Latest / Elon Musk Thinking / Elon Musk latest updates of Neuralink!!!

Transcript

Elon Musk latest updates of Neuralink!!! · Elon Musk Thinking

0:00
Speaker 1: Welcome to the Neuralink presentation. This is an update for
0:04
Speaker 1: the progress from the neuralink team. It's been an incredible
0:07
Speaker 1: amount of progress.
0:08
Speaker 2: This is.
0:11
Speaker 1: We're going to start off high level generally describing what
0:15
Speaker 1: neuralink is doing, and then we're going to have a
0:18
Speaker 1: very deep technical dive so you can actually get an
0:21
Speaker 1: understanding of what exactly we're doing at a granule level
0:24
Speaker 1: and what we can do to enhance human capabilities and
0:29
Speaker 1: ultimately build a great future for humanity. So that's a neuronspiring.
0:35
Speaker 1: It's funny thing that me talking right now is a
0:39
Speaker 1: bunch of neurons firing that then result in speech.
0:45
Speaker 2: That you here. Of course, neurons to fire in your brain.
0:51
Speaker 3: Part of.
0:53
Speaker 1: This presentation is about demystifying the brain. It is a
0:58
Speaker 1: remarkable organ. I mean we are the brain. Basically when
1:01
Speaker 1: you say you that really is uh, you're the brain.
1:06
Speaker 1: Like you can you can get a heart transplant, you
1:08
Speaker 1: can get a kidney transplant, but I don't know anyone
1:11
Speaker 1: who's gotten a brain transplant.
1:14
Speaker 2: So you are your brain.
1:19
Speaker 1: And your experiences are these neurons firing with the trillions
1:24
Speaker 1: of synapses.
1:28
Speaker 2: That somehow lead to conscious comprehension of the world.
1:32
Speaker 1: This is something that we have only begun to understand.
1:37
Speaker 1: We're really just barely at the beginning of understanding of
1:39
Speaker 1: what is the nature of consciousness. And I've thought a
1:43
Speaker 1: lot about what what is consciousness?
1:45
Speaker 2: What is it? Where does consciousness arise?
1:51
Speaker 1: Because if you start at the beginning of the universe,
1:54
Speaker 1: assuming physics is true, the physic current, the standard model
1:57
Speaker 1: of physics is true, then you have this.
2:01
Speaker 2: Big bang, the matter condensing.
2:05
Speaker 1: Into stars, those stars exploding, a lot of the atoms
2:10
Speaker 1: that are in your body. Right now, we're once at
2:12
Speaker 1: the center of stars, those stars exploded, recondensed. Fast forward
2:18
Speaker 1: thirteen point eight billion years, and here we are. And
2:24
Speaker 1: somewhere along that very long journey to us, at least
2:31
Speaker 1: consciousness arose or the molecules started talking to each other.
2:38
Speaker 1: And it begs the question of what is consciousness? Is
2:47
Speaker 1: is everything conscious? Maybe it's hard to say where along
2:50
Speaker 1: that line that there's no sort of discrete point where
2:54
Speaker 1: consciousness didn't exist, and then suddenly it does exist. It
3:00
Speaker 1: seems to be maybe you have a condensation of matter
3:04
Speaker 1: that has a density of like we don't know what
3:09
Speaker 1: really the real answer is, we don't know what consciousness is.
3:13
Speaker 1: But with the neural link and the progress of the
3:17
Speaker 1: company's making, will begin to understand a lot more about
3:20
Speaker 1: consciousness and what does it mean to to be. Along
3:30
Speaker 1: the way, we're we're going to solve a lot of
3:34
Speaker 1: a lot of brain issues where the brains get uh
3:40
Speaker 1: injury or damaged in some way or didn't develop in
3:43
Speaker 1: quite the right way. But there's yeah, there's there's a
3:47
Speaker 1: lot of brain in spine injuries that will so along
3:51
Speaker 1: the way. And I do want to have size that
3:53
Speaker 1: this is all going to happen quite slowly, meaning you'll
3:56
Speaker 1: you'll see it coming. Sometimes people think that suddenly there
3:59
Speaker 1: will be vast numbers of neural links all over the place.
4:04
Speaker 1: This is not going to be sudden. You'll be able
4:06
Speaker 1: to watch it happen, you know, over the course of
4:10
Speaker 1: several years. And we go through exhaustive regulatory approvals. So
4:17
Speaker 1: this is not something that we're just doing there by
4:20
Speaker 1: ourselves without governor oversight. We work closely with the regulators
4:26
Speaker 1: every sepular way. We're very cautious with the neural links
4:30
Speaker 1: and humans. That's the reason we're not moving faster than
4:34
Speaker 1: we are is because we're taking great care with each
4:37
Speaker 1: individual to make sure we never miss and so far
4:41
Speaker 1: we haven't, and I hope that continues into the future.
4:43
Speaker 1: Every single one of our implants and humans is working
4:47
Speaker 1: and working quite well, and you'll get to hear from
4:50
Speaker 1: some of the people that have received the implants and
4:54
Speaker 1: care it in their words. So what we're creating here
4:58
Speaker 1: with a neuralink device is a generalized input output.
5:04
Speaker 2: Technology for the brain.
5:06
Speaker 1: So it's how do you get information into or out
5:11
Speaker 1: of the brain and do so in a way that
5:16
Speaker 1: does not damage the brain or you know, of course
5:19
Speaker 1: any negative side effects. So it's a very hard problem
5:23
Speaker 1: and generally the reactions I've seen to this range from
5:28
Speaker 1: it's impossible to it's already been done before those people
5:33
Speaker 1: should meet. Actually, the reality is that there actually have
5:39
Speaker 1: been limited brain to computer interfaces for several decades on
5:47
Speaker 1: a very basic basis. Just what we're doing with neurlink
5:52
Speaker 1: is dramatically increasing the bandwidth by many or as the magnitude,
5:58
Speaker 1: so you can you can a human bandwidth output is
6:03
Speaker 1: less than one bit per second over the course of
6:05
Speaker 1: a day, so there's eighty six seconds in a day.
6:09
Speaker 1: It's very rare for a person to do more than
6:11
Speaker 1: eighty six thousand, four hundred bits of output per day.
6:15
Speaker 1: You'd have to be really talking a lot or typing
6:17
Speaker 1: all day and you might exceed that. So what we're
6:22
Speaker 1: talking about here is going from maybe one bit per
6:25
Speaker 1: second to ultimately megabits and then gigabits per second, and
6:31
Speaker 1: the ability to do conceptual consensual tlepathy. Now, the input
6:39
Speaker 1: to the rate is much higher because especially because of vision,
6:44
Speaker 1: depending upon how you counted, it might be on the
6:47
Speaker 1: order of a megabit or in the megabt range for input,
6:52
Speaker 1: primarily due to site, but even for input, we think
6:59
Speaker 1: that can be traumatic increased to the.
7:02
Speaker 2: Gig of it plus level.
7:05
Speaker 1: And a lot of the thinking that we do is
7:10
Speaker 1: which we take a concept in our mind and we
7:13
Speaker 1: compress that into a small number of symbols. So when
7:17
Speaker 1: you're trying to communicate with somebody else, you're actually trying
7:20
Speaker 1: to model their mind state and then take peraps quite
7:26
Speaker 1: a complex idea that you have, maybe even a complex
7:29
Speaker 1: image or scene or kind of mental video, and try
7:33
Speaker 1: to compress that into a few woods or a few keystrokes,
7:36
Speaker 1: and it's necessarily.
7:37
Speaker 2: Going to be very lossy.
