1 00:00:00,960 --> 00:00:01,320 Speaker 1: Sure. 2 00:00:02,759 --> 00:00:06,240 Speaker 2: Well, this is going to sound somewhat esoteric and maybe 3 00:00:07,120 --> 00:00:09,640 a bit strange, but I was actually trying to figure 4 00:00:09,640 --> 00:00:14,119 out how to mitigate the risk of digital superintelligence to 5 00:00:14,199 --> 00:00:17,399 the agree that that we can improve our bandwidth to 6 00:00:17,519 --> 00:00:23,000 our digital tertiary self. I think we can better align 7 00:00:23,320 --> 00:00:26,039 artificial intelligence with a collective human will. 8 00:00:26,839 --> 00:00:29,120 Speaker 1: That's going to sound very strange, but so. 9 00:00:29,239 --> 00:00:33,039 Speaker 2: You could think of like, basically our intelligence is being 10 00:00:33,039 --> 00:00:35,920 divided into roughly three areas. That's sort of like a 11 00:00:36,439 --> 00:00:40,479 like you know, Olympic system, like like like the sort 12 00:00:40,520 --> 00:00:45,119 of instinctual elements that this sort of like the cortex 13 00:00:45,159 --> 00:00:48,359 and the planning part. Then we also have a tertiary layer, 14 00:00:48,520 --> 00:00:55,320 which is only computers and bones applications software that we use, 15 00:00:56,079 --> 00:00:58,359 so that you have a digital tertiari self. 16 00:00:58,399 --> 00:00:59,960 Speaker 1: Basically we were already an android. 17 00:01:00,119 --> 00:01:03,960 Speaker 2: Defictively, I think people feel this when they forget their phone. 18 00:01:05,200 --> 00:01:07,079 Forgetting a phone, leaving a phone behind, it is like 19 00:01:07,200 --> 00:01:10,799 having missing limb syndrome. You're missing your part of your 20 00:01:10,799 --> 00:01:16,239 digital tertiaries. The constraint on human machine symbiosis is bandwidth. 21 00:01:16,599 --> 00:01:18,959 Speaker 1: What is the especially output bandwidth? 22 00:01:19,280 --> 00:01:22,400 Speaker 2: The output bandwidth of a human is less than one 23 00:01:22,400 --> 00:01:24,280 bit per second over the course of a day. 24 00:01:24,439 --> 00:01:26,439 Speaker 1: So if you have eighty six four hundred. 25 00:01:26,159 --> 00:01:29,159 Speaker 2: Seconds in a day, the number of output bits that 26 00:01:29,200 --> 00:01:31,640 you produce. Maybe there's some rare cases where it's above 27 00:01:31,680 --> 00:01:36,920 one bit per second, but very few people produce eighty six. 28 00:01:36,760 --> 00:01:39,120 Speaker 1: Thousand, four hundred output bits. 29 00:01:39,319 --> 00:01:41,760 Speaker 2: So most people like our averaging less than one bit 30 00:01:41,840 --> 00:01:44,480 per second over twenty four hour period. And when we 31 00:01:44,519 --> 00:01:46,879 do speak, they say the number of symbols per second 32 00:01:47,079 --> 00:01:50,319 of speech typing is quite low, especially if it's going 33 00:01:50,319 --> 00:01:52,319 through a phone. Then you just sort of have two 34 00:01:52,359 --> 00:01:55,840 slow moving meatsticks that are trying to type letters on 35 00:01:55,959 --> 00:02:00,359 a phone, so you really have just a few per 36 00:02:00,400 --> 00:02:02,319 second of characters, so that your phone is like a 37 00:02:02,359 --> 00:02:05,640 supercomputer in your hands and it is desperately trying to 38 00:02:05,640 --> 00:02:06,840 figure out what you want to say. 39 00:02:07,400 --> 00:02:10,639 Speaker 3: I'll tell you I've personally experienced that phantom limb syndrome 40 00:02:10,680 --> 00:02:12,560 when I actually can't find my phone. And I hadn't 41 00:02:12,599 --> 00:02:15,400 thought of myself as a cyborg until you challenge. 42 00:02:15,159 --> 00:02:17,560 Speaker 4: Me to think that way. But you're in a room 43 00:02:17,879 --> 00:02:19,120 of folks who've devoted their. 44 00:02:19,000 --> 00:02:22,159 Speaker 3: Lives to neurologic disease, and I must confess to you 45 00:02:22,240 --> 00:02:24,280 that I had never actually thought of the output of 46 00:02:24,280 --> 00:02:26,680 the brain in terms of bits per second. But when 47 00:02:26,719 --> 00:02:29,560 you frame it that way, it makes it really clear 48 00:02:30,000 --> 00:02:32,919 why there may be a broader opportunity to make that 49 00:02:33,000 --> 00:02:35,560 virtual cyborg that we have now with our phone a 50 00:02:35,560 --> 00:02:38,479 little bit more efficient. So that's as a starting point, 51 00:02:38,479 --> 00:02:40,400 what prompted your interest in neuralink. 52 00:02:41,759 --> 00:02:46,000 Speaker 1: Yeah, so basically I thought. 53 00:02:45,840 --> 00:02:50,680 Speaker 2: Okay, in order to have better human AI semiosis, we 54 00:02:50,800 --> 00:02:54,680 must solve the bandwidth problem. Below a certain bandwidth, we 55 00:02:54,759 --> 00:02:58,960 are basically just stationary to a computer and at one 56 00:02:59,000 --> 00:03:01,479 bit per second. Know, that's the very low data rate 57 00:03:01,520 --> 00:03:04,120 when computers are doing brillions of bits per second. 58 00:03:04,240 --> 00:03:06,680 Speaker 3: So when you think about brain machine interface, why did 59 00:03:06,680 --> 00:03:08,319 you select the technical approach you did? 60 00:03:08,479 --> 00:03:09,919 Speaker 4: I know a lot of thoughts gone into that. 61 00:03:11,479 --> 00:03:11,719 Speaker 1: Yeah. 62 00:03:11,759 --> 00:03:14,039 Speaker 2: So if you say, like, okay, we need to have 63 00:03:14,240 --> 00:03:18,439 ultimately a million bits per second or a billion bits 64 00:03:18,560 --> 00:03:21,840 per second to gig a bit per second interface, then 65 00:03:22,680 --> 00:03:24,560 that means you really you can't. 66 00:03:25,159 --> 00:03:26,680 Speaker 1: You need an implant and. 67 00:03:26,840 --> 00:03:29,719 Speaker 2: Ultimately will need to replace the skull and it's going 68 00:03:29,800 --> 00:03:31,599 to be a zillion wire. I mean, this is some 69 00:03:31,639 --> 00:03:35,400 sci fi, bizarre sci fi stuff, and I'm not this 70 00:03:35,560 --> 00:03:37,400 is certainly optional. 