain and body. So imagine people like h Stephen Hawking, who you know,
2:23
I imagine if he if he could communicate at the same speed as someone
2:25
who had still had the connection to their brain and body. So it's it's
2:30
really something that can help millions of people around the world, and it's a
2:35
it's part of our overall goal of enabling a very high bandwidth connection between the
2:42
brain and and your and the rest of the world in your computers. The
2:47
long term goal, which sounds a little esoteric, is to mitigate the the
2:52
risk of of the civil civilizational risk of AI by having a sort of closer
3:00
symbiosis between human intelligence and digital intelligence. But but that that'll that will take
3:06
many years. Along the way, we're going to help solve a lot of
3:13
brain injury or spinal injury issues. Uh so uh and with without first product
3:19
sympathy, that's that's going to be really quite profound. Uh that there is
3:23
also potential long term for bridging the gaps if there are damaged or severe neurons,
3:31
being able to span the gap between the brain's motor cortex to the spine
3:36
to enable someone to use their body again. I think that would be very
3:39
exciting, and it's you know that that is something that is possible in the
3:46
long term and they're not. Our second product, which we've demonstrated to work
3:53
with monkeys, is blind sight, which would enable someone who is completely blind
4:00
lost both eyes or completely lost their optic noeve to be able to see.
4:04
So that's that's something that we hope to demonstrate in the future. So this
4:10
just gives you a sense of what the device is. Uh. A way
4:14
to think about the neuralink device is kind of like a fobit or an Apple
4:16
watch with with tiny wires or electrodes. Those those tiny wires are implanted in
4:24
the in the brain and they read and write electrical signals. So a lot
4:30
of people think the brain is this incredibly mysterious thing. It's it is mynsterious
4:35
a lot of ways, but but it is actually it does operate with like
4:39
electrical signals. So if you can read and write those electrical signals, you
4:43
can interface with the brain. And the devices is sized so that it is
4:49
the same size as the as the piece of skull that is removed. So
4:54
if it's like a few centimeters diameter of skull skull that's removed, we replaced
4:59
that with device after implanting the tiny wires with the surgical robot, and that
5:04
enables read write capability to the neurons completely wirelessly. YEA, yes, exactly,
5:12
it's completely wirelessly. So like I could. I could have a neural
5:16
ink right now, you wouldn't know, and it charges inductively, so you
5:20
could just basically have an electromagnetic add that that that you charge the device with.
5:28
So yeah, it's like an Apple watch exactly. So uh, except
5:32
that it's actually a much harder technical challenge to solve that there's limit as to
5:38
how much heat the brain tissue whereason for phones and really care sitting on the
5:44
people. Yeah, so so it's got to go through skin and stuff as
5:47
well. And okay, so it's it is a tougher challenge to to charge
5:51
and to have iband with communications given that it's got to go through skin inherent
5:58
stuff we have we have yeah, so yeah, yeah, So our first
6:05
step with the Telepaphy is basically to unlock digital independence for people with prolysies and
6:13
to allow them to control the computer just with their mind without moving their body.
6:17
And our goal is to provide them the same level of control, functionality,
6:23
and reliability that I have when I'm using a computer, even better than
6:28
the level of control I have. Just to be clear, this guy he's
6:32
been following, that's with his brain, so he's not like you can't see
6:35
his hands in this video, but he's not using a loss and keyboard thinking
6:40
about how to move the cursor and time Civilizations. No eye tracker, right,
6:43
there's no I tracking from just thinking like a couple of days a cursor
6:50
move here. Yeah. Yeah, it's like the last standard two nights ago.
6:54
Yes, yeah, I think I think the way he also described it
6:57
is. Yeah, he is many more videos on his the platform. Definitely
7:03
checked them out. Yeah, so you can he's string that live and also
7:08
can talk and like movies had with that. Yeah, you guys, like,
7:12
if you join this live stream, you can ask him questions. You'll
7:14
you'll tell you all about what it's like to right. Also, I think
7:17
I haven't played Civilization myself, but I think this is actually not easy.
7:21
But this is expert Emperor mode. Emperor mode if you have played stiff ePRO
7:26
mode is like the highest. Just the point is that this is a kind
7:29
of way demanding task and while live streaming playing the hardest movie game, and
7:32
he's able to do that while movie talking, engaging with the audience. One
7:39
of the other games he likes to play a lot of chess. I think
7:42
it gets lost sometimes that he's actually playing speed chess against me, which requires
7:47
an incredibly high fidelitygree of control and speed of control in order to be able
7:54
to win. Also another cool stuff about our devices that use it anywhere anytime,
8:03
also on a plane during a flight while reading really cool memes of care.
8:09
Also, our device unlocks things that previously were impassible for our participants.
8:16
For example, we're able to connect him to his living consoles which play Marricat
8:22
with friends and family, and it was lovely to see them playing together after
8:26
years that couldn't do it since he's injury. I know if you're sitting over
8:31
from this guy playing look over in a cat meme, not no hands,
8:37
no movement even away over. Yeah, it's change strange time. Yeah.
8:45
And he loves playing using the device and using independently daily to watch videos and
8:50
read play games using the baser. And the key metrics that we care is
8:58
to make sure our device is actually useful to basically the amount of hours to
9:01
use the device daily and weekly, and we track it weekly since the since
9:05
the surgery, and on weeks that he's not too busy and not traveling,
9:11
he can even reach seventy hours using the device a week. This is amazing
9:16
and he would of course left to use it more, but need to run
9:20
result sessions. He needs to sleep sometimes and also of course to charge the
9:24
device once in a while. Hopefully will improve that over time. I think
9:28
maybe not obvious to people who are watching this, like it's a normal map
9:31
book he's controlling. This isn't like some limited thing that there's only a few
9:35
options. You can just do anything that you can go on. Man,
9:37
the same one I have on my desk. Actually it's the exact same one.