7:39
Speaker 1: Your ability to communicate is very limited by how fast
7:43
Speaker 1: you can talk and how fast you can type, and
7:45
Speaker 1: what we're talking about is unlocking that potential to enable
7:49
Speaker 1: you to communicate, like I said, thousands, perhaps millions of
7:53
Speaker 1: times faster than is currently possible. This is an incredibly
7:58
Speaker 1: profound breakthrough. This would this would be a fundamental change
8:03
Speaker 1: to what it means to be a human. So we're
8:08
Speaker 1: starting off with reducing human suffering, so or addressing issues
8:15
Speaker 1: that people have, save if they've been in an accident,
8:17
Speaker 1: or they have some uh neural disease that's degenerative, so
8:22
Speaker 1: they're losing capability to move their body, or some some
8:26
Speaker 1: kind of injury. Essentially, so enabling the first products is
8:31
Speaker 1: called telepathy, and that enables someone who has lost the
8:37
Speaker 1: ability to command their body to be able to communicate
8:40
Speaker 1: with a computer and move the mouse and actually operate
8:43
Speaker 1: a computer with roughly the same dexterity, ultimately much more
8:47
Speaker 1: dexterity than a human with working hands. Then our next
8:55
Speaker 1: product is blind sight, which will enable those who have
8:58
Speaker 1: total loss of visioning they've lost their eyes or the
9:01
Speaker 1: optic nerve or maybe i've never seen, we're even blind
9:05
Speaker 1: from birth, to be able to see again, initially low resolution,
9:09
Speaker 1: but ultimately very high resolution, and then in multiple wavelengths,
9:13
Speaker 1: so you could be like Jordie LaForge in Star Trek
9:16
Speaker 1: and you can see in radar, you can see infrared, ultraviolet,
9:19
Speaker 1: superhuman capabilities, sebernetic enhancement essentially, and then along the way
9:28
Speaker 1: this should help us understand a lot more about consciousness,
9:32
Speaker 1: what does it mean to be a conscious creature? Well,
9:37
Speaker 1: understand vastly more about the nature of consciousness as a
9:40
Speaker 1: result of this, and then ultimately I think this helps
9:43
Speaker 1: mitigate the civilizational risk of artificial intelligence. We are actually
9:53
Speaker 1: already we're ready to sort of have three layers of thinking.
9:57
Speaker 1: There's the olympic system, which is your kind of your instincts,
10:02
Speaker 1: your cortical system, which is your higher level planning and thinking.
10:06
Speaker 1: And then the tertiary layer, which is the computers and
10:10
Speaker 1: machines that you interact with, like your phone, all the
10:13
Speaker 1: applications you use.
10:16
Speaker 2: So people actually are already a cyborg.
10:19
Speaker 1: You can maybe have an intuitive sense for this by
10:23
Speaker 1: how much you miss your phone if you leave it behind.
10:28
Speaker 1: Leaving your phone behind is like it's almost like missing
10:30
Speaker 1: limbs syndrome.
10:32
Speaker 2: Your phone is somewhat of an extension of yourself, as
10:35
Speaker 2: is your computer.
10:36
Speaker 4: So you.
10:38
Speaker 1: Already have this digital tertie layer, but the bandwidth between
10:42
Speaker 1: your cortex and your digital tertie layer is limited by
10:46
Speaker 1: speech and by and by half fast you can move
10:49
Speaker 1: your fingers, and how fast you can consume information visually.
10:53
Speaker 1: So so, but I think it's actually very important for
10:58
Speaker 1: us to address that input output boundwidth constraint in order
11:03
Speaker 1: for the collective will of humanity to match the will
11:07
Speaker 1: of artificial intelligence. That's my intuition at least. So let's see,
11:17
Speaker 1: and what this presentation is mostly about is attracting smart
11:27
Speaker 1: humans to come and work with us on this problem.
11:30
Speaker 1: So this is not a presentation to raise money or
11:34
Speaker 1: anything like that.
11:36
Speaker 2: We're actually, you know, very well funded.
11:38
Speaker 1: We have a lot of great investors, some of the
11:42
Speaker 1: smartest people in the world are invested in neur Link.
11:45
Speaker 1: But we we need smart humans to come here and
11:48
Speaker 1: help solve this problem. So with that, let's uh, let's proceed.
12:04
Speaker 5: Hey everyone, my name is DJ I'm my co founder
12:09
Speaker 5: and president of Neuralink. And as Elon mentioned, well, actually
12:13
Speaker 5: we're standing in the middle of our robust space. We
12:16
Speaker 5: have a stage set up, but you know, this is
12:19
Speaker 5: actually where some of the next generation, most advanced surgical
12:23
Speaker 5: robots are being built, so welcome to our space. It's
12:33
Speaker 5: important to highlight that this technology is not being built
12:36
Speaker 5: in the dark. This is not a secret lab where
12:39
Speaker 5: we're not sharing any of the progress. In fact, we're
12:41
Speaker 5: actually sharing you know, the progress very openly and as
12:46
Speaker 5: well as also telling you exactly what we're going to
12:48
Speaker 5: be doing, and we're hoping to progress on that as
12:52
Speaker 5: as diligently and as safely and as carefully as possible.
12:56
Speaker 5: So to start off, two years ago, when we did
13:00
Speaker 5: our previous fundraising round, we outlined this path and timeline
13:04
Speaker 5: to first human and we currently have a clinical trials
13:08
Speaker 5: in the US for a product that we call Telepathy,
13:11
Speaker 5: which allows users to control phone or computer purely with
13:15
Speaker 5: their thoughts. And you're going to see how we do
13:17
Speaker 5: this and what the impact that this has had. And
13:21
Speaker 5: not only have we launched this clinical trial, but as
13:24
Speaker 5: of today we have not just one, but seven.
13:27
Speaker 6: Participants and we have an approval.
13:36
Speaker 5: And we also have an approval to launch this trial
13:39
Speaker 5: in Canada, UK and the UAE. So I guess before
13:51
Speaker 5: we dive into what this technology is and what we built,
13:55
Speaker 5: but I wanted to quickly share a video with you
13:57
Speaker 5: guys of when our first five parts participants met each
14:01
Speaker 5: other for the first time.
14:02
Speaker 2: So here you go.
14:03
Speaker 6: All right, we have everyone together up.
14:06
Speaker 2: Guys, thanks everybody joining. Want to use of you.
14:10
Speaker 7: Yeah, I'm Nolan aka P one.
14:13
Speaker 8: My name is Alex. I am the second participant in
14:16
Speaker 8: the Neuralink study.
14:18
Speaker 2: I am Brad Smith. Let's cycle P three.
14:22
Speaker 9: My name and my G four have.
14:29
Speaker 4: Less like.
14:31
Speaker 3: Uh yeah, I'm marj I'm P five and I just yes,
14:37
Speaker 3: I'm kind of to one of the team heres. So yeah,
14:40
Speaker 3: appreciate it. No trail blazer, you know, somebody's got to
14:43
Speaker 3: get first. Man, that was you.
14:44
Speaker 2: Appreciate that.
14:45
Speaker 10: What's been your favorite thing you've been able to do
14:47
Speaker 10: with the neuralink so far.
14:49
Speaker 7: I've just had a good time being able to use
14:51
Speaker 7: it as I travel buying and drawn.
14:53
Speaker 2: A little mustache on a cat.
14:55
Speaker 7: Had a lot of fun doing that. I mean, I've
14:57
Speaker 7: just had a good time playing around with it.
14:59
Speaker 2: Oh you know what, But I.
15:00
Speaker 7: Do know what My favorite BCI feature is probably not
15:03
Speaker 7: a feature, but I just I love web Grid more
15:07
Speaker 7: than I love anything in my life. Probably I think
15:10
Speaker 7: I could play that game NonStop forever.
15:14
Speaker 8: Has to be Fusion three sixty being able to design parts,
15:18
Speaker 8: design the hat logo with the BCI.
15:22
Speaker 2: That's what's up?
15:23
Speaker 11: Pretty sweet?
15:24
Speaker 2: That's sweet?
15:25
Speaker 8: Yeah, Yeah, I have a little uh are do we know?
15:30
Speaker 8: That takes input from my quad stick converts it into
15:35
Speaker 8: a PPM signal to go to an RC truck.
15:39
Speaker 9: Cool little rock crawler.
15:42
Speaker 8: Well with the BCI, I room code to drive the
15:48
Speaker 8: plane with the quad stick.