71 00:03:37,479 --> 00:03:41,159 Speaker 3: Plan mandatory replacement of my skull whatever problem. 72 00:03:41,319 --> 00:03:44,039 Speaker 1: Mandatory chip and brain is not what we're saying, Yeah, 73 00:03:44,039 --> 00:03:44,759 for sure. 74 00:03:44,919 --> 00:03:47,599 Speaker 2: But at some point you're you say like, okay, how 75 00:03:47,599 --> 00:03:52,879 many electrodes are needed in order to interface with have 76 00:03:52,960 --> 00:03:54,039 a whole brain interface? 77 00:03:54,759 --> 00:03:57,319 Speaker 3: Yeah, you know, I've heard you mentioned that larger goal 78 00:03:57,400 --> 00:04:00,560 of whole brain interface. One thing that's really me by 79 00:04:00,680 --> 00:04:03,840 the approach that's been taken is, I think as a resurgence, 80 00:04:03,879 --> 00:04:06,840 we often contemplate the natural history of the disease and 81 00:04:06,879 --> 00:04:10,520 competing risk and benefit in neuralink as a company has 82 00:04:10,560 --> 00:04:13,960 started with folks who have als and spinal cord injury. 83 00:04:14,080 --> 00:04:16,360 These kind of first steps in terms of technical approach. 84 00:04:16,560 --> 00:04:19,000 So we'd love to hear a little bit more about that. 85 00:04:19,959 --> 00:04:20,120 Speaker 4: Yeah. 86 00:04:20,160 --> 00:04:25,519 Speaker 2: Absolutely, long tem goal, like I said, is mitigating civilizational 87 00:04:25,639 --> 00:04:30,839 risk associated with a divergence of biological and digital intelligence. 88 00:04:30,879 --> 00:04:32,879 That's the long tim goal. Obviously. Then you've got to 89 00:04:32,879 --> 00:04:34,079 parse that, h L, Well, what are we going to 90 00:04:34,120 --> 00:04:38,959 do tomorrow? Yeah, So the starting point with the first 91 00:04:39,040 --> 00:04:42,879 new link device is a thousand electrodes, and with just 92 00:04:42,920 --> 00:04:45,120 one hundred of those electrodes are active, if it takes 93 00:04:45,160 --> 00:04:47,879 up our first few patients, you know, we're also having 94 00:04:47,920 --> 00:04:50,480 world records. Admitically, these are world records that are pretty low, 95 00:04:50,720 --> 00:04:52,079 but we're getting around. 96 00:04:51,839 --> 00:04:54,600 Speaker 1: Ten ves per second, and that does a path to 97 00:04:55,519 --> 00:04:56,439 one thousand us per. 98 00:04:56,360 --> 00:04:59,519 Speaker 2: Second, which would be literally one hundred times more than 99 00:04:59,680 --> 00:05:03,360 the next record. So we want to do the implants 100 00:05:03,480 --> 00:05:06,519 in where there's the highest gain and the least risk. 101 00:05:06,839 --> 00:05:09,800 So we call the first implant to telepathy, which really 102 00:05:09,839 --> 00:05:15,319 just interfacing with the motor cortex, and it's basically looking 103 00:05:15,319 --> 00:05:17,439 at signals as though somebody moved their own and just 104 00:05:17,480 --> 00:05:20,600 reading that signal and then sending that signal to the 105 00:05:20,639 --> 00:05:23,959 patient's phone or computer so they can then move the 106 00:05:24,000 --> 00:05:27,040 cursor around just by thinking. If you will have seen 107 00:05:27,079 --> 00:05:29,759 the videos of Noland, that's pretty impressive what he can do. 108 00:05:30,040 --> 00:05:32,480 In fact, shortly after getting the implant, he spent all 109 00:05:32,560 --> 00:05:34,439 night playing video games just by thinking. 110 00:05:35,160 --> 00:05:37,720 Speaker 3: Yeah, and those are the records you're talking about in 111 00:05:37,759 --> 00:05:41,120 those first two prime patients, where you're able to extract 112 00:05:41,160 --> 00:05:45,360 signals from their brain at record bits per second and 113 00:05:45,839 --> 00:05:48,160 enable them to work in the world. 114 00:05:48,360 --> 00:05:50,959 Speaker 4: Is those of us who lose their phone would use today. 115 00:05:51,879 --> 00:05:52,199 Speaker 1: Yeah. 116 00:05:52,240 --> 00:05:54,759 Speaker 2: Absolutely, And I think we'll get to the point pretty 117 00:05:54,800 --> 00:05:58,959 quickly where someone with a neuralink implant will outperform somebody 118 00:05:58,959 --> 00:06:00,839 who's using their hands play a video game. 119 00:06:01,839 --> 00:06:04,199 Speaker 4: What do you think the timeline for that is? We 120 00:06:04,279 --> 00:06:05,319 won't hold you to it. 121 00:06:05,839 --> 00:06:08,639 Speaker 2: Sure, I mean I do have a habit of being 122 00:06:08,639 --> 00:06:11,800 optimistic with respect to timelines. But if I wasn't optimistic, 123 00:06:11,839 --> 00:06:13,199 I wouldn't be starting these companies. 124 00:06:13,199 --> 00:06:15,040 Speaker 4: Probably, Yeah, that's fair. 125 00:06:16,639 --> 00:06:19,160 Speaker 1: But I think given that we're already. 