Latest / Elon Musk Thinking / Elon Musk Interview Talked About The Purpose of Neuralink.

Transcript

Elon Musk Interview Talked About The Purpose of Neuralink. · Elon Musk Thinking

• 0:00
  How the brain works is very limited. You know, we've got fMRI,
• 0:06
  which is that that's kind of like putting us, you know, a stethoscope
• 0:10
  on the outside of a factory wall and been putting it like all over the
• 0:13
  factory wall, and you can sort of hear the sounds, but you don't
• 0:16
  know what the machines are doing. Really it's hot. You can infer a
• 0:20
  few things, but it's very broad breaststroke. In order to really know what's
• 0:24
  going on in the brain, you really need you have to have high precision
• 0:27
  sensors, and then you want to have stimulus and response, Like if you
• 0:31
  trigger a neuron, what, how do you feel? What do you see?
• 0:34
  How does the change of perception of the world you're speaking to physically?
• 0:38
  Just getting close to the brain, being able to measure signals from the brain
• 0:41
  will give us sort of open the door inside the factory. Yes, exactly.
• 0:47
  Being able to have high precision sensors that tell you what individual neurons are
• 0:52
  doing, and then being able to trigger the neuron and see what the response
• 0:57
  is in the brain, so you can see the consequences of of of a
• 1:02
  if you fire this neuron, what happens? How do you feel? What
• 1:04
  has changed. It's it'll be really profound to have this in people, because
• 1:10
  people can articulate their change, like if there's a change in mood or if
• 1:15
  they you know, if they can tell you if they can see better or
• 1:19
  hear better, or be able to form sentences better or worse, or you
• 1:26
  know, their memories are jogged or that kind of thing. So on the
• 1:30
  on the human side, there's this incredible general malleability plasticity of the human brain.
• 1:34
  The human brain adapts, adjusts, and so on. So that's not
• 1:38
  that plastic you're totally frank, so there's a firm structure, but there nevertheless
• 1:42
  there's some plasticity. And the open question is, so if I could ask
• 1:47
  a broad question, is how much that plasticity can be utilized? Sort of
• 1:52
  on the human side, there's some plasticity in human brain, and on the
• 1:56
  machine side, we have networks machine learning are ficial intelligence. It's able to
• 2:02
  adjust and figure out signals. So there's a mysterious language that we don't perfectly
• 2:07
  understand that's within the human brain, and then we're trying to understand that language
• 2:13
  to communicate both directions. So the brain is adjusting a little bit. We
• 2:17
  don't know how much, and the machine is adjusting. Where do you see
• 2:23
  as they try to sort of reach together, almost like with an alien species,
• 2:27
  try to find a protocol, communication protocol that works. Where do you
• 2:30
  see the biggest the biggest benefit arriving from on the machine side or the human
• 2:36
  side? Do you see both of them working together? I should think the
• 2:38
  machine side is far more malleable than the biological side by huge amount. So
• 2:46
  it'll be the machine that adapts to the brain. That's the only thing that's
• 2:50
  possible to brain can't adapt that well to the machine. You can't have neurons
• 2:55
  start to regard an electrode as another neuron, not just that this like the
• 3:00
  pulse and so something else is pulsing. So there is that, there is
• 3:06
  that that elsticity in the interface, which we believe is something that can happen.
• 3:10
  But the vast majority of the malleability will have to be on the machine
• 3:14
  side. But it's interesting when you look at that synaptic plasticity at the interface
• 3:19
  side, there might be like an emergent plasticity because it's a whole nother It's
• 3:24
  not like in the brain. It's a whole nother extension of the brain.
• 3:28
  You know, we might have to redefine what it means to be malleable for
• 3:31
  the brain, so maybe the brain is able to adjust to external interfaces.