15:51
Speaker 6: That's awesome.
15:52
Speaker 1: The best thing I like Abunerline is being able to
15:58
Speaker 1: continue to provide.
16:02
Speaker 7: For my family and continue working.
16:07
Speaker 3: I think my favorite things probably been able to turn
16:10
Speaker 3: on my TV. Yeah, like the first time in two
16:13
Speaker 3: and a half years. I was able to do that.
16:15
Speaker 3: That's pretty sweet with a right shooting the all base.
16:18
Speaker 3: That's that's sound nice. Excited to see it. BCIs got
16:21
Speaker 3: going on?
16:22
Speaker 11: They got a car?
16:23
Speaker 8: What's your shirts say?
16:25
Speaker 11: Is it?
16:25
Speaker 3: I do a thing called whatever I want.
16:39
Speaker 6: Now.
16:39
Speaker 5: One of the major figure of merits that we have
16:42
Speaker 5: is to keep track of monthly hours of independent DCI use. Effectively,
16:47
Speaker 5: are they using the BCI and not at the clinic
16:50
Speaker 5: but at their home? And what we have noticed, and
16:53
Speaker 5: this is a plot of all of the different participants,
16:56
Speaker 5: first five participants and their usage per month over the
17:00
Speaker 5: course of the last year and a half and we're
17:03
Speaker 5: averaging around fifty hours a week of usage and in
17:06
Speaker 5: some cases peak usage of more than one hundred hours
17:08
Speaker 5: a week, which is pretty much every waking moments. So
17:18
Speaker 5: I think it's been incredible to see all of our
17:20
Speaker 5: participants demonstrating greater independence through their use of PCI. Not
17:28
Speaker 5: only that, we've also accelerated our implantation cadence as we've
17:31
Speaker 5: amassed evidence of both clinical safety as well as value
17:35
Speaker 5: to our participants. So to date, we have four spinal
17:38
Speaker 5: cord injury participants as well as three ALS participants, with
17:42
Speaker 5: the last two surgeries happening within one week of each other,
17:48
Speaker 5: and we're just beginning.
17:50
Speaker 6: This is just tip of the iceberg.
17:52
Speaker 5: Our end goal is to really build a whole brand interface.
17:56
Speaker 5: And what do we mean by whole brand interface. We
17:59
Speaker 5: mean being able to listen to neurons everywhere, be able
18:03
Speaker 5: to write information to neurons anywhere, be able to have
18:07
Speaker 5: that fast data wireless transfer, to enable that high bandwidth
18:10
Speaker 5: connection from our biological brain to the external machines, and
18:15
Speaker 5: be able to do all of this with fully automated surgery,
18:19
Speaker 5: as well as enable twenty four hours of usage, and
18:25
Speaker 5: towards that goal, we're really working on three major product types.
18:29
Speaker 5: Elon mentioned earlier that our goals to build a generalized
18:32
Speaker 5: input output platform and technology to the brain so to
18:37
Speaker 5: afford the output portion of it, which is extremely slow,
18:40
Speaker 5: through our meat sticks, as Elon calls them.
18:46
Speaker 6: Meet hands that are holding the mics.
18:50
Speaker 5: We're starting out with helping people with movement disorders, either
18:53
Speaker 5: through where they lost a mind body connection, either through
18:56
Speaker 5: a spinal cor injury ALS or a stroke, be able
18:59
Speaker 5: to regain some of that digital as well as physical
19:02
Speaker 5: independence through a product that we're building called Telepathy, and
19:06
Speaker 5: this is our opportunities to build a high channel read
19:09
Speaker 5: and output device. On the input side of things, there's
19:15
Speaker 5: opportunities for us to help people that have lost the
19:17
Speaker 5: ability to see be able to regain that site again
19:21
Speaker 5: through a product that we're calling blind Site, and this
19:24
Speaker 5: is our opportunity to build high channel right capabilities and
19:29
Speaker 5: last but not least, be able to also help people
19:32
Speaker 5: that are suffering from neurological debilitating dysregulation or psychiatric conditions
19:39
Speaker 5: or neuropathic pain by inserting our electros in reaching any
19:45
Speaker 5: brain regions to be able to insert them not just
19:48
Speaker 5: on the cortical layer, but into the sulk guys as
19:51
Speaker 5: well as deeper parts of the brain, the so called
19:54
Speaker 5: limbic system, to really enable better opportunities to just regain
19:58
Speaker 5: some of that independence. Our north Star metrics is one
20:03
Speaker 5: increasing the number of neurons that we can interface with,
20:07
Speaker 5: and second to expand to many diverse areas any parts
20:11
Speaker 5: of the brain, starting with microfabrication or lithography to change
20:15
Speaker 5: the way in which we can actually increase the number
20:18
Speaker 5: of neurons that we can see from a single channel,
20:21
Speaker 5: and also doing mixed signal chip design to actually increase
20:26
Speaker 5: the physical channel counts to increase more neurons that we
20:31
Speaker 5: can interface to to sort of allow more information from
20:35
Speaker 5: the brain to the outside world.
20:37
Speaker 2: And then you.
20:38
Speaker 5: Know, everything we built from day one of the company
20:41
Speaker 5: has always been read and write capable, and with Telepathy
20:45
Speaker 5: our first product, the focus has been on the read
20:48
Speaker 5: capabilities or the output and we want to hone in
20:51
Speaker 5: on our right capability and also show that through accessing
20:55
Speaker 5: deeper regions within the visual cortex that we can actually
21:00
Speaker 5: achieve functional vision. So now just to step you through
21:15
Speaker 5: what the product evolution is going to look like in
21:16
Speaker 5: the next three years, today, what we have is one
21:20
Speaker 5: thousand electrodes in the motor cortex, the part of the
21:23
Speaker 5: small part of the brain that you see in this
21:25
Speaker 5: animation called the hand knob area that allows participants control
21:29
Speaker 5: computer cursors as well as gaming consoles.
21:33
Speaker 6: Next quarter, we're planning to implant in.
21:36
Speaker 5: The speech cortex to directly decode attentive words from brain
21:41
Speaker 5: signals to speech. And in twenty twenty six, not only
21:48
Speaker 5: are we going to triple the number of electrodes from
21:52
Speaker 5: one thousand to three thousand for more capabilities, we're planning
21:56
Speaker 5: to have our first blind site participant to enable navigation.
22:08
Speaker 5: And in twenty twenty seven we're going to continue increasing
22:11
Speaker 5: channel counts, probably another triple, so ten thousand channels, and
22:16
Speaker 5: also enable for the first time multiple implants, so not
22:21
Speaker 5: just one in mortal cortex, speech cortex, or visual cortex,
22:24
Speaker 5: but all of the above. And finally, in twenty twenty eight,
22:31
Speaker 5: our goals to get to more than twenty five thousand
22:34
Speaker 5: channels per implant, have multiple of these, have ability to
22:38
Speaker 5: access any part of the brain for psychiatric conditions, pain dysregulation,
22:43
Speaker 5: and also start to demonstrate.
22:45
Speaker 6: What it would be like to actually integrate with AI,
22:56
Speaker 6: and all of.
22:57
Speaker 5: This is to say that we're really building towards set
23:00
Speaker 5: of fundamental foundational technology that would allow us to have
23:04
Speaker 5: hundreds of thousands, if not millions, of channels with multiple
23:06
Speaker 5: implants for whole bit interfaces that could actually solve not
23:10
Speaker 5: just these devoltating neurological conditions, but be able to go
23:13
Speaker 5: beyond the limits of our biology.
23:15
Speaker 6: And this vertical integration.
23:17
Speaker 5: And the talented team that we have at Neuralink has
23:20
Speaker 5: been and will continue to be the key recipe for
23:22
Speaker 5: rapid progress that we will be making. Just to recap
23:25
Speaker 5: real quick, neuralink is implanted with precision surgical robot. It's
23:30
Speaker 5: physically invisible, and one week later users are able to
23:34
Speaker 5: see their thoughts transform into actions and to share more
23:37
Speaker 5: about what that experience is like. I like to welcome
23:41
Speaker 5: to Hedge to the stage.