9:43
And maybe another interesting point is that on the first day he used BCI
9:50
control, he was able to break the previous world record for curser control just
9:54
by using the brain and recently he even doubled it and was able to outperform
10:01
about ten percent of our engineer at New Link. And you can be sure
10:05
that we are very good in this game and very week And if you want
10:09
to check out how well how well you can do it, you can do
10:11
it on our website. And it's very addictive games. Yeah, it's it's
10:16
a very simple game. You just have to click on the square. But
10:20
it's it's it's it's actually, even though it sounds silly, it's it's quite
10:24
a Yeah, it can be quite It sounds like it'd be quite addictive,
10:28
and it's especially if you get a low score and you think there's no way
10:31
I got it. So I mean, any anyone who wants to try this,
10:35
I recommend going to the neural link dot com website and seeing seeing if
10:37
you can beat Nolan's record, and it's that you will find that's actually quite
10:41
difficult to do so, and this is really with version one of the device
10:46
and with only a small percentage of the electrodes that are that are working.
10:50
So this is uh, this is really just the beginning, but even the
10:54
beginning is twice as good as the world record. This is important to size
11:01
the you know, the media has a habit of saying that the glasses ten
11:05
percent empty, but actually it's ninety percent full. So I think it's really
11:11
quite accomplishment of the Neuraling team to have achieved with the first patient, first
11:16
device, twice the world record for the brained computer madwidth. That's really an
11:26
astonishing, an amazingly great outcome, and it's only going to get better from
11:31
here. So the potential is to ultimately get I think to megabit level.
11:37
So that's that's partly the long term goal of improving the bandwidth of the brain
11:43
computer interface. If you think about, like how low the bandwidth normally is
11:48
between a human and a device, it's the average bandwidth is extremely low.
11:52
It's it's i say, less than one bit per second over the course of
11:58
a day. If there are eighty six thousand, four undred seconds in a
12:00
day, you're outputting less than that number of bets to any any given device
12:05
except half very rare circumstances. So this is it's actually quite important for AI.
12:18
You know, basic for human AI sembiosis is just being able to communicate
12:26
at a speed. Yeah, I could follow. So yeah, just to
12:35
emphasize again, he's performing at this extremely high level with about fifteen percent of
12:41
his channels functional, and so we want to mitigate any of the problems that
12:48
led to that situation. So, you know, the brain is a fascinating
12:52
organ. Share with you some of the secrets about the brain. During any
12:56
typical brain surgery, a small amount of air is introduced into the skull.
13:03
That's because neurosurgeons like to have as much room as possible around the brain,
13:05
and so there's this little known control mechanism of allowing the CO two concentration in
13:16
the blood to rise a bit, which allows the brain to either expand or
13:20
contract depending on where you target that CO two. But typically neurosurgeons will have
13:26
the brain shrink by lowering CO two. What we're going to do in our
13:31
future surgeries is keep the CO two concentration actually quite normal, maybe even slightly
13:37
elevated, and I'll allow the brain to stay its normal size and shape during
13:41
surgery. That should eliminate this air pocket that we saw in the first participant.
13:46
That air pocket we think may have contributed to eating up some of the
13:48
thread slack as the air bubble migrated to be under the implant push the brain
13:54
away from the implant, and so that's easy enough to fix. Another consideration
14:03
that we want to focus on for our upcoming participants is that the brain.
14:09
Think of it like a really complex folded onion. It's layer upon layer of
14:15
sheets of neurons all over the surface of the brain, folded into this odd
14:20
looking shape. The folds of the brain traveled down deep into the brain and
14:26
along with it go those onion layers of neurons. And if we insert very
14:31
close to one of the folds, where there may be very useful information encoded
14:37
in neurons, we may end up traveling with our threads parallel to some of
14:43
the layers of neurons that were most interested in avoiding them entirely. To avoid
14:46
that possibility, we're going to insert in our future participants more close to the
14:52
middle of the apex of the folds, ensuring that we're crossing the layers of
14:58
interest layer five cortex. I also think that it's important to highlight here those
15:05
tiny wires that Elon mentioned, their their fraction of a human hair. They're
15:09
very flexible intentionally so because you know, brain is constantly moving and you want
15:13
the electros to be moving with the brain causing less of the scarring. And
15:18
it's actually impossible for human eurosurgeon, however talented matthew Is to actually maneuver them
15:26
by so we have a surgical robot that we built that can actually precisely target
15:30
them in any three dimensional space x y as well as see with micronic level
15:33
precision while avoiding bascoature, so that you don't disrupt and cause immune response from
15:39
happening. So we actually have the technology to be able to place them exactly
15:43
where we want in Yeah, it was truly amazing to see the surface of
15:48
the brain after the robot had inserted all the electrodes on the first participant without
15:54
a drop of blood in sight. It's really quite an achievement. Yes,
16:00
probably most people don't realize is that the brain appears to be sort of somewhat
16:03
undifferentiated. So if you look at the cortex, it looks like a whole
16:07
bunch of folds that where you know, maybe like that it's it's it's not
16:12
obviously just looking at to take a picture of the brain that uh that that
16:15
it's the brain is highly differentiated. That there's you pretty much know exactly where
16:22
the part of the brain is that controls your right hand and your left hand
16:26
and your leg and that that kind of thing or or vision. It's it's
16:30
actually quite precisely located. It's not some people like might might think look the
16:37
brain like, oh, could be it could be anywhere, but actually we
16:41
it it's it's your brain is is highly differentiated, even though it doesn't look
16:45
it. It's. Yeah, do you want to describe how we actually were,
16:51
like how we identify word all the queen? So we can we can
16:55
put a patient that is considering this implant into an f MRI. So a
17:00
functional magnetic resonance imaging machine and ask them to imagine hand movements that, you
17:06
know, because of the spinal cord injury, don't happen, But just imagining
17:10
those hand movements causes these areas of the brain to light up in the fMRI
17:15
scanner. So we have a pretty good idea based in fact, for each
17:21
individual participant, which part of their brain is going to, you know,
17:26
respond to imagined movements of the hand, and so we can map those imagined
17:30
movements much as we all do when moving a mouse to controlling a cursor on
17:34
a screen, even without the use of a mouse. Yeah, but I
17:40
think it's kind of an important point that, like, it's not like the
17:44
part of your brain that controls your hand might be anywhere in the cortex.