126 00:06:20,360 --> 00:06:22,439 Speaker 2: Pretty much out a point where we're pretty close to 127 00:06:22,680 --> 00:06:26,240 on par with the video game. Basically you can play 128 00:06:26,279 --> 00:06:30,959 a video game at a comparable competent level to someone 129 00:06:31,040 --> 00:06:35,319 with hands. I think with our second generation device, which 130 00:06:35,360 --> 00:06:37,639 we'll have three thousand electrodes, and we'll get a lot 131 00:06:37,680 --> 00:06:41,439 better at placing those electrodes so only it's only one 132 00:06:41,480 --> 00:06:45,120 hundred electrodes being effective, we'll both improve the yield and 133 00:06:45,160 --> 00:06:48,360 will increase the number of electrodes. So we'll go from 134 00:06:48,399 --> 00:06:51,279 say one hundred electrodes that are reading to I don't know, 135 00:06:51,439 --> 00:06:54,120 out of three thousand electrodes, maybe fifteen and a half, 136 00:06:54,240 --> 00:06:56,800 So like fifteen hundred are reading. So at that point 137 00:06:56,800 --> 00:06:59,560 the data rate is far in excess of what someone 138 00:07:00,160 --> 00:07:02,399 video game with their hands could do, and we can 139 00:07:02,720 --> 00:07:05,079 reduce the latency the moment you think of a move, 140 00:07:05,839 --> 00:07:09,519 it happens instantly on the computer, as opposed to for 141 00:07:10,000 --> 00:07:12,279 you know, currently, if you're a human play a video game, 142 00:07:12,360 --> 00:07:13,959 you have to move your hand so that that's like 143 00:07:13,959 --> 00:07:15,480 you've got to send signals to the muscles. 144 00:07:15,519 --> 00:07:16,519 Speaker 1: The muscles have to move. 145 00:07:17,839 --> 00:07:19,839 Speaker 2: Your finger takes a certain amount of time to move, 146 00:07:20,560 --> 00:07:22,519 so you've got to be you basically got to move 147 00:07:22,560 --> 00:07:25,040 the meat puppet. If you don't have to move actuate 148 00:07:25,160 --> 00:07:27,240 the muscles in your hand or your finger's going to 149 00:07:27,279 --> 00:07:29,680 move at a certain rate and set like milimeters per second. 150 00:07:30,240 --> 00:07:31,680 But if you don't have to do any of that, 151 00:07:31,839 --> 00:07:33,800 you can literally think it immediately with no latency. 152 00:07:33,839 --> 00:07:36,120 Speaker 1: You'll outperform someone who has to use a hands. 153 00:07:36,480 --> 00:07:38,639 Speaker 3: Yeah, you know, I think as surgeons we really take 154 00:07:38,680 --> 00:07:40,839 pride in being efficient and using your hand. But when 155 00:07:40,839 --> 00:07:42,959 you're a reductionist like that, it actually makes me feel 156 00:07:42,959 --> 00:07:45,839 like I'm actually not particularly efficient. You could just if 157 00:07:45,879 --> 00:07:47,959 you just think and do it, I think I'd probably 158 00:07:47,959 --> 00:07:49,639 get a lot I'll get a little lot more done. 159 00:07:49,839 --> 00:07:50,000 Speaker 1: You know. 160 00:07:50,120 --> 00:07:52,199 Speaker 3: One of the things that struck me in terms of 161 00:07:52,240 --> 00:07:55,519 the technical approach is obviously you have the implant and 162 00:07:55,519 --> 00:07:58,879 then you're extracting those signals and have a recording algorithm, 163 00:07:59,079 --> 00:08:02,319 and then you're actually affecting an action and you know, 164 00:08:02,360 --> 00:08:04,959 in one of the patients you actually had a lead 165 00:08:05,040 --> 00:08:07,519 or traction. But then we're able to tune the recording 166 00:08:07,519 --> 00:08:10,120 algorithm to actually recover that function. 167 00:08:10,519 --> 00:08:12,079 Speaker 4: Could you maybe say a little bit about. 168 00:08:11,920 --> 00:08:14,680 Speaker 3: That kind of vertically integrated approach and how that's going 169 00:08:14,759 --> 00:08:15,959 to let you scale a little bit? 170 00:08:17,199 --> 00:08:20,720 Speaker 2: Sure? Well, since you know, really none of this stuff 171 00:08:20,759 --> 00:08:24,160 existed before, we had to design and build everything from scratch, 172 00:08:24,759 --> 00:08:29,360 and I mean it's basically like having an apple watcher 173 00:08:29,439 --> 00:08:31,920 a fit, but that replaces a piece of skull. And 174 00:08:31,920 --> 00:08:36,440 then you've got these electrodes, very very fine electrodes that 175 00:08:36,720 --> 00:08:38,759 are implanted with a surgical robot. 176 00:08:38,960 --> 00:08:41,639 Speaker 3: I mean, we can share a little bit about the robot, 177 00:08:41,799 --> 00:08:42,720 the R one robot. 178 00:08:42,720 --> 00:08:44,519 Speaker 4: That two is to implant the threads. 179 00:08:45,360 --> 00:08:47,360 Speaker 2: Yeah, so the threads are really too small to be 180 00:08:47,399 --> 00:08:51,120 manipulated by hand, and they need to be placed with 181 00:08:51,200 --> 00:08:54,440 extreme precision, very quickly. The brain is moving all the 182 00:08:54,440 --> 00:08:57,399 time due to breathing and heartbeat or just not just 183 00:08:57,440 --> 00:09:00,720 sitting there. It's like a pulsing thing, and you're trying 184 00:09:00,759 --> 00:09:03,879 to get an electrode to a specific depth, while this, 185 00:09:05,039 --> 00:09:08,840 you know, jello balloon is just moving around all over 186 00:09:08,879 --> 00:09:12,399 the place. So it's it's it's kind of an impossible, 187 00:09:12,559 --> 00:09:14,279 really an impossible thing to do by hand. These these 188 00:09:14,320 --> 00:09:17,480 spades are just too tiny and the level of precision 189 00:09:17,720 --> 00:09:21,759 required is beyond what people can do. I maybe liken 190 00:09:21,799 --> 00:09:26,440 it to be being similar to computer controlled machining or 191 00:09:27,159 --> 00:09:29,679 three D metal printing with we've better a lays of 192 00:09:29,759 --> 00:09:32,279 welding tiny bits of metal dust. It's just there's just 193 00:09:32,320 --> 00:09:34,679 no way that humans just do not have the level 194 00:09:34,679 --> 00:09:38,639 of precision necessary to implant the electrodes, you know, to 195 00:09:39,000 --> 00:09:42,320 fractions of a millimeter of x y z position. 