23:56
Speaker 12: What's up, guys. My name is Sahedge. I'm from the
23:59
Speaker 12: Brain Computer Inner team here at Neuralink, and I'm going
24:01
Speaker 12: to be talking about two things today. The first thing
24:04
Speaker 12: is what exactly is a neuralink device capable of doing
24:07
Speaker 12: right now? And the second one is how does that
24:10
Speaker 12: actually impact the day to day lives of our users?
24:14
Speaker 12: Very simply put, what the neuralink device does right now
24:19
Speaker 12: is it allows you to control devices simply just by thinking. Now,
24:23
Speaker 12: to put that a bit more concretely, I'm about to
24:26
Speaker 12: play a video of our first user. His name is Noland,
24:29
Speaker 12: if you remember from DJ section, And what Nolan is
24:32
Speaker 12: doing is he's looking at a normal off the shelf
24:36
Speaker 12: MacBook Pro and with his neuralink devices, you're going to
24:39
Speaker 12: see he's going to be able to control the cursor
24:42
Speaker 12: simply with his mind, no eye tracking, no other sensors.
24:46
Speaker 12: And what's special about this particular moment is this is
24:49
Speaker 12: the first time someone is using a neuralink device to
24:52
Speaker 12: fully control their cursor. This is not your ordinary brain
24:59
Speaker 12: controlled cursor. This is actually a record breaking control, literally
25:04
Speaker 12: on day one, beating decades of brain computer research. And
25:08
Speaker 12: I'm about to show you the clip on day one,
25:11
Speaker 12: Nolan breaking the b c I world record.
25:20
Speaker 9: Ship good.
25:26
Speaker 12: Man, he's a new world record holder.
25:38
Speaker 2: There's one of surprise.
25:40
Speaker 9: I thought it was higher.
25:41
Speaker 7: I thought I would have to get to five or something.
25:44
Speaker 9: Oh my gosh, that's crazy.
25:48
Speaker 11: It's pretty cool.
25:52
Speaker 2: Uh yeah.
25:55
Speaker 12: Another really fun thing you could do with the neuralink
25:58
Speaker 12: device outside of controlling a computer cursors. You can actually
26:01
Speaker 12: plug it in through USB through a lot of different devices.
26:04
Speaker 12: And here we actually have Nolan playing Mario Kart. Now,
26:08
Speaker 12: what's special about this particular clip is Nolan is not
26:11
Speaker 12: the only cyborg playing Mario Kart.
26:13
Speaker 2: In this clip.
26:14
Speaker 12: We actually have a whole community of users as mentioned earlier,
26:17
Speaker 12: and this is literally five of our first users of
26:20
Speaker 12: neuraling playing Mario Kart together over call. Yeah, now, yeah,
26:33
Speaker 12: Mario Kart.
26:34
Speaker 2: Is it's cool?
26:35
Speaker 12: You know you're using one joystick and then you're clicking
26:37
Speaker 12: like a couple of buttons to throw items. What would
26:40
Speaker 12: be even cooler is what if you could control two
26:43
Speaker 12: joysticks at once simultaneously with your mind. When I'm about
26:47
Speaker 12: to show you, and I think this is for the
26:48
Speaker 12: first time someone playing a first person shooter game with
26:51
Speaker 12: a brain computer interface. This is Alex and r J
26:55
Speaker 12: playing Call of Duty, controlling one joystick to move and
26:59
Speaker 12: then the other stick to I think, point your gun,
27:01
Speaker 12: and then shooting people as a button. Uh, here's the
27:04
Speaker 12: larger here's Alex shooting another person.
27:09
Speaker 3: Oh dear God, I don't know what to do, and
27:13
Speaker 3: I'm wanting to freaking shoot you. When I do, I
27:18
Speaker 3: notice shot in the face.
27:22
Speaker 12: Now that we have a bit of a sense of
27:24
Speaker 12: what the bc I can do, a very important question
27:27
Speaker 12: to answer is how does this impact the day to
27:29
Speaker 12: day lives of the people that use it every day.
27:33
Speaker 12: So I'm about to show you a clip going back
27:36
Speaker 12: to Nolan for a second, where he talks. We simply
27:39
Speaker 12: just asked him randomly during the day how he enjoys
27:42
Speaker 12: using the bc I a couple of months ago, and
27:45
Speaker 12: this is his candid reaction.
27:46
Speaker 13: I work basically all day from when I wake up
27:52
Speaker 13: trying to wake up at like six or seven am,
27:56
Speaker 13: and I'll do work until session. I'll do session, and
28:00
Speaker 13: then I'll work until you know, eleven, twelve pm or
28:06
Speaker 13: twelve am. While I'm doing, like I'm learning my languages,
28:14
Speaker 13: I'm learning my math, I'm like relearning all of my math.
28:19
Speaker 13: I am writing, I am doing the class that I
28:25
Speaker 13: sign up for, And I just I wanted to point
28:28
Speaker 13: out that, like, this is not something I would be
28:31
Speaker 13: able to do out like without the NEURALNK.
28:35
Speaker 12: Next, I want to talk a bit about Brad. You
28:37
Speaker 12: guys may already know him as the ALS cyborg, and
28:40
Speaker 12: Brad also has ALS and what separates him from our other
28:43
Speaker 12: users is he's actually nonverbal, so he can't speak. Why
28:47
Speaker 12: this is pretty relevant is he relies, at least before
28:50
Speaker 12: the neuralink on an igaze machine to communicate, and a
28:54
Speaker 12: lot of I Gaze machines you can't use outdoors. You
28:57
Speaker 12: really need like a dark room. So what this means
28:59
Speaker 12: is for the last six years since Brad's been diagnosed
29:02
Speaker 12: with alas, he's really unable to leave his house. Now
29:06
Speaker 12: with the neuralink device, We're gonna show you a clip
29:08
Speaker 12: of him with his kids at the park, shot by
29:11
Speaker 12: Ashley Vance and the team.
29:12
Speaker 3: Okay here ready, you absolutely do.
29:17
Speaker 14: I am absolutely doing more with Neuralink than I was
29:20
Speaker 14: doing with I Games. I have been a batman for
29:25
Speaker 14: a long time, but I go outside now. Going outside
29:29
Speaker 14: has been a huge blessing for me, and I can
29:32
Speaker 14: control the computer with telepody.
29:34
Speaker 9: Dad's watching.
29:36
Speaker 15: Look, he's watching on the camera.
29:39
Speaker 14: One of the arms.
29:40
Speaker 12: The last user I want to talk about is Alex.
29:43
Speaker 12: You've seen some clips of him earlier. What's specially about
29:46
Speaker 12: Alex to me is he's a fellow left handed guy
29:49
Speaker 12: who writes some cursive all the time. And what he
29:51
Speaker 12: mentioned is since his spinal cord injury from like three
29:54
Speaker 12: four years ago. He's been unable to just like draw
29:58
Speaker 12: or write, and he always brags about how good his
30:01
Speaker 12: handwriting was. So we actually got to put it in
30:03
Speaker 12: the test. We gave him a robotic arm. And I
30:05
Speaker 12: think this is the first time you tried using the
30:07
Speaker 12: robotic arm to write anything. And this is a spouta
30:10
Speaker 12: version of writing at the Convoy trial and drawing something.
30:16
Speaker 2: Sounds like that.
30:22
Speaker 12: Now, Yeah, controlling a robotic arm is cool. Uh, but
30:26
Speaker 12: this one has a clamp. And what would be cooler
30:29
Speaker 12: is if you could decode the actual fingers, the actual wrist,
30:31
Speaker 12: all the muscles of the hand in real time.
30:35
Speaker 2: Just in the past couple of weeks.
30:37
Speaker 12: Uh, we were able to do that with Alex, and
30:39
Speaker 12: you're about to see him and his uncle, uh playing
30:42
Speaker 12: a game.