17:47
It's this is not the case. It's going to be in a very specific
17:48
region, and it's going to be extremely common across people. Precision is keyt
17:53
Yeah, so the left handed right handed, to my mind, like,
18:00
if you're right handed, you want the fights on the left side, the
18:03
lato side to it. Yeah, the left side of your brain controls right
18:07
side of Another of the risk mitigations we're looking at in the future is that
18:15
you know, the implant has a certain size. The depth of the bottom
18:18
of the implant is actually thinner than the average human skull, and so what
18:25
we want to be able to do is control the size of the gap under
18:29
the implant, give the threads that travel from the implant into the brain as
18:34
much slack as possible. We didn't do this in the first participant because we
18:41
didn't want to manipulate any of their tissue that we didn't absolutely have to.
18:45
In upcoming implants, our plan is to sort of sculpt the surface of the
18:51
skull very intentionally to minimize the gap under the implants such that the bottom of
18:56
the implant travels perfectly flush with the normal contour of the inner side of the
19:02
skull. That will put the implant closer to the brain, will eliminate some
19:07
of the tension on the threads. We think it will reduce some of the
19:11
tendency of threads to retract. Right, And we actually built a tool to
19:15
do right. Yeah, this is actually this is a very important detail.
19:21
You really want the the inner contur with the skull to be blush, so
19:26
that so the implant there's there's no the brain doesn't want to pucker up into
19:30
the into the gap. That's really quite a big deal. So like like
19:37
minimize the air procket and the implant being flushed with the inside contour of the
19:42
skull are two very important improvements. The additional benefit here is that, uh,
19:48
you know, you do see some amount of stick up what we call
19:51
stick ups, so you might have bumped in the head, but this actually
19:55
eliminates it even further. Yeah, yeah, I mean it's like, really
19:57
our goal is that that if you run your hand over the top of the
20:03
skull, you don't feel any momp, you don't feel any device, and
20:07
that even if someone was bold, you want to really even notice it.
20:11
And then from the inner inner contra of the skull that the brain from physical
20:15
standpoint doesn't really notice that there's a divot in the skull because there's no dib
20:21
it. Okay. Another aspect of the human brain that you know obviously differs
20:26
from any of the animals that we tested in is that the human brain is
20:30
a lot bigger. And so you may not realize that that means the human
20:33
brain moves quite a bit more than any of these other smaller brained creatures.
20:41
And so when we open the skull. We see the brain travel toward and
20:48
away from the robot about three millimeters in total as the heart beats and the
20:52
breathing takes place, and so that movement, you know, it adds a
20:56
small challenge for the robot in precisely choosing a depth to insert each thread.
21:03
It's not an enormous challenge, and we've already upgraded the robots capabilities to be
21:07
able to even more precisely target depth in even a very rapidly moving brain with
21:15
a high amplitude of movement. You may think the most obvious mitigation for threads
21:22
that pulled out of the brain is to insert them deeper. We think so
21:25
too, and so we're going to broaden the range of depths at which we
21:30
insert threads. So for the very first participant, we had an enormous amount
21:37
of data from our animal work, and we had very highly optimized our insertion
21:41
depth to maximize the crossing of the layers of interest in the cortex with the
21:48
electrodes that we're recording from. Now that we know retraction is a possibility,
21:55
we're going to insert at a variety of depths that even in several cases of
22:00
different amounts of retracting threads. We're going to have electrodes at the proper depth
22:06
and with the deepest threads be able to track how much retraction has occurred across
22:11
the surface of the brain from from each thread. And so we're going to
22:15
you know, both have more threads in the right layer and have better data
22:21
on how much retraction has occurred. If you're a pc I nerd, you
22:23
might know that being able to control individual z the depth or thread it's not
22:27
something that most neural interface devices offer. Most neural inface devices are kind of
22:33
the static fixed Rigi array that you push in and all the letters are on
22:37
depth. Right. To be able to do this is actually pretty pretty novel
22:40
part of the robots. Yeah, the historical approach is to actually pound in
22:45
a sort of bed of nails with an air hammer into the plant. Looks
22:47
it looks crazy that that is, yeah, just with a with a aumount
22:52
of camera. That's that's it sounds come somewhat by byric. This is not
22:56
what we do, but this is the wat's been done before. Is that's
23:00
literally just hammering and what looks like a better nails with the brain actually works.