• 3:36
  There will be some adjustments to the brain because there's gonna be something reading and
• 3:38
  simulating the brain, and so it will adjust to that thing. But most
• 3:47
  the vast majority of the adjustment will be on the machine side. This is
• 3:53
  just, this is just. It has to be that otherwise it will not
• 3:55
  work. Ultimately, like we currently operate on two layers. We have sort
• 3:59
  of lumbaco like prime primative brain layer, which is where all of our kind
• 4:03
  of impulses are coming from. It's sort of like we've got we've got like
• 4:08
  a monkey brain where a computer stuck on it. That's that's the human brain,
• 4:13
  and a lot of our impulses and everything are driven by the monkey brain
• 4:15
  and the computer. The cortex is constantly trying to make the monkey monkey brain
• 4:20
  happy. It's not the cortex that's steering the monkey brains. The monkey brain
• 4:25
  is steering the cortex, you know. But the cortex is the part that
• 4:30
  tells the story of the whole thing. So we convince ourselves it's more interesting
• 4:33
  than just the monkey brain. The cortex is like what we call like human
• 4:38
  intelligence, you know. So it's like the that's like the advanced computer relative
• 4:42
  to other creatures. Other creatures do not have either really they don't have the
• 4:48
  computer, or they have a very weak computer relative to humans. But it's
• 4:56
  it's like it sort of seems like, surely the really smart thing should can
• 5:00
  the dumb thing, but actually, don't think controls a small thing. So
• 5:04
  do you think some of the same kind of machine learning methods or whether that's
• 5:09
  natural language processing applications are going to be applied for the communication between the machine
• 5:15
  and the brain to learn how to do certain things like movement of the body,
• 5:20
  how to process visual stimuli and so on. Do you see the value
• 5:26
  of using machine learning to understand the language of the two way communication with the
• 5:30
  brain? Sure? Yeah, absolutely. I mean we're neural net and that
• 5:35
  you know, AI is basically neural net. So it's like digital neural net
• 5:41
  will interface with biological neural net and hopefully bring us along for the ride.
• 5:46
  But the vast majority of our of our intelligence will be digital. There's like
• 5:54
  like like think of like the difference in intelligence between the cortex in your olympics
• 6:00
  system is gigantic. Your olympic system really has no comprehension of what the hell
• 6:06
  the cortex is doing. You know, it's just literally hungry, you know,
• 6:13
  or tired or angry or taxi or something. You know, it's just
• 6:19
  and then that in case, that's that impulse to the cortex and tales the
• 6:25
  cortex to go satisfy that. So then a lot of a great deal of
• 6:29
  like a massive amount of thinking, like truly stupendous amount of thinking has gone
• 6:33
  into sex without purpose, without procation, without procreation, which which which is
• 6:43
  actually quite a silly action in the absence of procreation. It's it's a bit
• 6:48
  silly. Why are you doing it? Because it makes the limbic system happy,
• 6:53
  that's why. That's why. But it's pretty absurd, really, well,
• 6:58
  the whole of existence, that's pretty absurd in some kind of sense.
• 7:01
  Yeah, But I mean, this is a lot of computation has gone into
• 7:05
  how can I do more of that with approcreation not even being a factor?
• 7:11
  This is I think a very important era of research, finds FW an agency
• 7:17
  that should receive a lot of funding, especially after this conversation. If I
• 7:20
  propose the formation of a new agency, Oh boy, what is the most
• 7:30
  exciting or some of the most exciting things that you see in the future impact
• 7:33
  of neuralink both on the science and engineering and societal broad impact. So neuralink,
• 7:40
  I think at first we'll solve a lot of brain related diseases. So
• 7:45
  it could be anything from like autism, schizophrenia, memory loss, like everyone
• 7:48
  experiences memory loss at certain points in an age. Parents can't remember their kids'
• 7:54
  names and that kind of thing. So there's a tremense amount of good that
• 7:59
  neuralink can do in solving critical critical damage to the brain or the spinal cord.
• 8:07
  There's a lot that can be done to improve quality of life of individuals,
• 8:11
  and that will be those will be steps along the way, and then
• 8:15
  ultimately it's intended to address the risk existential risk associated with a digital superintelligence,
• 8:24
  like we will not be able to be smarter than a digital supercomputer. So
• 8:31
  therefore, if you cannot beat them, join them and release, we won't
• 8:35
  have that option. So you have hope that your link will be able to
• 8:41
  be a kind of connection to allow us to merge to ride the wave of
• 8:46
  the improving AI systems. I think the chance is above zero percent, so
• 8:52
  it's non zero. There's a chance, and that's what have you seen,
• 8:56
  Doumin Dummer? Yes, yes, so I'm saying there's a chance. He's
• 9:01
  saying one in a billion or one in a million whatever. It was a
• 9:03
  dumb and dumber you know. It went from maybe one in a million to
• 9:07
  improving. Maybe it'll be one in a thousand, and then one one hundred,
• 9:09
  then one in ten. Depends on the rate of improvement of neuralink and
• 9:13
  how fast we're able to do make progress. You know. Well, I've
• 9:18
  talked to a few folks here that quite brilliant engineers, so I'm excited.
• 9:22
  Yeah, I think it's like fundamentally good, you know, you know,
• 9:24
  giving somebody back full motor control after they've had a spinal cord injury, you
• 9:31
  know, restoring brain functionality after a stroke, solving debilitating genetically orange brain diseases.
