访谈·中文精读

马斯克 Neuralink“Show and Tell”发布会

Elon Musk’s Neuralink “Show and Tell” event

中文精读由 AI 基于原始视频与文稿改写,非官方翻译,可能存在偏差或错误;表述以原视频为准。说明

六个月,把一块硬币大小的芯片放进人的大脑。这个时间表在 Neuralink 的"Show and Tell"发布会上被马斯克说得像一件已经排上日程的事,语气平淡得不像在宣布医疗史上的第一次。

比这个承诺更早出现的画面,是一只叫 Sake 的猴子。它坐在屏幕前,两只手都没碰键盘,屏幕上的光标却稳稳地移向一个又一个高亮的字母,把单词一个字母一个字母拼了出来。

马斯克紧接着把话往回收了一半。他说,严格来讲 Sake 还不会拼写,这套系统只是把光标移到指定位置,真正的拼写能力属于下一代产品,他不想把这件事说过头。

这个自我纠正的动作,比演示本身更能说明 Neuralink 现在处在什么位置。它已经能让一只猴子用意识控制光标,但离读写语言还有距离。这场发布会的全部内容,基本上就是在讲这段距离有多长,以及他们打算怎么走过去。

目标是一整颗大脑,起点是一个光标

Neuralink 的长期目标,马斯克用的词是"全脑接口"(whole brain interface)。一个通用的输入输出设备,长期来看可以和大脑的每一个部分交互;短期来看,先做到能和大脑的任意一个特定区域交互,解决那些让人失去正常生活能力的病症。

这里有两个方向的动作:读取大脑的信号,写入大脑的信号。最终把这两件事覆盖到整个大脑,再往外延伸到神经系统。脊髓受伤的人、颈部受伤的人,都属于这个延伸范围。

所以这场发布会讲的东西,从最底层看只有一件事:怎么在大脑和外部世界之间,接一根带宽足够高、又足够安全的线。

带宽这个词在脑机接口里不是修辞。人的神经信号密度极高,一个通道只能读到一小片区域里一小群神经元的活动。通道越多,能还原的信息越细。这也是为什么 Neuralink 一路往上堆通道数,而不是把力气花在别的地方。

原型和产品之间,隔着一百到一千倍

马斯克在这场活动里说了一句和发布会主题不太搭的话。他说原型很容易,量产很难,从原型走到一个安全、可靠、在各种环境下都能工作、价格能被人接受、还能规模生产的设备,难度要高出 100 到 1000 倍。

这句话解释了为什么一家 2016 年成立的公司,到 2022 年底还在讲猴子和机器人。脑机接口的演示从来不难,难的是把一根电极稳稳地插进几千个人的大脑里,十年后还不坏。

进度上,Neuralink 表示已经把大部分文件提交给了美国食品药品监督管理局(FDA),预计大约六个月后可以把设备植入人体。发布会的后段,团队又给出了一个略保守的说法:希望在未来六个月内,在美国启动第一项人体临床试验。一个是"植入人体",一个是"启动试验",这两个时间点不是同一件事,但都被放进了同一个六个月里。

动物实验之前,先做完能做的台架测试

发布会对动物实验的质疑做了正面回应。Neuralink 团队的说法是,在考虑把设备放进动物体内之前,他们把所有能做的严格台架测试(benchtop testing)都做完了。设备进入羊、猪、猴子体内时,团队希望每一次植入都能起到确认作用,不拿活体动物试错。

"验证性"和"探索性"在伦理上是两种不同重量的词。前者意味着前面已经堆了一堆数据,动物身上只是最后确认一遍;后者意味着拿动物去撞运气。Neuralink 想强调的是前者,也用了"极其小心"这样的措辞来描述整个过程。

自上一场演示以来,实验猴群扩大到了六只。原来的 Pager 做了硬件升级,猴群在做目标追踪任务。团队反复强调的词是稳定、可复现、长期不退化。对植入式设备来说,这三个词的分量比任何一段炫酷视频都重。

六只猴子,一只会打字

Sake 是这次的主角。它用意识把光标移到屏幕上高亮的按键上,完成输入。马斯克把它称作"心灵感应式打字"(telepathic typing),紧接着又承认这套系统还不能真正拼写。

能做的事已经很清楚了:控制光标,控制鼠标。对四肢瘫痪或者四肢功能丧失的人来说,这意味着能操作手机、能操作电脑。Neuralink 团队给出的判断是,一个几乎没有其他方式和外界交互的人,用这套设备操作手机会比一个双手正常的人更快。这句话如果成立,脑机接口的价值就从"恢复"跨到了"超过"。

发布会还花了点篇幅说明猴子们的状态。Sake 自己喜欢做这个演示,没有被绑在椅子上逼着做。马斯克说这些猴子过得挺开心,还顺口夸了它们一句,说它们在水果这件事上是反应很快的决策者。