23:04
It's astonishing that it actually some people like DBS props you're just sticking in
23:11
by hand. Those are several several orders of magnitude more volume of brain tissue
23:18
that you're destroying compared to what we're doing. But that deep brain simulation stuff
23:21
does actually work and it actually helps people a lot. Yeah, yeah,
23:25
that's a great point. Yeah, I mean I think we'll be able to
23:29
do a much more finessed version of that down the road. So, I
23:36
mean it's really difficult. Like the neuraling device is something that really absolutely minimizes
23:41
damage to the brain, absolutely minimizes the load on the patient, and the
23:45
goal is to allow someone to live a completely normal life. They that you
23:49
won't even notice that someone even has the device. So, like I said,
23:56
restoring the ability to control your computer and phone, that's a slapathy.
24:00
And then next device being able to allow people to see that could not see
24:04
before. In fact you could, you could allow people to see kind of
24:07
like Jordan of the Forge and Star Trek in any whatever infrared, ultraviolet radar.
24:17
So so, so I think another way of saying it is that we
24:22
want to give people superpowers. So it's not just that we're restoring your prior
24:27
brain functionality, but that you actually have functionality far greater than a normal human.
24:33
That's a super big deal. And I also think you know, oftentimes
24:37
the questions that we get a lot is why do you have to actually go
24:38
into the brain? What if you place it on the surface or all sides
24:42
of school Basically the long story short the physics of how it works. You
24:47
really need to get the sensors, which are these facing in the brain next
24:52
to the source, which neuron as close to it as Otherwise what you get
24:56
is you get a population respault and not be able to kind of do the
25:00
level of controls that we believe. Yeah, I mean maybe a good sort
25:06
of analogy would be like if you're trying to understand what goes on in a
25:10
factory, you kind of need to go into the factory. You can't just
25:11
put us death scope on the wall and try to figure ot what's going like
25:15
anything on the outside of the Trying to read things from the outside is like
25:19
putting a statoscope on the wall of a factory trying to understand what's going in
25:22
the factory. It's not going to be effective. You've got to be threads.
25:29
I got to be in there. So but I just want to be
25:33
emphasized again, like the goal is to give people superpowers, not not just
25:37
to restore prior functionality. So I think it's very exciting and I think that
25:44
should give hope to a lot of people in the world that the future is
25:47
going to be exciting and inspiring and the technology is going to give them superpowers.
25:53
I mean, that's that's amazing. Yeah, and could can you multitask
25:57
with it? Yeah? In fact, if you look at Nolan's streaming and
26:00
you can just check out Nolan's streams on on the X platform, he's multitasking
26:07
all the time. So he's playing video games while talking and uh podcast the
26:11
podcast. Yeah, yeah exactly. So it's truly just like if you're using
26:18
your hands and you can be, you know, playing a video game while
26:22
talking. So I mean, don't take a word for it. They just
26:22
go watch. Yeah, he's out there on the internet doing the thing.
26:26
Yeah exactly. So can you do keyboard shortcuts or is just a mouse?
26:32
Yeah, that's actually we're working on right now. Oh sure, so Courtley
26:36
is walking the mouse, but we are also exploring more dimensions out from the
26:40
new activity multiple clicks, so shortcuts or able to do more games games with
26:51
an ex boot controller. But also in the future we'll expand. We plan
26:55
to expand to the code text, not just the mouse control, but also
26:57
allow we'll put the spend too. I put much fast to it. Yeah,
27:04
actually, so maybe going back to the discussion of third attraction, you
27:07
know, one of the very exciting parts to me about the story is that
27:11
we're able to do so much with fifteen percent of channels. You have more
27:14
channels. What that actually offers you is not just faster and mouse control,
27:17
because in the motor cortext neurons don't all represent the same thing. So if
27:21
you're trying to understand, like you know, what an individual finger is trying
27:23
to do, you might or might not have an electro next to it.
27:26
And the more channels you have in the brain, in the higher likelihood you
27:29
have representation or decodability of all fingers on the hand. And so if you're
27:33
trying to do something like I'll put texts at a fast rate's something that matters
27:37
a lot for people who are completely locked in, who cannot speak at all,
27:40
who are trying to, you know, just say I love you to
27:41
a love one in their family or ask for a glass of water or scratch
27:45
or whatever, you know, being able to type it a faster it's extremely
27:47
important. And the more fingers you have access to, higher probability you can
27:51
do that official And so yeah, you know, I'm super excited about how
27:55
high the ceiling is. Can we can get to as we solve this dead
27:59
attraction too. Yeah, I mean we're currently at approximately ten tenus per second
28:03
p great, but ultimately we want to get to megabit and I think,
28:08
say ultimately the whole brain interface, I think, you know, many years
28:14
from now. I think gigabit level as possible, So that's pretty soonishing now,
28:22
you know, with there is still version of one of our device.
28:25
As we mentioned, it's version one with only fifteen percent of the threads working.
28:30
The current device has sixty four threads with sixteen electrodes on each thread.
28:34
Our next device has one hundred and twenty eight threads with eight electrodes per thread.