196 00:09:42,679 --> 00:09:45,519 Speaker 3: Well, you know what's interesting obviously is a group of surgeons, 197 00:09:45,600 --> 00:09:48,840 many of us to varying stages, have incorporated robotics into 198 00:09:48,879 --> 00:09:53,519 our practice. When you hear a precision exceeding human capacity, 199 00:09:53,759 --> 00:09:55,960 do you think is this going to be a disruption 200 00:09:56,080 --> 00:09:59,080 or is this an augmentation to what surgeons do? And 201 00:09:59,120 --> 00:10:01,840 I know you have some thoughts around that, and there's 202 00:10:01,840 --> 00:10:04,120 maybe some analogies and ophthalmology, so it would love to 203 00:10:04,120 --> 00:10:05,159 hear that perspective. 204 00:10:06,200 --> 00:10:09,559 Speaker 2: Yeah, So I think the ophalmology analogy is the right 205 00:10:09,559 --> 00:10:14,320 one with laseric and ophthalmologists will oversee perhaps half a 206 00:10:14,360 --> 00:10:17,559 dozen or a dozen laser machines and to make sure 207 00:10:17,559 --> 00:10:21,600 the machine is is the patient getting the right operation 208 00:10:22,240 --> 00:10:25,879 in the correct eye, and is the is the machine 209 00:10:25,879 --> 00:10:29,679 operating properly. But thereafter the you know, patils in the 210 00:10:29,759 --> 00:10:33,960 laser chair and the robots going to basically laser rival. 211 00:10:34,480 --> 00:10:37,200 And now this is much better than someone getting a 212 00:10:37,240 --> 00:10:41,200 hand laser and laid hand lasering arrival, which would have 213 00:10:41,320 --> 00:10:44,279 varying results. I think it will be something similar to LASIC, 214 00:10:44,279 --> 00:10:47,240 where you perhaps a neurosurgeon overseeing half a dozen or 215 00:10:47,240 --> 00:10:50,879 a dozen of the neuralink robots that are doing the 216 00:10:51,159 --> 00:10:53,279 implants and just obviously making sure it's the right implant 217 00:10:53,279 --> 00:10:55,720 and the right location for the right purpose, and that 218 00:10:55,759 --> 00:10:58,519 everything's okay with the patient. So it would be like 219 00:10:58,720 --> 00:11:03,159 a massive amplification, I think, and it's kind of necessary 220 00:11:03,159 --> 00:11:06,039 that it'd be a massive appiplication because there's simply not 221 00:11:06,200 --> 00:11:07,600 enough neurosarchence to do this whole by hand. 222 00:11:07,639 --> 00:11:08,799 Speaker 1: It's like physically impossible. 223 00:11:09,200 --> 00:11:13,080 Speaker 2: Yeah, because we're talking about ultimately doing tens of millions 224 00:11:13,120 --> 00:11:16,000 of these things, like maybe this eight billion people in 225 00:11:16,000 --> 00:11:16,600 the world. 226 00:11:16,840 --> 00:11:18,120 Speaker 1: I don't know, maybe at. 227 00:11:18,120 --> 00:11:20,320 Speaker 2: Least a few billion are going to want this, maybe 228 00:11:20,360 --> 00:11:23,720 more so, then how do you get billions of devices 229 00:11:24,240 --> 00:11:25,399 unless you got the robots. 230 00:11:25,480 --> 00:11:26,279 Speaker 1: It's not happening. 231 00:11:26,480 --> 00:11:28,960 Speaker 3: I've heard you frame the introduction of the robot is 232 00:11:29,000 --> 00:11:31,759 not just a precision issue, but an interest of workforce 233 00:11:31,799 --> 00:11:34,320 and scale. And there's obviously a little over three thousand 234 00:11:34,440 --> 00:11:37,000 of us nationally, so that would be a little bit challenging. 235 00:11:37,360 --> 00:11:39,559 Can you share a little bit in this early journey 236 00:11:39,679 --> 00:11:42,919 with BCI what some of the challenges have been, what 237 00:11:42,960 --> 00:11:46,360 you've encountered technically, I know, a biological environment, the saltwater 238 00:11:46,399 --> 00:11:47,559 problem is very hostile. 239 00:11:47,960 --> 00:11:49,279 Speaker 4: Things with energy treaties. 240 00:11:50,000 --> 00:11:52,360 Speaker 3: Would love to hear your thoughts on that and how 241 00:11:52,360 --> 00:11:53,840 your team's taking those things on. 242 00:11:55,480 --> 00:11:59,960 Speaker 2: Yeah, I mean, yeah, as everyone obviously talking to people 243 00:12:00,039 --> 00:12:02,600 that know a lot more about the brain and that 244 00:12:03,039 --> 00:12:07,639 than I do, but I accidentally come to understand more 245 00:12:07,639 --> 00:12:09,840 than most people. The challenge is You've got a device 246 00:12:09,879 --> 00:12:12,360 that's going to live there for years. It's an electrical 247 00:12:12,399 --> 00:12:17,320 device that has to transmit radio essentially, you know, it 248 00:12:17,399 --> 00:12:19,840 has to transmit photons to your computer. 249 00:12:20,639 --> 00:12:23,200 Speaker 1: It's subcutaneous, it's got to be charged. 250 00:12:23,679 --> 00:12:26,559 Speaker 2: It's got electrodes that are reading and writing, so it's 251 00:12:26,600 --> 00:12:28,720 not like it can't just be electrically isolated. 252 00:12:28,919 --> 00:12:31,240 Speaker 1: In fact, you're fighting two things. You want. 253 00:12:31,279 --> 00:12:34,200 Speaker 2: You really are desperately trying to read these neurons, but 254 00:12:34,240 --> 00:12:37,159 you also don't want to be corroded. So it's like 255 00:12:37,399 --> 00:12:40,399 the very difficult thing to have just the minimum amount 256 00:12:40,399 --> 00:12:44,320 of insulation necessary to not be corroded, but not be 257 00:12:44,440 --> 00:12:47,279 so insulated that you can't hear the neurons. So there's 258 00:12:47,320 --> 00:12:50,279 a very challenging materials problem with our latest electrodes that 259 00:12:50,320 --> 00:12:53,559 will be silicon carbide coded, but even the silicon cartibide is. 260 00:12:53,519 --> 00:12:54,799 Speaker 1: A very difficult material to work with. 261 00:12:54,919 --> 00:12:56,600 Speaker 2: It's awesome, but it's very difficult, and you've got to 262 00:12:56,600 --> 00:13:00,759 make sure the coding is extremely precise. It's be you know, 263 00:13:00,840 --> 00:13:03,360 canvy tooth thin or tooth thick anywhere. It's going to 264 00:13:03,440 --> 00:13:08,799 be very evenly you applied to the threads. So it's 265 00:13:09,399 --> 00:13:13,720 the cher number of iterations necessary to actually have this 266 00:13:13,799 --> 00:13:18,679 device be medically sealed and survived in the body and 267 00:13:18,759 --> 00:13:21,879 not fail in some way, and then have to be 268 00:13:21,879 --> 00:13:24,960 able to transmit to your phono computer at a high 269 00:13:25,000 --> 00:13:29,080 data rate without burning down the battery is very difficult. 