30:43
Speaker 16: Rock Favor Scissors, Shoot damn It, rock vapor saissors, Shoot
30:51
Speaker 16: rock paper says or shoot rock paper saissor shoot that.
31:07
Speaker 12: Some more cool controlling. Yeah, that's pretty dope.
31:21
Speaker 11: I don't know.
31:26
Speaker 12: And uh, controlling a robotic hand on screen is obviously
31:30
Speaker 12: not super helpful for most people. Fortunately, we have connections
31:35
Speaker 12: with Tesla who have the Optimist hand, and we're actually
31:40
Speaker 12: actively working on giving Alex an Optimist hand so that
31:45
Speaker 12: you could actually control it in his real life. And
31:48
Speaker 12: here's actual replay of the end of that video using
31:51
Speaker 12: Alex's neural signals on an optist hand.
31:53
Speaker 2: Sean, if you want to play that.
32:02
Speaker 11: Yeah, actually let me.
32:07
Speaker 1: Let me maybe add a few things to that, which
32:10
Speaker 1: is so as we advance the neural link devices, your
32:16
Speaker 1: Chevelte actually have a full body control and sensors from
32:21
Speaker 1: an Optimist robot, So you could basically inhabit an Optimist robot.
32:26
Speaker 1: It's not just the hand, the whole the whole thing.
32:30
Speaker 1: So you could like basically mentally remote into an Optimist
32:36
Speaker 1: robot and.
32:37
Speaker 2: And uh, be kind of cool. The future is going
32:40
Speaker 2: to be weird, but but but pretty cool.
32:47
Speaker 1: And then now another thing that can be down also
32:51
Speaker 1: is like for people that have say lost a limb,
32:55
Speaker 1: lost an armor like or something like that, then we
32:59
Speaker 1: think in the future will be able to attach an
33:02
Speaker 1: optimus armor legs. And so you kind of like, I
33:06
Speaker 1: remember that scene from Star Wars where Luke Skywalker gets
33:10
Speaker 1: his hand, you know, chopped over with the lightslaber and
33:13
Speaker 1: he gets kind of a robot hand, and I think
33:16
Speaker 1: that's the kind of thing that we'll be able to
33:18
Speaker 1: do in the future working with the neuralink in Tesla,
33:22
Speaker 1: so that it goes far beyond just operating a robot hand,
33:26
Speaker 1: but replacing limbs and having kind of a whole body
33:28
Speaker 1: robot experience.
33:30
Speaker 2: And then I think.
33:30
Speaker 1: Another thing that will be possible, I think is very
33:35
Speaker 1: likely in the future, is to be able to bridge
33:39
Speaker 1: where the damaged neurons are, so you can take the
33:41
Speaker 1: signal from the brain and transmit that signal past where
33:46
Speaker 1: the neurons are damaged or strained to the rest of
33:49
Speaker 1: the body, so you could reanimate the body, so that
33:52
Speaker 1: if you have a neurallink implant in the brain and
33:54
Speaker 1: then one in the spinal cord, then you can actually
33:58
Speaker 1: bridge the signals and you could walk again and have
34:03
Speaker 1: full body functionality.
34:04
Speaker 2: Obviously that's what people would prefer.
34:06
Speaker 1: To be clear, we realize that that would be the
34:08
Speaker 1: preferred outcome, and so that even if you have a
34:13
Speaker 1: broken neck or if you saw we believe I'm actually
34:18
Speaker 1: at this point I'd say fairly confident that at some
34:20
Speaker 1: point in the future, well, we're able to restore full
34:22
Speaker 1: body functionality.
34:31
Speaker 4: Yeah, so hello, hello everyone. My name is Near and
34:34
Speaker 4: I am leading the BCA application group. And I think
34:38
Speaker 4: the video is that just shared with you. I probably
34:42
Speaker 4: watch them maybe thousands of time, but still I get
34:45
Speaker 4: a goose bump every time I watch them. And I
34:47
Speaker 4: think this is one of the cool perks here at
34:49
Speaker 4: new Link when you get the job, is that you
34:51
Speaker 4: might get goose bump every week or maybe every few
34:53
Speaker 4: days in good weeks. And and this is really fun
34:59
Speaker 4: as an engine. It's really cool because you can build
35:02
Speaker 4: a new feature. You can build a new machine learning
35:04
Speaker 4: model and new software feature and test it on the
35:07
Speaker 4: same day with a participant and get feedback. And you
35:12
Speaker 4: already saw with our first device, Telepathy, that we can
35:16
Speaker 4: address a very diverse needs of the different users that
35:19
Speaker 4: we have for moving a cursor, to playing games, to
35:23
Speaker 4: move everybody can with multiple fingers, and we could not
35:27
Speaker 4: have done it without the new Link device. The neural
35:29
Speaker 4: Link device gives us something that no other device can
35:32
Speaker 4: give us, which is a single neuron recording from thousands
35:37
Speaker 4: of channels simultaneously. The Telepathy products is basically recording the
35:41
Speaker 4: neural activity from the small area and the motor context
35:44
Speaker 4: that involve in execution of hand and arm movements. But
35:48
Speaker 4: if we go only about two or three inches below
35:51
Speaker 4: there's another branaria that's involved in execution of speech and
35:55
Speaker 4: with the same device, with the same machine learning model architecture,
35:59
Speaker 4: the same soft through a pipeline, the same surgical robot.
36:02
Speaker 4: We can have a new application and we can do
36:04
Speaker 4: it very quickly. It's really interesting that if we can
36:07
Speaker 4: decode someone intention to speak silently and non vocal communication,
36:13
Speaker 4: we can use that to revolutionize the way we interact
36:16
Speaker 4: with computers, with technology and with information. Instead of typing
36:20
Speaker 4: with your finger or like moving the mouse or talking
36:23
Speaker 4: to your phone, you'll be able to interact with computer
36:26
Speaker 4: with the speed of thoughts. It will make this interaction
36:30
Speaker 4: much more, much faster, and much more intuitive. The computers
36:33
Speaker 4: will understand what you want to do. And we can
36:36
Speaker 4: also expand that to AI. We can now build an
36:39
Speaker 4: interface with AI that you will be able to achieve information,
36:44
Speaker 4: will be able to start our thoughts anywhere, anytime, privately
36:48
Speaker 4: and silently. Again, because we build a fundamental technology a
36:53
Speaker 4: platform and we do everything in house, we own the
36:56
Speaker 4: intense stack from neons to pixels on the users. Now
37:01
Speaker 4: I'll pass passy to rouse to talk about UI for be.
37:04
Speaker 17: Said, Thank you nil each spike that our implant detects
37:15
Speaker 17: goes on a fairly remarkable journey to ultimately form a
37:19
Speaker 17: pixel on a participant's display, and that experience starts with,
37:24
Speaker 17: of course, unboxing, the very first time that a participant
37:28
Speaker 17: pairs to and meets their implant, this invisible part of
37:32
Speaker 17: their body and sees their own spikes materialize across the display.
37:39
Speaker 17: From there, they'll go into body mapping and actually imagine
37:43
Speaker 17: moving their arm again and get a feel for what
37:46
Speaker 17: feels natural to them and what doesn't, and they'll take
37:49
Speaker 17: that into calibration, using one of those motions to actually
37:56
Speaker 17: move a cursor again, iteratively refine their control as they
38:01
Speaker 17: go throughout this process, until finally they're teleported back to
38:07
Speaker 17: their desktop and can experience the magic of neural control
38:12
Speaker 17: for the very first time. And our control interfaces is
38:16
Speaker 17: where the OS integration that we do really shines, letting
38:21
Speaker 17: us adapt both control and feedback for every interaction. So
38:27
Speaker 17: for familiar interactions like scrolling, we can surface an indicator
38:31
Speaker 17: over the scrollable parts of the display at a touch
38:34
Speaker 17: of gravity to automatically pop a participant's cursor onto that
38:38
Speaker 17: indicator as they approach.
38:41
Speaker 9: Show the actual velocities.