• 9:37
  These are all incredibly great, I think. And in order to do
• 9:41
  these, you have to be able to interface with neurons at detailed level and
• 9:46
  need to be build fire the right neurons, read the right neurons, and
• 9:50
  and then effectively you can create a circuit, replace what's broken with with silicon
• 9:58
  and such a fill in the the missing functionality, and then over time we
• 10:05
  can have we develop a tertiary layer. So if like Olympic system is a
• 10:09
  primary layer, then the cortex is like the second layer. And I said
• 10:13
  that, you know, obviously the cortex is vastly more intelligent than the Olympic
• 10:16
  System. But people generally like the fact that they have Olympic System and cortex.
• 10:20
  I haven't met anyone who wants to lead either one of them. They're
• 10:24
  like, okay, I'll keep them both. That's cool. The Olympic System
• 10:26
  is kind of fun. That's what the fun is. Absolutely, And then
• 10:31
  people generally don't lose the cortex either, right, so they're like having the
• 10:35
  cortex and the Olympic System. Yeah, and then there's a tertiary layer,
• 10:39
  which will be digital superintelligence. And I think there's room for optimism given that
• 10:45
  the cortex, the cortex is very intelligent and the Olympic System is not.
• 10:50
  Yet they work together. Well, perhaps they can be a tertiary layer where
• 10:56
  digital superintelligence lies, and that will be vastly more intelligent than the cortex,
• 11:00
  but still coexist peacefully and in of a nine manner with the cortex Olympic System.
• 11:05
  That's a super exciting future, both in low level engineering that I saw
• 11:09
  as being done here and actual possibility in the next few decades. It's important
• 11:16
  that neuralinks solved this problem sooner rather than later, because the point at which
• 11:20
  we have digital superintelligence, that's when we pass the singularity and things become just
• 11:24
  very uncertain. It doesn't mean that they're necessarily bad or good. For the
• 11:26
  point of which we passed singularity, things become extremely unstable. So we want
• 11:31
  to have a human brain interface before the singularity, or at least not long
• 11:35
  after it, to minimize existential risk for humanity and consciousness as we know it.
• 11:43
  But there's a lot of fascinating actual engineering and low level problems here in
• 11:46
  your link that yeah, quite exciting. The problems that we face neuralink are
• 11:54
  material science, electrical engineering, software, mechanical engineering, micro vocation. It's
• 12:01
  a bunch of engineering disciplines essentially. That's what it comes down to is you
• 12:05
  have to have a tiny electrode, so it's so small it doesn't hurt neurons,
• 12:13
  but it's got to last for as long as a person, so it's
• 12:16
  gonna last for decades. Uh. And then you've got to take that signal
• 12:20
  you've got to process that single signal locally at low power. So we need
• 12:26
  a lot of chip design engineers that, you know, because we've got to
• 12:31
  uh signal processing and do so in a very power efficient way so that we
• 12:37
  don't heat your brain of because the brain is very heat sensitive. And then
• 12:41
  and then we've got to take those signals, we're going to do something with
• 12:43
  them. And then we've got to stimulate and stimulate the back to to you
• 12:48
  know, so you could bi directional communication. So if somebody's good at material
• 12:54
  science, software, mechanical engineering, electrical engineering, trip design, microfabrication,
• 13:01
  that's what those are the things we need to work on. We need to
• 13:05
  your good at material science so that the we can have tiny electrodes that last
• 13:07
  long time. And it's a tough thing with the science problems is a tough
• 13:11
  one because you're trying to read and simulate electrically in an electric an electrically active
• 13:20
  area. Your brain is very electrically active, electrochemically active. So how do
• 13:24
  you have say, a coding on the electrode that doesn't dissolve over time and
• 13:31
  uh and is safe in the brain. This is a very hard problem.
• 13:35
  And then and then how do you collect those signals in a way that is
• 13:41
  the most efficient because you really just have very tiny amounts of power to process
• 13:46
  those signals, you know, and then we need to automate the whole thing.
• 13:50
  So it's like laser, you know. So it's it's it's not if
• 13:54
  this is done by neurosurgeons. There's no way it can scale to a large
• 13:58
  numbers of people, and it needs to scales large numbers of people because I
• 14:01
  think ultimately we want the future repeated to be determined by a large number of
• 14:07
  humans. Do you think that's this has a chance to revolutionize surgery period,
• 14:13
  So neurosurgery, and yeah, for sure, it's gotta be like laser.
• 14:18
  Like if laser had to be hand done done by hand by a person,
• 14:22
  that wouldn't be great. You know, it's done by a robot and then
• 14:30
  arphomologist kind of just needs to make sure your your heads in the right position
• 14:33
  and then they just press the button and go