270 00:13:29,120 --> 00:13:32,159 I'd say there's many many technical challenges in that. So 271 00:13:32,320 --> 00:13:35,919 I mean I do have slightly criviolized by saying it's 272 00:13:35,919 --> 00:13:37,799 sort of like a fitbit or an Apple Watch in 273 00:13:37,840 --> 00:13:39,879 your brain. But if you actually put those things in 274 00:13:39,919 --> 00:13:42,600 your brain, neither your brain nor the Apple watch or 275 00:13:42,600 --> 00:13:43,600 fitbot would be happy. 276 00:13:43,960 --> 00:13:46,600 Speaker 3: So this feels like the right place to ask. I 277 00:13:46,600 --> 00:13:49,519 think one of the more interesting questions we received. So 278 00:13:49,919 --> 00:13:52,879 as someone who's in a position of authority to comment 279 00:13:52,919 --> 00:13:55,639 on both, can you settle the age old question, what's 280 00:13:55,679 --> 00:13:58,519 actually more difficult brain surgery or rocket science? 281 00:14:00,679 --> 00:14:02,440 Speaker 2: Well, both of my challenging. It's bizarre that I'm in 282 00:14:02,480 --> 00:14:05,000 bold in both. I mean, I think there are similar 283 00:14:05,799 --> 00:14:07,399 magnitude of difficulty. 284 00:14:07,080 --> 00:14:09,120 Speaker 4: Especially story the story checks out. 285 00:14:10,200 --> 00:14:12,519 Speaker 2: Yes, I think nobody's out there thinking, you know, what's 286 00:14:12,559 --> 00:14:14,159 easy brain surgery and rockets. 287 00:14:14,279 --> 00:14:17,519 Speaker 4: Okay, perfect, Thanks, thanks for backing us up. We appreciate it. 288 00:14:17,559 --> 00:14:20,120 Speaker 2: Yeah, and unmercent, Now that's a legit. Rain surgery is 289 00:14:20,159 --> 00:14:22,360 super hard, and rockets it's super hard. And there's a 290 00:14:22,399 --> 00:14:25,639 reason that there are idiomatic expressions. This is no accident, 291 00:14:26,000 --> 00:14:28,919 especially as you try to scale the electrodes number of electrodes, 292 00:14:29,720 --> 00:14:33,200 and I don't we don't know how to say. 293 00:14:33,039 --> 00:14:34,759 Speaker 1: Like ultimately get to say, how do we do a 294 00:14:34,759 --> 00:14:35,759 million electrodes? 295 00:14:35,799 --> 00:14:37,559 Speaker 2: This is we don't know how to do that yet 296 00:14:37,639 --> 00:14:41,320 except that hopefully it is physically possible. If you want 297 00:14:41,360 --> 00:14:43,559 to have a hind Man with a whole brain interface, 298 00:14:43,879 --> 00:14:47,200 then I think probably the right automagtude is something like 299 00:14:47,200 --> 00:14:49,120 a million electrode and that that still has a very 300 00:14:49,159 --> 00:14:53,639 high ratio of neurons to electrodes, So that means you've 301 00:14:53,679 --> 00:14:57,159 got to read you try to add like any given 302 00:14:57,399 --> 00:15:00,960 electrode has to be able to read neurons from you 303 00:15:00,960 --> 00:15:04,679 know several like I don't know one hundred or one 304 00:15:04,679 --> 00:15:08,080 thousand neurons. So if you can do if you've got 305 00:15:08,080 --> 00:15:11,240 a million electrodes and each electroid can read a thousand neurons, 306 00:15:11,240 --> 00:15:13,120 so you've got access to a billion neurons. 307 00:15:13,679 --> 00:15:17,759 Speaker 3: Well, the goal with a whole brain interface is this 308 00:15:18,679 --> 00:15:21,840 potential for long term augmentation or symbiosis. But you know, 309 00:15:21,879 --> 00:15:24,120 in the more immediate term, something that we think a 310 00:15:24,120 --> 00:15:27,159 lot about as surgeons is how is technology can allow 311 00:15:27,240 --> 00:15:29,879 us to treat problems that we aren't able to treat 312 00:15:29,960 --> 00:15:35,200 now and there's this whole family of diseases, psychiatric conditions, 313 00:15:35,879 --> 00:15:40,000 neurodevelopmental conditions, you know, folks who are neurodiverse and nerd 314 00:15:40,000 --> 00:15:43,440 degenerative conditions like Alzheimer's and so as we get a 315 00:15:43,440 --> 00:15:46,399 better picture of not just the structure of the brain, 316 00:15:46,440 --> 00:15:48,320 but you know, for lack of better term, the music 317 00:15:48,360 --> 00:15:48,840 of the brain. 318 00:15:49,679 --> 00:15:52,080 Speaker 4: Do you see those as intermediate steps? Would love to 319 00:15:52,080 --> 00:15:53,120 hear your perspective on it. 320 00:15:53,320 --> 00:15:55,399 Speaker 2: Yeah, I mean, I think we should be able to 321 00:15:55,799 --> 00:15:59,120 solve any problem over time that is a result of 322 00:16:00,320 --> 00:16:02,080 you know, like if you think of the brain like 323 00:16:02,080 --> 00:16:04,879 a computer effect, like a circuit board or something like that, 324 00:16:04,919 --> 00:16:07,519 you can say, like if you're given a circuit board 325 00:16:07,559 --> 00:16:10,840 and there were some short circuits or some circuits that 326 00:16:10,840 --> 00:16:13,159 should be there but aren't there. If there are any 327 00:16:13,399 --> 00:16:17,799 circuits that shouldn't be there, and and some that that 328 00:16:17,879 --> 00:16:20,559 are there but shouldn't, we can fix those. So basically, 