28:42
Because as we get more confident about how we're exactly to place the electrode
28:48
the thread, you need fewer electrodes per thread. So we can essentially with
28:52
the current device without substantial changes, potentially double the band if we are accurate
28:59
with the with the placement of the threads, and then our next generation device
29:03
will have maybe even more channels. Yeah yeah, so next advice to airing
29:10
for yeah, three thousand channels. So this will just keep getting better and
29:15
better, really moving up. I think and for as magnitude in factors of
29:19
ten, basically in what kind of dwidth. So I think it won't be
29:25
it won't be all that long before someone with a neuralink device can communicate faster
29:30
than someone who is has a fully functional body. And yeah, so I
29:37
think, you know, faster than the fastest speed typist or auctioneer. The
29:41
esports tournaments are going to be like you literally won't be able to speak faster
29:45
than someone can communicate with a neural link to lepathy device. It may,
29:49
it may be a very interesting part of this. Basically, we currently connect
29:53
startup inputs to the computer to hours and keyboards. Very soon as we will
29:59
have a part of then we need to think about new ways to actually build
30:03
the interface for devices. Yeah no, that's that's a good point because the
30:08
you can't input devices are centered around human hands. So it's like we've got
30:15
these you know, little meat sticks that we move and there's a certain rate
30:21
at which you can move your little move your fingers, and and so we've
30:23
got like the mouse and the keyboard and or the doystick control, you know,
30:26
like Xbox control or something like that. But you really don't need that.
30:30
You can actually you don't. You don't you don't need since you're no
30:34
longer if you're if you're not trying to use your hands, you don't you
30:37
actually don't need those conventional uh control mechanisms. And so that's why I like,
30:45
ultimately I think you'll be able to do conceptual it telepathy, like where
30:52
you can communicate entire concepts uncompressed to someone else within your link or to the
30:57
computer. Even today, we have some problems here. We're like, you
31:02
know, if you don't feel the mouse clicking under your finger, how do
31:03
you know it actually happened, Because you know, you're you're seeing it on
31:07
the screen, but you don't actually feel the mouse. You don't have the
31:08
appropriate acceptive feedback of you know, the keys under your fingertips or the trackpad
31:11
on your So there's all sorts of interesting UX challenges to actually give the user
31:17
some sense of what their decoder is actually doing, or what the earth is
31:19
actually doing they're so wireless. Yeah, it's Bluetooth, just a bluetoth connection,
31:33
just like how your normal Apple mouse or like Apple Magic keyboard connects to
31:37
your computer, same exact thing in fact. In yeah, we can basically
31:41
have this exposed as an HID interface if we HID is just the name of
31:45
the protocol for like sending bits from a mouse into a computer. Yeah,
31:48
I can plug into basically anything. Yeah. Yeah, I mean I think
31:52
we chose that interface because it's ubiquitous. Yeah, basically any devices are have
31:57
Bluetooth capabilities. Our long term goal is to actually have our own protocol that
32:02
is safe and secure, but for now chosen it for interoperability. So the
32:08
question is kind of neuralink to prepare paralysis in the long term. You know,
32:14
we can't do that right now. We have done sort of preliminary work
32:17
planting a second neuralink in the spinal cord, and we can restore naturalistic looking
32:25
hand and leg movements in animal models, but this isn't something that is you
32:32
know, don't don't hold your breath waiting for it. It's going to be
32:35
a while. We've got a lot of work to do. Yes, there's
32:38
no reason in theory that we can't repair paralysis. Yeah, I mean essentially
32:46
to to I mean there's there's no there's no physics barrier to fully solving paralysis.
32:52
That is pers a way to say that you've got signals coming from your
32:58
motor cortex, uh that if they are transferred past the point where the neves
33:06
are damaged. Essentially just it's basically a communications bridge to bridge the communications from
33:10
the motor cortex past the point in the necrospine where the nose are damaged.
33:17
And you should like it is physics. It is possible from a physics standpoint
33:22
to restore full body functionality. For a physic standpoint, it's a very hard
33:25
technical problem, but there is nothing that it prevents it happening from a physics
33:30
standpoint. So in terms of next phases of roll out, well, we
33:36
really want to make sure that we make as much progress as possible between each
33:39
neuralink patient. So this is we're only just moving now to our second neuralink
33:45
patient. But we hope to have you know, if things go well high
33:51
single digits this year, and I don't know maybe if this is so somewhat
33:57
dependent on regulatory approval and how how much technical progress we make but within a
34:01
few years, hopefully thousands. Yeah, And I think one thing that is
34:07
important to highlight is that, you know, it's not that we built only
34:10
one device and one surgery. We've built bunters of surgery. We've built thousands
34:15
and thousands of devices even for just the ability to unearth any sort of low
34:20
frequency failure mode. So we have already been investing very heavily in infrastructure to
34:24
be able to scale this thing the device manufacturing side as well as on the
34:29
surgery side of things. We want to be able to help as many people
34:31
as quickly as Yeah, we go through obviously the appropriate hurdles that are regulatory
34:37
challenges. Yeah, and the device implantation really needs to become almost entirely,
34:44
if not entirely automatic, in the same way that's a laser eye surgery is
34:47
done. You know, you don't have an optomologist with a laser cutter by
34:52
hand, that that would be crazy, But the optomologist oversees the a laser
34:59
machine and make sure that the settings are correct, and then the machine does
35:02
everything and restores your eyesight. That's really remarkable how many people have had their
35:07
eyesight restored with laser And I think there's another one called smile. It's they
35:12
keep making it better. We need to have something similar for a neuralin implantation,
35:15
so that you basically sit down and whatever the whatever kind of upgrades or
35:22
you know, brain fixes are needed. That's that's reviewed by medical expert.