38:42
Speaker 17: That we decode inside of it and add a bit
38:45
Speaker 17: of momentum to those velocities to carry them forward as
38:48
Speaker 17: they glide across the page. They are also unique interactions
38:52
Speaker 17: that we need to solve for in this space. For example,
38:55
Speaker 17: when a participant is watching a movie or just talking
38:58
Speaker 17: to somebody next to them, the brain is very active
39:02
Speaker 17: still and that activity can actually induce motion in the cursor,
39:05
Speaker 17: distracting them from that moment. So when a participant wants
39:09
Speaker 17: to just get their cursor out of the way, they
39:11
Speaker 17: can push it into the edge of the display to
39:13
Speaker 17: park it there, and of course we add gravity to
39:16
Speaker 17: sort of hold it still, but they can push it
39:19
Speaker 17: out with either just a firm push or in this case,
39:22
Speaker 17: a gesture, and of course it's what it goes. Without
39:27
Speaker 17: saying that, all of these control interfaces are designed hand
39:30
Speaker 17: in hand with our participants, so huge shout out to
39:32
Speaker 17: both Nolan and Brad for helping us design these two
39:36
Speaker 17: and those control interfaces, of course extend typing. We have
39:39
Speaker 17: a great software keyboard that does everything you'd expect it to,
39:43
Speaker 17: popping up when a participant clicks on a text field,
39:45
Speaker 17: giving them feedback about the click along the surface of
39:48
Speaker 17: the key and supporting both dictation and swipe.
39:58
Speaker 18: Hi everyone, I'm Harrison and engineer here at Neuralink, and
40:01
Speaker 18: I must say being an mL engineer at Neuralink is
40:04
Speaker 18: a bit like being a kid in a candy store.
40:08
Speaker 18: When you think of the inputs to most mL systems
40:11
Speaker 18: out there, you might think of pixels of tokens or
40:13
Speaker 18: of a user's Netflix watch history. The input to our
40:17
Speaker 18: systems is a little different. It is pure raw brain power.
40:22
Speaker 18: And when we think about the mL systems we can
40:24
Speaker 18: build here at Neuralink, really we're limited by our imagination
40:27
Speaker 18: and our creativity. There's no reason our mL systems can't
40:30
Speaker 18: do anything that the human brain can do, such as
40:33
Speaker 18: controlling a phone, typing, or even gaming.
40:38
Speaker 6: Right here to my.
40:39
Speaker 18: Left is actual footage of Alex, one of our participants,
40:42
Speaker 18: playing at first person shooter against RJ, another one of
40:45
Speaker 18: our participants. Now, for those unfamiliar with first person shooters,
40:49
Speaker 18: this is not a trivial feat. It requires two fully
40:52
Speaker 18: independent joysticks or four continuous degrees of control, as well
40:57
Speaker 18: as multiple reliable buttons. Now, contrary to popular belief, the
41:02
Speaker 18: Neuralink does not simply read people's minds it's simply reading
41:07
Speaker 18: neuronal activations corresponding to motor intent.
41:10
Speaker 9: So one of the fun.
41:11
Speaker 18: Challenges with this project was figuring out which motions we're
41:14
Speaker 18: going to give out to the joystick. We started with
41:17
Speaker 18: the typical left thumb and right thumb, but quickly found
41:19
Speaker 18: the dominant hand overshadow the non dominant hand. My personal
41:23
Speaker 18: favorite is we had one of our participants imagine walking
41:26
Speaker 18: for the left joystick and aiming for the right joystick,
41:29
Speaker 18: so in game, they were simply doing naturalistic motions like
41:33
Speaker 18: you might do in virtual reality in Ready Player one,
41:35
Speaker 18: and that was really cool to watch.
41:37
Speaker 9: What we ended up on was the thumb for.
41:39
Speaker 18: The left joystick and the wrist for the right joystick,
41:43
Speaker 18: and I challenged the audience to try to replicate their motions.
41:46
Speaker 18: I'm really in all of them being able to pull
41:47
Speaker 18: this off. I want to talk a bit about the
41:50
Speaker 18: progress to our cursor calibration experience.
41:53
Speaker 2: To my left.
41:53
Speaker 18: Here you can see RJ completing his first ever cursor
41:56
Speaker 18: calibration with a redesigned open up flow where he first
41:59
Speaker 18: gathering for about his intent and how to map the
42:01
Speaker 18: neural activity to the first time he controls a cursor
42:04
Speaker 18: to the final product where he has smooth and fluid
42:06
Speaker 18: control of his computer. And most remarkably, this experience took
42:11
Speaker 18: only fifteen minutes from a start to finish, fifteen minutes
42:15
Speaker 18: from not fifteen minutes from no control to fluid computer use.
42:26
Speaker 18: Contrast that to a year and a half ago with
42:28
Speaker 18: P one, where that was multiple hours to get to
42:30
Speaker 18: the same level of control and several engineers standing around
42:33
Speaker 18: a table pulling their hair out. There was virtually no
42:36
Speaker 18: need for neuralink engineers to even be at this session.
42:38
Speaker 18: This was basically it out of the box experience for
42:41
Speaker 18: our participant, and even more remarkably, we're continuing to smash
42:48
Speaker 18: day one records, with RJ being able to achieve seven BPS.
42:52
Speaker 9: On his very first day with a neuralink.
42:58
Speaker 18: Now, such an effective and fishing calibration process is only
43:01
Speaker 18: made possible by high fidelity estimations of a user intention
43:05
Speaker 18: or labels. And to briefly illustrate just how challenging a
43:10
Speaker 18: problem that is, this is an animation of myself trying
43:13
Speaker 18: to draw circles on my desktop with a mouse. Now,
43:16
Speaker 18: the task was simple, draw uniform circles at a constant speed,
43:20
Speaker 18: repeatedly and as you can see by that animation, I
43:23
Speaker 18: am horrible at that.
43:25
Speaker 2: Even though my.
43:25
Speaker 18: Intent was pretty obvious unambiguous, the execution was really poor.
43:29
Speaker 18: There is a ton of variation in both speed and
43:33
Speaker 18: the shape itself. To visualize this a little differently, each
43:37
Speaker 18: row here is one of those circles unwound in time
43:40
Speaker 18: with synchronized starts, and you can just see how much
43:43
Speaker 18: variation there is in the timing of each circle as
43:45
Speaker 18: well as what I'm doing at any given point in time.
43:49
Speaker 18: Orthogonal to the labeling problem is neural nonstationarity, or the
43:53
Speaker 18: tendency of neural signals to drift over time. And I
43:56
Speaker 18: think that's honestly a beautiful thing, right If your neurals
44:00
Speaker 18: signals didn't drift, you couldn't grow. When you wake up
44:02
Speaker 18: the next day, you're not the same person you were
44:04
Speaker 18: the day before. You've learned, you've grown, You've changed, and
44:08
Speaker 18: so too must your neural data change. This animation here
44:11
Speaker 18: is a simple illustration of the learned representation by the
44:14
Speaker 18: decoder and how it drifts the further away we get
44:16
Speaker 18: from the day it was trained on. This is one
44:18
Speaker 18: of the key challenges we need to solve here At
44:20
Speaker 18: neuralink to unlock fluid and a product level experience for
44:23
Speaker 18: our users.
44:32
Speaker 15: Everyone here, everyone, My name is Joey. Blind Sight is
44:41
Speaker 15: our project to build a visual prosthesis to help the
44:44
Speaker 15: blind see again. Users would wear a pair of glasses
44:50
Speaker 15: with an embedded camera and receive an implant in their
44:54
Speaker 15: visual cortex. Scenes from the environment are recorded by the
44:58
Speaker 15: camera and processed in the patterns of stimulation delivered to
45:03
Speaker 15: the brain, causing visual perception and restoring functionality. Now blind
45:10
Speaker 15: Site will be enabled by placing our implant into visual cortex.