329 00:16:20,600 --> 00:16:24,159 if if if it's it's like fixing a circuit board. Now, now, 330 00:16:24,159 --> 00:16:26,320 if the circuit board is all melted, it's going to 331 00:16:26,360 --> 00:16:28,840 be hard to fix a melted circuit board. You can 332 00:16:28,840 --> 00:16:31,879 fix the circ board with a few issues, but you 333 00:16:31,919 --> 00:16:34,519 can't fix it if it's been melted. But the vast 334 00:16:34,600 --> 00:16:39,279 majority of diseases or brain issues I think are fixable 335 00:16:39,399 --> 00:16:42,240 with your within your rolling device. It's it's a it's 336 00:16:42,240 --> 00:16:47,480 a fine grained means of reading and writing electrical signals 337 00:16:47,480 --> 00:16:50,559 in the brain at a road with high precision. And 338 00:16:50,639 --> 00:16:53,120 so that means like if there's an electrical storm, some 339 00:16:53,200 --> 00:16:55,960 kind of apilepasy or something, you can interrupt that storm 340 00:16:56,480 --> 00:16:59,120 if you can, if there are a set of signals 341 00:16:59,639 --> 00:17:02,600 to like in the case of blindness, that if somebody's 342 00:17:02,679 --> 00:17:05,839 lost their optic noble both eyes, you can still stimulate 343 00:17:05,920 --> 00:17:08,359 the visual cortex. 344 00:17:08,240 --> 00:17:10,960 Speaker 1: Basically anything that is a function of signals in er out. 345 00:17:11,000 --> 00:17:12,799 Speaker 2: If that is the nature of the problem, it can 346 00:17:12,799 --> 00:17:14,720 be fixed ultimately with a neuraling device. 347 00:17:14,880 --> 00:17:15,720 Speaker 4: Yeah, well I know you. 348 00:17:15,960 --> 00:17:19,880 Speaker 3: Neuralink just got FDA breakthrough designation for blindsight week and 349 00:17:19,920 --> 00:17:22,759 a half before this meeting. One thing that I heard 350 00:17:22,799 --> 00:17:25,119 you talk about that I thought was so interesting when 351 00:17:25,160 --> 00:17:28,759 I think about neurodiversity or NeuroD degenerative disease, is this 352 00:17:28,880 --> 00:17:32,640 idea of imagine, if someone of the intellective of Stephen 353 00:17:32,680 --> 00:17:36,240 Hawking was able to communicate more efficiently, how much more 354 00:17:36,240 --> 00:17:39,440 would society have benefited from those insights, and so when 355 00:17:39,440 --> 00:17:42,960 I think of people with neurodiverse conditions, I always think 356 00:17:43,000 --> 00:17:46,920 that they have this amazing potential to potentially be unlocked, 357 00:17:47,079 --> 00:17:50,000 and maybe this implant could be a digital bridge to that. 358 00:17:51,279 --> 00:17:53,559 Speaker 2: Absolutely, so I think it can help a lot of people, 359 00:17:53,839 --> 00:17:57,240 like really ultimately help tens millions of people, maybe one 360 00:17:57,279 --> 00:17:59,960 hundreds of millions of people. I should say, also this potential, 361 00:18:00,119 --> 00:18:02,319 we'll go beyond the brain to like if somebody's got 362 00:18:02,400 --> 00:18:05,799 a sort of spinal cord injury, that being able to 363 00:18:05,799 --> 00:18:07,440 transmit the signals, so. 364 00:18:07,640 --> 00:18:10,559 Speaker 1: You know, like the ideal. I think what most people 365 00:18:10,599 --> 00:18:11,160 that have blusted the. 366 00:18:11,119 --> 00:18:13,000 Speaker 2: Connection between their brain and their body would like is 367 00:18:13,039 --> 00:18:16,200 to reanimate their body. Sure you know there are there 368 00:18:16,240 --> 00:18:18,680 are there are some approximations of that where you can 369 00:18:18,720 --> 00:18:21,880 animate see a robot suit or a robot arm or 370 00:18:21,920 --> 00:18:24,319 something like that. But if I think most people will 371 00:18:24,359 --> 00:18:25,920 be asking them, like, what would you prefer, I'd like 372 00:18:25,960 --> 00:18:28,799 my body to work again if provided the neurons are 373 00:18:28,799 --> 00:18:31,640 still kind of there, It's it's simply physically possible to 374 00:18:32,119 --> 00:18:35,319 shunt the signals, frown the motor cortex past the point 375 00:18:35,440 --> 00:18:38,319 where the damage has occurred to the neurons that then 376 00:18:38,359 --> 00:18:40,599 interface with your muscles and your. 377 00:18:40,480 --> 00:18:41,279 Speaker 1: Arms and legs. 378 00:18:41,720 --> 00:18:43,960 Speaker 2: If you think of it just like an electrical and 379 00:18:44,000 --> 00:18:48,279 communication system, like if you severed some ethernet cables, what 380 00:18:48,279 --> 00:18:48,680 would you do? 381 00:18:48,759 --> 00:18:49,839 Speaker 1: Well, you bridge the signal? 382 00:18:50,319 --> 00:18:53,720 Speaker 2: Okay, great, that the same thing can be done with 383 00:18:53,839 --> 00:18:57,079 the human body is bridge the electrical signals and the 384 00:18:57,119 --> 00:19:01,319 communication signals. So you've got sensors and actuators and the 385 00:19:01,359 --> 00:19:04,839 signals the bi directional signals for sensors and actuators are 386 00:19:04,880 --> 00:19:07,079 being interrupted, and I said, if you shunt the signals, 387 00:19:07,480 --> 00:19:09,079 you will be able to renovate the body. 388 00:19:09,640 --> 00:19:11,880 Speaker 3: One other issue that comes up with implants that you 389 00:19:11,880 --> 00:19:15,440 were mentioning our iPhones when you're committing someone to an implant, 390 00:19:15,480 --> 00:19:19,440 obviously there's a whole issue around upgrades or the cycle 391 00:19:19,480 --> 00:19:22,079 time or iteration and technology. So you can maybe say 392 00:19:22,079 --> 00:19:25,119 a little bit about reversibility and how we should be 393 00:19:25,119 --> 00:19:27,759 thinking about these things as we enter an era where 394 00:19:27,880 --> 00:19:29,839 bci'll become more widespread. 