35:29
Obviously, we want to make sure that that is reviewed correctly, but it
35:32
really needs to be automatic. So you sit down and within ten minutes you
35:37
have a neural device installed, very very fast. I mean it's very sort
35:43
of cyberpunk, you know diosex if you played those games and will annually to
35:47
interface with adult devices like wheelchair. It's a great question. We're currently focusing
35:52
on turning computers unlock independence in the virtual world. Of course, our plan
35:59
is as you mentioned, little body and wheelchair to knock independence. Physical world
36:04
is of course at additional risk if you make computer yeahs to that, but
36:09
we're welcome with the FDA to allow us to what seems like if the wheelchair
36:15
has an app, well the wheelchair just needs to have an airface. So
36:21
if wheelchair has a Bluetooth interface, you could just Bluetooth interface to the wheelchair
36:25
and that's probably something we should do. We're working on pretty soon paperwork safely.
36:34
You don't, Well, I think, well, we can limit the
36:37
speed so it doesn't go careating an authentic disaster, but you know, so
36:43
just it go slowly at first. But yeah, so being able to sort
36:47
of it. Really, the New York device just should work generally for anything
36:52
that's got a Bluetooth theraterface, including potentially an optimism. Yes, yeah,
37:01
yes, you could communicate with optimists. Yep, absolutely optimist. Well we
37:09
also also be able to talk to optimists, but like, but you could
37:15
just yeah, instead of talking, just you could just be in it directly.
37:17
Or if if someone has lost the use of speech, then then they
37:21
can still communicate to an optimist. They can mumunicate telepathically to optimists or Bluetooth
37:29
and and and so even if someone has completely lost the ability to speak,
37:35
they could still control optimists or their computer or phone. We also like,
37:38
if you have an optimist and you have a nerve, and you can just
37:42
mapped the insignal to control of the physical armor, and that's a very meaningful
37:45
thing, like if you're no Folks that have spon ecoan injury, one of
37:47
the biggest requests to be able to scratch yourself. It's something that quite annoying
37:52
actually, And if you have a scratch on your face, you can't fall
37:55
asleep until you scratch it. You know, it's very convenient to be able
37:59
to move something physically towards you too. Similar things like eating food. You
38:02
know, if you need somebody to feed you very hard to have dinner with
38:06
friends in a way that is sort of a normal social experience. And so
38:09
if you can feed yourself, pick up a fork and actually eat a pizza
38:14
chicken on your own, yeah, that's a big deal. It prevents and
38:17
saves a lot of interaction with caretakers and other people in your life that you
38:22
rely on to take care of. It really increases you. I think.
38:27
An exciting possibility long term also is to say if you take parts of the
38:30
optimous Optimius humanoid robot and you combine that with a neural link. Let's say
38:34
somebody has lost their arms or legs, Well, we could actually attach an
38:38
optimus armor optimist legs and do a neuralink implant so that the motor commands from
38:45
your brain that goat would goat to your biological arms, now go to your
38:50
robot arms or robot legs and again, you'd have basically cybernetic superpowers. Actually,
38:55
so the linksy from the neuralink to your hand would probably be faster than
39:00
it is to go to your physical hands. So you can imagine, like
39:02
if you're a piano player or anything that requires extremely fast hand movements, that
39:07
you could actually have a pretty imbalanced right hand robotic arm control versus left hand
39:12
physical hand control. That's one of them. Yeah, Like I said,
39:15
it's just kind of a cyberpunk dio sex in the future where you have cybernetic
39:20
upgrades that are actually better than your biological limbs. And it's certainly that we'll
39:31
have a much you know as particularly as we expand to a large number of
39:38
customers or patients for neuralink, the understanding of the brain will improve dramatically because
39:45
really there isn't a fine, very fine grained understanding of the brain today because
39:52
it's just the sensors aren't good enough. You've got fMRI, which is pretty
39:55
good, but it's still not as good as actually having I bound with like
40:00
TOIDs in the brain. Yeah, I think this is underappreciated as a research
40:02
tool to move that whole effort forward of really knowing you know what the physical
40:08
substance of human thought is we don't know to the to the degree that we
40:15
need to. Neuralink is actually a very powerful research tool. Yeah, I
40:21
mean we I think we can ultimately understand and and fix it's quite severe psychosis
40:29
or like if if somebody's got like the if somebody's got like a like a
40:35
delusion that they have a chip in their brain. I was if you're going
40:38
to mention that one. We just want to be clear if there's only one
40:43
person with a new link hip in their brain. So for people out there
40:46
who think we've put a chip in their brain, would like to assure you
40:50
or what it's worth, you probably won't believe us, but we did not
40:52
put each op in your brain. Okay, it's so there's actually a remarkable
41:00
number of people who think we have put a trip in their brain, but
41:02
we have not. But in the future, if you would like us to
41:07
put a trip in your brain, which will perhaps help with the issue of
41:10
thinking that you have a tip in your brain, then we will be able
41:14
to do so. So there are people that have severia, schizophrenia, they've
41:19
got basically things that their brain is malfunctioning in some way and this is actually
41:25
due to really like physical circuitry issues. You can think of the brain as
41:30
like, really it's a biological computer, and if if some of the circuits
41:36
are crussed, it's gonna you know, it's going to crash, it's or
41:39
it's gonna have issues that it's not work. But with a neuralic device,
41:45
we can fix those issues. And you know, give someone who I think
41:52
has to say severe scisophraina or psychosis of some kind, allow them to live
41:57
a normal life. I think that is one of the likely things in the
42:00
future. So yeah, I mean, yeah, you can certainly imagine,
42:12
like I'm sure people have like parents, grandparents who have you have memory that's
42:20
not working as well as it used to be. Sometimes they forget who who
42:23
their grandchildren are or what day it is, and this is something that in
42:28
your like device could help fix. I mean, that's actually one of the
42:32
personal reasons in many way like forms of you're literally losing part of your right
42:44
then, yeah, just a very verat Yeah, and it's really just it's
42:52
a glitch in the biological computer that is a fixable glitch, like like it's
42:58
a short circuit. Essentially, how does the device charge and how long does
43:02
the charge last? Yeah, so the current version that Nolan has lasts but
43:07
four to five hours on a charge, and it takes about forty five minutes
43:10
to charge. The thing we've learned from Nolan is that that's actually one of
43:14
the main limitters for him using it more. It's actually pretty hard to use
43:16
a product more than like seventy hours a week. But that's about what he
43:21
has used it for it in some weeks. Yeah, seventy hours a week.