45:13
Speaker 15: This is a new brain area for us, and this
45:16
Speaker 15: brings new opportunities and challenges. So the surface of the
45:21
Speaker 15: brain for visual cortex represents just a few degrees of
45:24
Speaker 15: angle in the center of the visual field. Larger fields
45:28
Speaker 15: of view are represented deep within the cortical folds of
45:31
Speaker 15: the calcarine fissure. Our threads are able to access these
45:36
Speaker 15: deeper structures, providing the possibility of restoring vision over a functional,
45:41
Speaker 15: useful visual field. So the N one implant has had
45:44
Speaker 15: experimental stimulation capabilities for quite some time, but our new
45:49
Speaker 15: S two chip is designed from the ground up for stimulation.
45:53
Speaker 15: It provides over sixteen hundred channels of electrical stimulation, high
45:58
Speaker 15: dynamic range recording abilities, and a wide range of microstimulation
46:03
Speaker 15: currents and voltages. We can achieve these capabilities because we're
46:08
Speaker 15: vertically integrated and we design this CUSTOMASIC in house. Similarly,
46:13
Speaker 15: we design and fabricate our electrode threads in house, and
46:16
Speaker 15: here you can see one of our standard threads designed
46:20
Speaker 15: for recording in an electron micrograph. For blind site, our
46:24
Speaker 15: requirements are a little different and our vertical integration allows
46:28
Speaker 15: us to rapidly iterate on the design and manufacturing of
46:31
Speaker 15: these threads for this new purpose. So here I'm using
46:35
Speaker 15: red arrows to highlight the electrode contacts which are optimized
46:40
Speaker 15: for stimulation, and as you can see, they're a little
46:42
Speaker 15: bit larger, which results in a lower electrical impedance for
46:45
Speaker 15: safe and effective charge delivery, which is important for blind site. Now,
46:52
Speaker 15: how can we calibrate our implant for blind site? So
46:55
Speaker 15: here's one way. We stimulate on the array, picking say
46:59
Speaker 15: three different channels. The user perceives something say three spots
47:04
Speaker 15: of light somewhere in their visual field and points at them.
47:07
Speaker 15: We track their arm and eye movements and repeat this
47:10
Speaker 15: process for each of the channels on the array, and
47:15
Speaker 15: here's what a simulated example of a blind site vision
47:18
Speaker 15: could look like after calibration. Now I showed you how
47:32
Speaker 15: for blind site we need to insert threads deeper into
47:35
Speaker 15: the brain than we have previously and doing this requires
47:38
Speaker 15: state of the art medical imaging. So we worked with
47:42
Speaker 15: Semens to get some of the best scanners on earth.
47:45
Speaker 15: We built out our imaging core from scratch in the
47:48
Speaker 15: past year. Actually it was faster than that. It was
47:50
Speaker 15: about four months from dirt to done. Since bringing the
47:54
Speaker 15: scanners online, we scanned over fifty internal participants, building out
47:58
Speaker 15: a database of human structural and functional anatomy. What can
48:02
Speaker 15: we do with the imaging information from these scanners? So
48:05
Speaker 15: medical imaging can be used for surgical placement. It lets
48:08
Speaker 15: us parcelate out brain regions by their function, and we
48:11
Speaker 15: use our imaging capabilities to refine the placement for telepathy.
48:15
Speaker 15: It also gives us the capability of target new brain
48:18
Speaker 15: regions for future products such as blind site or speech
48:20
Speaker 15: pres thesis, and we're working towards more capabilities. So one
48:24
Speaker 15: click automated planning of surgery from functional images to robot
48:28
Speaker 15: insertion targets. Here you can see a screen capture from
48:31
Speaker 15: one of our in house tooling to do end to
48:33
Speaker 15: end surgical planning. You can see a region of motor
48:35
Speaker 15: cortex known as hand knob and the thread trajectory plans
48:39
Speaker 15: that'll be sent directly to the robot. This is a
48:44
Speaker 15: really incredible degree of automation that's only possible because we're
48:47
Speaker 15: controlling the system from one end to the other.
48:54
Speaker 10: My name is John and I lead the robot mechanical team.
48:57
Speaker 10: This is our current R one robot. It was used
49:00
Speaker 10: to implant the first seven participants. This robot works really well,
49:04
Speaker 10: but it has a few flaws, one of which is
49:06
Speaker 10: the cycle time is rather slow, so to insert each
49:09
Speaker 10: thread it takes in the best case scenario seventeen seconds,
49:13
Speaker 10: and many cases external disturbances cause us to have to
49:16
Speaker 10: retry to grasp that thread and then reinsert it. To
49:20
Speaker 10: scale our number of neurons access through higher channel acount
49:24
Speaker 10: increase numbers of threads, we need to have a much
49:26
Speaker 10: faster cycle time. So let me introduce our next generation robot,
49:31
Speaker 10: which is right here. Through rethinking the way that we
49:43
Speaker 10: hold the implant in front of the robot, by holding
49:45
Speaker 10: it directly in front on the robot, head. We will
49:48
Speaker 10: achieve an eleven times cycle time improvement, so each thread
49:51
Speaker 10: takes one and.
49:52
Speaker 2: A half seconds.
49:53
Speaker 10: We also scale up a lot of surgery workflow process
49:58
Speaker 10: improvements through deleting the separate an operator station and implant
50:01
Speaker 10: stand Now the outside of the robot looks pretty similar
50:05
Speaker 10: between the two, but what's inside that really counts. Each
50:08
Speaker 10: system has been redesigned from the ground up with a
50:11
Speaker 10: focus on re liability, manufacturability, serviceability, and using a lot
50:16
Speaker 10: of our vertical integration techniques. It's enabled us to have
50:19
Speaker 10: a lot more control of the system and to end
50:22
Speaker 10: now that fast cycle time doesn't mean much if it's
50:25
Speaker 10: not compatible with a significant portion of the human population.
50:28
Speaker 10: Prior to each surgery, we scan our participants anatomy and
50:32
Speaker 10: ensure that they will be compatible with the robot and
50:35
Speaker 10: vice versa. Unfortunately, the robot isn't compatible with everyone, so
50:39
Speaker 10: we had to extend the reach of the needle in
50:41
Speaker 10: the next generation robot, and now we're compatible with more
50:43
Speaker 10: than ninety nine percent of the human population. We've also
50:46
Speaker 10: increased the depth that the needle can insert threads. Now
50:49
Speaker 10: we can reach more than fifty millimeters from the surface
50:51
Speaker 10: of the brain, accessing and enabling new indications. We have
50:55
Speaker 10: to produce a ton of custom sterile components for each surgery.
50:59
Speaker 6: We actually supple more than twenty of these parts.
51:01
Speaker 10: Many of these parts are made through traditional C and
51:04
Speaker 10: C manufacturing capabilities, which we do just on the other
51:06
Speaker 10: side of this wall actually, and some custom developed processes
51:10
Speaker 10: like this pemptosecond laser milling used to manufacture the tip
51:13
Speaker 10: of the needle. Now, these processes take quite a bit
51:16
Speaker 10: of time, effort, and cost, So let's take a look
51:19
Speaker 10: at how we're going to reduce costs and time for
51:22
Speaker 10: one of the components. So the current needle cartridge has
51:25
Speaker 10: a total cycle time of about twenty four hours and
51:28
Speaker 10: the machine components cost about three hundred and fifty dollars.
51:31
Speaker 10: The final assembly is performed by a set of like
51:33
Speaker 10: highly skilled technicians. They have to glue a one hundred
51:36
Speaker 10: and fifty micron diameter canua onto this wireedm machined, seainless
51:40
Speaker 10: steeled base plate. They have to electro polish a forty
51:43
Speaker 10: micron wire into a sharp taper, and then they have
51:45
Speaker 10: to thread that forty micron wire into a sixty micron
51:49
Speaker 10: hole in the canua. This is done manually and then
51:52
Speaker 10: they finally have to laser weld all the components together.
51:57
Speaker 10: Next generation needle cartridge takes only thirty minute of cycle
52:00
Speaker 10: time and fifteen dollars in component. We were able to
52:03
Speaker 10: delete the wire idium machined base plate and the canuo
52:06
Speaker 10: Gulian step by switching to an insert mold it component.