395 00:19:31,079 --> 00:19:33,880 Speaker 2: Yeah, so we do think upgrades are pretty important, just 396 00:19:33,920 --> 00:19:35,839 as you would not once an iPhone one stuck in 397 00:19:35,880 --> 00:19:39,000 your head when there's an iPhone sixteen or whatever version 398 00:19:39,039 --> 00:19:40,559 iPhone and are on these days but I think it's 399 00:19:40,599 --> 00:19:41,720 like six it's pretty high. 400 00:19:41,960 --> 00:19:44,039 Speaker 3: I've lost track of what not for they're on I 401 00:19:44,079 --> 00:19:45,759 think you're I think you're up to date on the sixteen. 402 00:19:45,759 --> 00:19:48,119 Speaker 1: I think, okay, you know so so. 403 00:19:48,519 --> 00:19:50,480 Speaker 2: But I mean, now there's this, there's some sort of 404 00:19:50,519 --> 00:19:53,880 logarithmic you know, there's like as kind of goes by, 405 00:19:53,960 --> 00:19:57,400 the incremental gains from one, say iPhone to the next 406 00:19:57,400 --> 00:19:58,400 are are less. 407 00:19:58,480 --> 00:20:01,400 Speaker 1: It's kind of logarithmic gain, it would appear. 408 00:20:02,079 --> 00:20:06,000 Speaker 2: But that means that well, like I say, the first 409 00:20:06,119 --> 00:20:08,599 five or six versions, there are actually big jumps, and 410 00:20:08,640 --> 00:20:10,720 certainly that would be that is the case with your link. 411 00:20:10,839 --> 00:20:14,279 Speaker 1: So if somebody has say production design. 412 00:20:14,079 --> 00:20:16,599 Speaker 2: Version one, I think five years later they'll one to 413 00:20:16,599 --> 00:20:18,920 have production design version three or four. And so we 414 00:20:19,000 --> 00:20:21,880 designed the implant such that it can be removed but 415 00:20:22,000 --> 00:20:26,200 with hopefully minimal strip damage to the area, so that 416 00:20:26,240 --> 00:20:29,599 you can then then replace it with another one. And 417 00:20:29,640 --> 00:20:33,359 we have with in our animal studies, we've done I 418 00:20:33,359 --> 00:20:38,279 think three implants, and the third implant still worked quite well, meaning. 419 00:20:38,039 --> 00:20:40,279 Speaker 4: You've replaced the implant three times in the same. 420 00:20:40,200 --> 00:20:42,920 Speaker 2: Place three times. Yeah, and the third one was still 421 00:20:43,000 --> 00:20:44,440 working was working great. 422 00:20:45,359 --> 00:20:48,640 Speaker 3: So we've talked about the robot addressing the workforce problem. 423 00:20:48,720 --> 00:20:51,279 Speaker 4: We've talked about interchangeability. 424 00:20:51,400 --> 00:20:53,960 Speaker 3: You know, a lot of what your vision involves is 425 00:20:54,000 --> 00:20:56,839 being high performing but also affordable, so it would be 426 00:20:56,839 --> 00:21:00,440 accessible to people. How do you see bridging that gap? 427 00:21:02,200 --> 00:21:07,000 Speaker 2: Yeah, So the device itself in volume should should not 428 00:21:07,079 --> 00:21:09,839 be super expensive. I mean hopefully it's like, I don't know, 429 00:21:09,880 --> 00:21:14,240 five to ten thousand dollars and very high volume. It 430 00:21:14,240 --> 00:21:16,839 should sought to approximate the cost of an Apple watch 431 00:21:16,920 --> 00:21:20,319 or a phone, so maybe it's a thousand or two 432 00:21:20,319 --> 00:21:24,160 thousand dollars something like that. And then the if it's 433 00:21:24,200 --> 00:21:27,160 implanted with a robot, then that that surgical procedure should 434 00:21:27,240 --> 00:21:30,119 be fast. Like we do have a game plan for 435 00:21:30,160 --> 00:21:32,599 what I call this six hundred second surgery. So ten 436 00:21:32,680 --> 00:21:34,599 minutes you sit in the chair, in ten minutes the 437 00:21:34,640 --> 00:21:38,559 data you have an implant, and we're not violating physics. 438 00:21:38,599 --> 00:21:41,200 So it I mean just just has with laser. You know, 439 00:21:41,359 --> 00:21:43,160 it goes in a laser to a whole munch of 440 00:21:43,160 --> 00:21:46,680 things to rival. Now you'd have to automate basically everything here. 441 00:21:47,160 --> 00:21:50,359 But if you break it down second by second, date 442 00:21:50,480 --> 00:21:53,880 is possible to have a six hundred second or ten 443 00:21:53,920 --> 00:21:56,759 minute surgery. And so at that point, if it's being 444 00:21:56,759 --> 00:21:59,240 done by a robot and it's the whole thing takes 445 00:21:59,240 --> 00:22:03,000 ten minutes, I think it probably that the whole thing, 446 00:22:03,759 --> 00:22:06,440 all inclusive, ends up being you know, on the order 447 00:22:06,440 --> 00:22:08,680 of five thousand dollars maybe similar to Lasic. 448 00:22:09,559 --> 00:22:14,119 Speaker 3: You invoked physics, And one interesting insight that I gained 449 00:22:14,160 --> 00:22:16,839 in our time together is this idea that there's often 450 00:22:17,200 --> 00:22:18,279 a debate about the. 451 00:22:18,240 --> 00:22:20,200 Speaker 4: Possible and what's possible and what's not. 452 00:22:20,880 --> 00:22:25,039 Speaker 3: And I know you have the perspective that that shouldn't 453 00:22:25,079 --> 00:22:30,599 really be subject to debate, because if something's impossible, it's 454 00:22:30,640 --> 00:22:32,880 because it's a function of physics, and if not, then 455 00:22:32,880 --> 00:22:34,240 it is and you just have. 456 00:22:34,200 --> 00:22:34,799 Speaker 1: To figure it out. 457 00:22:36,119 --> 00:22:38,680 Speaker 2: If something like if you're breaking conservation of energy or 458 00:22:38,680 --> 00:22:42,039 momentum or charge or something like that, then you either 459 00:22:42,200 --> 00:22:44,559 have a Nobel prize or you're wrong, and most likely 460 00:22:44,640 --> 00:22:47,680 you're wrong. But provided you're not sort of trying to 461 00:22:47,680 --> 00:22:50,400 break the sound barrier or something like that, like you're 462 00:22:50,400 --> 00:22:51,079 not moving. 