43:23
Yeah, I mean just for context, like you sleep roughly eight hours
43:25
of night, so that's you know, we're doing better than the bed,
43:29
Like the bed is fifty six hours a week he used, So seventy hours
43:32
a week of uses. I challenge you to think about products you've actually used
43:36
for that duration. That's again, some of these points are worth like empasizing
43:39
again, like the that's Nolan, our first on yourlink recipient has used the
43:45
uralin device for seventy hours in a week, which is incredible. You probably
43:50
won't enjoy that I'm sharing his computer use publicly, but I mean I assure
43:55
you for productive things only. But actually, so one of things we've learned
44:00
is that in the next version of the device, who really need to like
44:01
double or increase that out of your life, And so I think DJ the
44:05
next version is going to be double actually actually do without without increasing the charge
44:09
correct in charging time to double, the better your life, meaning you should
44:13
get roughly eight hours of use, and the goal is to actually get to
44:15
all they use, so you can just charge, you know, maybe in
44:19
your sleep, right sleeping pillow, exactly as soon as you've got like sixteen
44:22
hours of usage, then you basically have twenty four hours of usage because you
44:25
can charge while you're sleeping. One other thing that's important I think to call
44:29
out here is if you're paralyzed, you can't you put the charger over your
44:31
head yourself, And so it's important to think about, like it's not just
44:35
a duration of better use, but also can you recharge it yourself independently.
44:38
We spend a lot of time thinking about how to make them peaceable because then
44:42
that means that you can and this is what no one does. You can
44:44
use the device, charge it, use the device, charge it, use
44:46
the device without needing anybody to come in and sort of help you with that,
44:50
which is a big deal. If you're trying to play save until five
44:52
am at night when your family's And the way in which he does that is
44:57
that there is a charge of coil. That's a bigger you know about this,
45:00
and we actually put it in the sleeve of a little beanie or a
45:06
beanie and then he wears it and then says with the voice command and charge
45:09
charge your energize. That's the one you like Howard writing work. Uh so
45:22
so uh yeah, the current device that Nolan has is reading, so it's
45:30
trying to read his essentially like worst movement from from one one hand. That's
45:35
Alost you know, with Worth pointing out like in the future, like we're
45:38
pretty cool to give Nolan a second implant that would allow the other hand to
45:44
be used and also have higher uh obviously higher active electro accounts, so then
45:49
you could play too, essentially play games two handed, because that's not only
45:51
how you play games and uh but then with with writing, it's really just,
45:57
uh, it's an electrical impulse. Instead of like reading electrical impulses from
46:00
the neurons, you issue an electrical impulse, which is obviously critical for vision.
46:05
So vision is writing, which is just triggering electrical impulse in the vision
46:10
part of the brain and that like activates a pixel. So we actually do
46:16
have this working in monkeys, and we have we've had it working with monkeys
46:20
for a while now, where you can sort of flash a pixel and then
46:22
you watch where the monkey. Obviously the monkey is like whatever. Mike's a
46:28
little surprised to see, like, hey, there's a flash here and a
46:30
flash here, but it's it's used to it after a while. But it
46:34
just you can see that that the pixel is in the right location because the
46:37
monkey's eyes will will dart to that location. It's not on the screen,
46:40
like there's no pixel on the screen, there's no picks on the screen in
46:43
the brain. Yeah, just like just verify that that there triggering pixel in
46:46
the right part of the brain. So you know, the initial resolution for
46:52
vision will be relatively low, you know, sort of a Tario graphics type
46:55
of thing, but over time it could potentially be better than normal vision.
47:00
And then I guess in terms of some additional applications for where writing to the
47:06
brain can be useful, what are applications. As Bliss mentioned, there is
47:12
feedback, there's appropriate receptive feedback, there's a tel feedback, especially on robot
47:15
army. You're trying to grasp ap you need to know about it to some
47:20
egg. It's very much a delicate balance of not just initiating the movement but
47:25
getting the feedback and controlling it accordingly. So there is some metasensory cortex that's
47:30
right adjacent to modern cort text it could be better motor movements. So any
47:46
changes in neural growth after the device is inserted, we don't see any any
47:51
signs of neural damage, but I and I guess we we have seen some
47:57
rebound on some the electrodes, right correct. And then also I mean,
48:04
I guess, I guess you know. Brain is very plastic, so it's
48:07
not that plastic. Well, it does diminished quite a bit after and why
48:15
through throughout childhood, especially when you get to about twenty five brain really done
48:23
cooking. Yeah, but that there there are there is a little bit of
48:30
damage done with each insertion, but it's a minuscule amount compared to anything else
48:36
out there, and so it's an easy amount of damage to recover from.
48:43
And it's really only detectable on cutting pieces of the brain after after the animal
48:51
is no longer live and looking at them under a microscope. You can't really
48:54
tell during life that there's been any brain. Another way to interpret this question,
49:00
have the fund need changes in neural growth after device is inserted. One
49:04
way to interpret that is like the user learning how to use the device.