52:09
Speaker 10: So we get a box of these base plates with
52:11
Speaker 10: the canules already installed, for like a thousand of them
52:13
Speaker 10: for like a couple five ten dollars apiece. We also
52:17
Speaker 10: deleted the electropolishing step with the revised needle tip geometry,
52:20
Speaker 10: which is also compatible with inserting the threads through the dura.
52:24
Speaker 10: We have a few revised manufacturing techniques to delete the
52:28
Speaker 10: manual threading through basically a funnel. Rather simple, but it
52:31
Speaker 10: has been a big impact. And then we're able to
52:33
Speaker 10: delete the laser welding through using crimping.
52:41
Speaker 2: By Hi.
52:47
Speaker 19: I'm Julian. I'm one of the leads on the implant team.
52:51
Speaker 19: So the way humans communicate today, if they want to
52:54
Speaker 19: output information is by using their hands and their voice
52:57
Speaker 19: as I'm doing right now, And if you want to
53:00
Speaker 19: receive information, you use your ears and your eyes, and
53:02
Speaker 19: of course that's how you're receiving this very torque but
53:07
Speaker 19: we've built this implant, and this implant is very special
53:11
Speaker 19: because it is the first time that we're able to
53:14
Speaker 19: add a completely new mode of data transfer into and
53:17
Speaker 19: out of the brain. If you look at this device
53:21
Speaker 19: in a nutshell, it's really just sampling voltages in the
53:24
Speaker 19: brain and sending them over radio. But if you zoom
53:27
Speaker 19: out and look at the system from end to end,
53:30
Speaker 19: what you actually see is that we're connecting your brain
53:34
Speaker 19: or a biological neural net to a machine learning model
53:38
Speaker 19: or a silicon neural net on the right hand side.
53:41
Speaker 19: And I actually think this is really elegant because the
53:45
Speaker 19: machine learning model on the right hand side is in
53:48
Speaker 19: fact inspired by neurons on the left hand side, and
53:51
Speaker 19: so in some sense, we're really extending the fundamental substrate
53:55
Speaker 19: of the brain for the first time. We're able to
53:58
Speaker 19: do this in a mass market product. That's a very
54:00
Speaker 19: very special piece of hardware. So these are some of
54:11
Speaker 19: the first implants that we ever built. There are electrodes
54:15
Speaker 19: that were made with our in house lithography tools. We
54:19
Speaker 19: have custom A six that we also designed in house,
54:21
Speaker 19: and this was really a platform for us to develop
54:24
Speaker 19: the technology that allows us to sense micro level vaults
54:28
Speaker 19: in the brain across thousands of channels simultaneously. We learn
54:32
Speaker 19: a lot from this. But as you'll notice in the
54:34
Speaker 19: right to images, there are USBC connectors on these devices.
54:38
Speaker 19: These were not really the most implantable implants. This next
54:43
Speaker 19: set of images are the wireless implants, and there was
54:47
Speaker 19: a complete evolution that we went through to add the battery,
54:50
Speaker 19: the antenna, the radio, and to make it actually fully implantable.
54:55
Speaker 19: Once it's implanted, it's completely invisible. It's very compact, it's modular,
55:00
Speaker 19: and it's a general platform that you can use in
55:01
Speaker 19: many places in the brain. Going from that top row
55:06
Speaker 19: to the bottom row is very challenging. The implant you
55:10
Speaker 19: see on the bottom right here is in fact the
55:12
Speaker 19: device that we have working in seven participants today and
55:15
Speaker 19: it's augmenting their brain every day and restoring their autonomy.
55:19
Speaker 19: But getting to that point involved a huge number of
55:23
Speaker 19: formidable engineering challenges. We first had to make hometic enclosure,
55:27
Speaker 19: passing one thousand separate conductors through the enclosure of the device.
55:31
Speaker 19: We had to figure out how to make charging seamless
55:33
Speaker 19: and work with very tight thermal constraints in a very
55:36
Speaker 19: very small area, and then we also had to scale
55:38
Speaker 19: up our testing infrastructure so that we could support large
55:41
Speaker 19: scale manufacturing and very safe devices and have confidence in
55:45
Speaker 19: our iteration cycle.
55:47
Speaker 9: So what's next.
55:48
Speaker 19: We're going to be increasing our manufacturing so that we
55:52
Speaker 19: don't just produce a certain like a small number of
55:55
Speaker 19: implants per year, but thousands and then eventually millions of
55:57
Speaker 19: implants per year. We're also going to be increase seeing
56:00
Speaker 19: channel count. More channels means more neurons are sensed, which
56:04
Speaker 19: means more capabilities. In some sense, we often think a
56:08
Speaker 19: lot about the Moore's law of neurons that were interacting
56:11
Speaker 19: with and in the same way that Moore's law propelled
56:14
Speaker 19: forward many subsequent revolutions in computing, we think that sensing
56:18
Speaker 19: more and more neurons will also completely redefine how we
56:22
Speaker 19: interact with computers and reality at large.
56:26
Speaker 9: I want to leave you with one final thought.
56:29
Speaker 19: When I was a child, I used a fifty six
56:31
Speaker 19: kilo bit moodem to access the Internet. If you remember
56:36
Speaker 19: what it's like, you would go to a website.
56:38
Speaker 2: You're lucky, You're lucky, bastard. Yeah, when I was a child,
56:43
Speaker 2: we had acoustic couplers.
56:45
Speaker 1: Oh, yeah, okay, so just just beat at each other.
56:48
Speaker 9: Yeah.
56:49
Speaker 19: The first modem was the acoustic Coppler. Incredible device honestly.
56:54
Speaker 19: But then if you I guess, if you're my age,
56:56
Speaker 19: you started with fifty six kit modem and you would
57:04
Speaker 19: go to a website and like there would be an
57:06
Speaker 19: image and it would it would scroll like slowly, it
57:10
Speaker 19: was loading pixel by pixel on the screen.
57:12
Speaker 9: So that's what it's like to be bandwidth limited.
57:15
Speaker 19: Now I imagine using the current Internet with that same modem.
57:19
Speaker 19: It's like it's inconceivable, it would be impossible to do.
57:23
Speaker 9: So what broadband Internet did to.
57:26
Speaker 19: The fifty six kilo bit modem is what this hardware
57:29
Speaker 19: is going to do to the brain. We're trying to
57:32
Speaker 19: drastically expand the amount of bandwidth that you have access
57:35
Speaker 19: to to have a much richer experience and superhuman capabilities.
57:46
Speaker 5: So I guess, just to kind of close out and
57:49
Speaker 5: to recap today, neuralink is working reliably and has already
57:54
Speaker 5: changed the lives of seven participants and making a real impact.
58:00
Speaker 5: And our next milestone is to go to market and
58:03
Speaker 5: enable scaling of this technology to thousands of people, as
58:07
Speaker 5: well as expand functionalities beyond just the movement to enable
58:11
Speaker 5: robotic sophisticated robotic arm control, speech vision, give site back,
58:17
Speaker 5: and even.
58:18
Speaker 6: Getting to the speed of thought.
58:20
Speaker 5: I hope you got a good sort of sample of
58:22
Speaker 5: our technology stack and the challenges that we have. And
58:27
Speaker 5: I like to handle over the mic to Elon for
58:30
Speaker 5: any closing remarks.
58:33
Speaker 1: Well, we're trying to give you a sense of the
58:38
Speaker 1: depth of.
58:38
Speaker 2: Talent at your link.
58:39
Speaker 1: There's a lot of really smart people working on a
58:42
Speaker 1: lot of important problems. This is one of the most
58:46
Speaker 1: difficult things to to actually succeed in creating and have
58:50
Speaker 1: at work and work at scale and be reliable and
58:53
Speaker 1: available for millions of people at an affordable price. So
58:59
Speaker 1: super heart problem and we'd like to have you come
59:03
Speaker 1: join and help.
59:04
Speaker 2: Us solve it. Thank you.