463 00:22:50,799 --> 00:22:53,480 Speaker 1: That fast that then you should come visit. Okay, that's 464 00:22:53,519 --> 00:22:56,440 probably going to be bad for the brain. If it's 465 00:22:56,519 --> 00:22:57,240 going super. 466 00:22:57,039 --> 00:22:58,960 Speaker 4: Sadic, that actually is starting to make a lot of sense. 467 00:22:59,359 --> 00:23:02,559 Speaker 2: Yeah. Yeah, but provided you're still subsonic and you're not 468 00:23:02,599 --> 00:23:06,279 just doing things so fast that it causes physical disturbances. 469 00:23:06,319 --> 00:23:09,279 That then you can get things don very quickly. Basically 470 00:23:09,400 --> 00:23:10,880 if you look at the things at a bind grain 471 00:23:10,960 --> 00:23:14,000 level and say, well, what is the size of the 472 00:23:14,559 --> 00:23:17,559 voltage difference that you're trying to detect in a neuron 473 00:23:17,599 --> 00:23:20,119 and how so that for like, how far away from. 474 00:23:20,000 --> 00:23:23,480 Speaker 1: An electrode could you detect a pulse? You know? And 475 00:23:23,759 --> 00:23:24,119 can you. 476 00:23:24,079 --> 00:23:28,240 Speaker 2: Distinguish one neuron from another neuron based on its signature, 477 00:23:28,319 --> 00:23:31,920 So like if one neuron has almost like an accent 478 00:23:32,039 --> 00:23:35,440 or a voice, if your sensors are precise enough, you 479 00:23:35,440 --> 00:23:37,799 can say, okay, that sort of faint voice we hear 480 00:23:38,160 --> 00:23:42,799 that faint signal is this neuron, This loud signal is 481 00:23:42,839 --> 00:23:43,680 a nearby neuron. 482 00:23:43,759 --> 00:23:46,240 Speaker 1: And you can actually figure out especially where these neurons 483 00:23:46,240 --> 00:23:51,799 are based on on slight differences and how they they fire. 484 00:23:52,799 --> 00:23:54,359 Speaker 3: And that's and that's how you're going to map the 485 00:23:54,359 --> 00:23:56,839 function of the brain and get a step closer to 486 00:23:56,839 --> 00:23:57,920 that whole brain interface. 487 00:23:58,960 --> 00:24:02,799 Speaker 2: Yeah, I mean, we definitely are venturing into deep sci 488 00:24:02,799 --> 00:24:05,359 fi here. If people are interested in some sci fi 489 00:24:05,400 --> 00:24:09,079 book recommendations I would recommend in banks, the culture books 490 00:24:10,200 --> 00:24:13,079 and in banks actually just have this concept of a 491 00:24:13,079 --> 00:24:16,759 neural lace where there's all the humans have a neural 492 00:24:16,799 --> 00:24:19,559 link or neural lays throughout their brain, and when somebody dies, 493 00:24:19,720 --> 00:24:21,720 their memories are being dynamically uploaded to. 494 00:24:22,680 --> 00:24:24,400 Speaker 1: The cloud or whatever the internet is. 495 00:24:24,400 --> 00:24:27,519 Speaker 2: In the future, they can reinstantiate into human body if 496 00:24:27,519 --> 00:24:28,039 they want. 497 00:24:28,640 --> 00:24:31,000 Speaker 1: Well, they can live in simulation, which we might be 498 00:24:31,000 --> 00:24:31,480 in right now. 499 00:24:31,559 --> 00:24:33,680 Speaker 2: If so, I'd just like to applaud the simulators on 500 00:24:33,720 --> 00:24:34,759 the excellent work they are doing. 501 00:24:34,920 --> 00:24:37,839 Speaker 3: This feels very immersive and high fidelity. So thank you 502 00:24:37,920 --> 00:24:38,759 to simulator. 503 00:24:39,480 --> 00:24:41,119 Speaker 1: Thank you simulators. Please don't turn us off. 504 00:24:41,240 --> 00:24:46,880 Speaker 3: Yeah, well, well listen, Elana, this has been a terrific conversation. 505 00:24:47,039 --> 00:24:51,000 You have all of neurosurgery in the room here, and 506 00:24:51,079 --> 00:24:53,519 so what are maybe some last thoughts you'd like to 507 00:24:53,599 --> 00:24:54,160 leave us with. 508 00:24:56,039 --> 00:24:59,279 Speaker 2: Well, I think this is going to be something that 509 00:24:59,319 --> 00:25:04,400 is an incredible powerful tool for neurosurgeons for helping fix 510 00:25:04,480 --> 00:25:07,359 things that are rare related issues. It's sort of like, 511 00:25:08,279 --> 00:25:11,839 you know, it might be like the difference between if 512 00:25:12,079 --> 00:25:15,400 it was a weapon situation, difference between like bows and 513 00:25:15,480 --> 00:25:19,680 arrows and jet airplanes, Like, it's a big difference, you know, 514 00:25:19,920 --> 00:25:21,000 so we want to give you. 515 00:25:21,359 --> 00:25:24,599 Speaker 4: I hope I have the airplane in that Yeah, in. 516 00:25:24,519 --> 00:25:25,839 Speaker 1: A positive constructive way. 517 00:25:26,119 --> 00:25:28,640 Speaker 2: I mean, one can only do as well as the 518 00:25:28,680 --> 00:25:30,480 tools that want to get you know what. 519 00:25:30,599 --> 00:25:31,960 Speaker 1: It's like, what tools do you have? 520 00:25:32,680 --> 00:25:36,799 Speaker 2: And I think with my essentially giving neurosurgeons a much 521 00:25:36,839 --> 00:25:40,599 more sophisticated, powerful tool like the neuralink device, you could 522 00:25:40,640 --> 00:25:44,640 really help a lot of people terrific. 523 00:25:44,680 --> 00:25:47,319 Speaker 3: And I know that's why we're all here to better 524 00:25:47,400 --> 00:25:50,440 characterize in our logic disease and to help people. So 525 00:25:50,799 --> 00:25:55,000 really value your perspective. Thank you for being are a 526 00:25:55,039 --> 00:25:57,440 puzzle lecturer for creativity innovation