49:06
I think on that side of things, there's been tremendous progress. He's put
49:09
in hundreds of hours trying to figure out the best way to use this device
49:14
because he really thinks that, you know, if he can figure this out
49:15
and you can help share this knowledge. I mean, he's like on Friday
49:19
night at eight pm, you know, he's starting a session of like,
49:22
you know, figuring out himself how to how to push his own performance to
49:24
the next level. And that's really a unique learning process because there's not many
49:29
people in the world that had the experience of moving something and so there's a
49:31
lot of nuance to like, Okay, how exactly should I imagine or attempt
49:36
to move my risk to get the thing too. Yeah, he's really dialed
49:39
that into it. Also, just a sheer number of hours that he's us
49:44
He's saying even in the past six months, right, Yeah, in many
49:47
ways, like he's using it in his travel, in his plane ride.
49:51
Effectively, PCI has left alive. Yeah, yeah, yeah, I mean
49:53
one of the questions is how close we can burning thoughts out of text.
49:58
I mean, right, right now, it's more about and curse from the
50:00
screen on a virtual keyboard. But long term you should be able to really
50:06
master about entire words faster than anyone could possibly time. I'm able to type
50:10
a low world today directly, but we're still in the early days making that
50:15
a pause experience. I mean, the other things that we're looking at is
50:20
silent. At the end of the day, it is a movement of and
50:23
into rights. Yeah. True, was the brain trying to naturally push the
50:30
threads out. I mean this is sort of a universal feature of any implant
50:34
in the body. The body tries to reject it, and the goal of
50:37
the surgeons and the technology team is to fight that. And so with artificial
50:43
hips and with screws in the spine, we've done a really good job of
50:46
finding biocompatible materials and techniques to fix those implants in the body. I mean,
50:53
past a certain age, it's getting hard to find someone without some kind
50:57
of implant, you know, the uh, some kind of screws in their
51:00
spine. And so we've got this problem pretty well solved. So to answer
51:08
your question. Yes, the body is trying to get rid of any implant,
51:12
but we can sure that basically canned. It's also worth highlighting that the
51:17
threads have not actually moved in the past five months. There's there's some still
51:23
minor movements in terms of like some maybe maybe getting pushed in a little bit,
51:27
pushed out a little bit, but it's it's more or less very stable
51:30
and been stable for and the reason for that is, you know, once
51:34
you once you do a brain surgery, it takes some time for the tissues
51:38
to come in and then and then our tissue or the neo membrane to actually
51:43
come in and then anchor the threads in place. And once that happens it
51:45
has been stable and seen much movement. That's where the world record performance starts
51:52
to come in. Yeah, that was a couple of weeks ago. Yeah,
51:57
the threads like it. It is important that the threads be extremely tiny.
52:00
If they're extremely tiny, then the brain does not The smaller they are,
52:04
the less like the brain is to react to them. So that's why
52:07
you want the thread to be extremely tiny and also to minimize any damage to
52:12
your unts. So on that note, we do plan to actually share some
52:17
of the tissue response in detail in some of the later upcoming updates. Yeah,
52:25
it is quite a challenging. It's challenging on many fronts to do something
52:30
like this because you're trying to read and read and write electrical signals, but
52:36
you need to have the threads themselves need to be electrically isolated and not subject
52:45
to corrosion in the body. So like the you know, just middle by
52:50
itself is somewhat supress subject to corrosion or being attacked. So it's it's it's
52:55
good. Like in terms of the pre various coatings and things. To actually
52:59
make this electrode work while not actually eroding its performance over time is very difficult.
53:05
Human bodies are very very harsh, very harsh. It's a bag of
53:07
salt water with bad sensors, that's elevated temperature that as well regulated. I
53:13
mean, I'm sure people have experienced dropping their electronic devices in a seawater and
53:16
in an instant. Yeah, yeah, so we'll we're better sort of wrap
53:22
this up soon. And if there's like a few a few last questions,
53:28
I guess, so a good question, So what about upgrades? So yeah,
53:32
we we do think it's going to be important to be able to upgrade
53:36
their device over time, just like you wouldn't want like an iPhone one stuck
53:39
in your brain forever. You know, if you've got an iPhone fifteen,
53:45
you probably want the iPhone fifteen, not the iPhone one. So I think
53:50
people over time will be able to upgrade their neural links. So we'll take
53:54
the neural link device out and put a new one in. And we have
54:00
done this with some of our animals, and they're actually in one case we
54:06
did with we upgradeed device three times and and uh with a pig. We
54:10
did with a monkey as well. Yes, and he's he's doing fine.
54:17
Point hit his I think his record with an upgrade. It still beat him
54:23
though, still beat him. Yes, this is true. Humans are top
54:27
of the species board right now. Pages like what like eight or something.
54:29
It was like eight point five bps. Okay, well that's a very high
54:32
school not trying to put Patriot down. And also to train a monkey to
54:36
do that is a full challenge on its own. We have like the best
54:37
animal care team world. Yeah, I just do want to have size.
54:40
We we we we do our absolute best to take care of the animals.
54:45
And uh, when we had like a USDA inspector come through. She said
54:51
that this was the nicest animal facility she has ever seen in her entire life.
54:55
So breakfast, the monkey orders room service. Yeah, we have monkey
55:05
room servants, which is rare fact. We're the only ones who offer monkey
55:09
room servants. So we really do everything we can to maximize well for the
55:15
animals. So all right with that, thank you everyone for tuning in.
55:20
I hope you're