为了让猴子愿意配合,团队把充电线圈嵌进了猴子的栖息环境,加了一个新的外部控制环路,还装了一台香蕉奶昔泵。猴群被训练成自己去充电。Pager 身上的 N1 植入体在实时回传诊断数据,当它爬到线圈下方坐下,充电器会自动检测到它的存在,从搜索状态切换到充电状态。屏幕上能看到 0 到 1 区间的调节功率输出,以及充进电池的电流。

这段细节看着像花絮,其实是整场发布会里最接近"产品体验"的部分。一个要陪着人过一辈子的设备,充电这件事必须做到不用想。

iPhone 1 不该永远留在你的头骨里

可升级性是马斯克在这场活动里讲得最具体的一个产品逻辑。他说第一批量产设备会很像 iPhone 1,而他很确定没有人愿意把 iPhone 1 一直卡在脑袋里,尤其是在 iPhone 14 已经上市的年份。

所以设备必须做到完全可逆、完全可升级。可以取出来,换成最新版本;如果它因为任何原因停止工作,也可以换掉。这是 Neuralink 对产品的硬性要求。

这条要求背后是一个常被忽略的现实:脑机接口的电极迟早会老化,生物组织也会在电极周围形成胶质疤痕。如果设备不能换,第一批用户就会被永久锁死在第一代硬件上。Neuralink 把"能换"写进设计目标,等于承认了这个物理事实。

第一批两个人:一个重新看见,一个重新动起来

Neuralink 在人体上首先要做的两个应用,一个落在视觉皮层,一个落在运动皮层。

视觉这一项,马斯克特意点出了一个反直觉的地方:即使一个人从来没有过视力,生下来就是盲人,Neuralink 相信仍然可以帮他恢复视觉。理由是视觉皮层还在,只是没收到信号。设备要做的是直接给这块皮层写入图像信息。

运动皮层这一项,目标是让几乎失去所有肌肉控制能力的人重新操作设备,马斯克用的参照是霍金。第一阶段是让他们用意识操作手机,速度能超过双手正常的人。

更进一步的版本是"桥接"。如果一个人的脖子断了,从运动皮层采集信号,再把这些信号传给植入在脊髓里的 Neuralink 设备,理论上就能把断掉的通路接回来。Neuralink 团队表示,在实现全身功能这件事上,他们看不到物理层面的限制。

"看不到物理限制"是一句很重的话。神经系统本质上就是电信号和化学信号的传递,如果两端都能读写,中间断掉的部分确实可以被电子设备替代。这句话的技术含量在于,它把问题从"能不能"换成了"什么时候"和"多贵"。

N1 长什么样

N1 是 Neuralink 的植入体,大小约等于一枚 25 美分硬币。它有超过 1000 个通道,能记录信号,也能施加刺激。这些通道做在一种柔性薄膜阵列上,Neuralink 管它叫"线程"(threads)。

整个设备完全可植入、完全无线。手术结束后,植入体藏在皮肤下面,外面看不见。它带一块可以无线充电的电池,用户在家里就能充。

1000 个通道在脑机接口领域属于很高的密度,把这么多通道塞进一枚硬币大小的无线设备里,意味着散热、功耗、封装全都要重新设计。柔性线程解决的则是另一个问题:硬质电极会随着大脑的搏动不断切割组织,柔性材料能跟着组织一起动。

充电难在两摄氏度的红线上

消费电子产品的充电可以很简单,插一根线,或者往充电板上一放。完全植入的设备没有这个选项。

Neuralink 列出的挑战有四条。充电要在较大的体积范围内完成,不能依赖磁铁做精确对准;系统要能抗干扰;充电要快,不能让人每天花很长时间等;最要紧的是安全,植入体外表面和脑组织接触的那一面,温升不能超过 2 摄氏度。

2 摄氏度这条线不是随便定的。脑组织对温度非常敏感,超过这个幅度就可能造成损伤。这意味着无线充电的效率必须足够高,损耗必须足够低,因为任何一点能量浪费都会变成热量,而热量在密闭的颅骨里面没有地方散。

当前的量产充电器做在一个铝制电池底座里,集成了驱动电路,外接一个可拆卸的远端线圈,尺寸是初代设备的四倍。线圈的开关频率提高了,耦合效率也跟着提高。远端线圈做大了,对准的容错空间就大了,用户不需要把充电器精确摆在某个位置上。

插进大脑的线,只比红细胞宽一点

线程的直径只有几个红细胞那么宽。这种尺寸的东西,手工插不进大脑,所以 Neuralink 造了 R1 手术机器人。

发布会现场做了一次完整的插入演示。演示对象是一个"大脑模型"(brain proxy),患者代号 Alpha。屏幕上分两个视图,左边是靶向视图,粉色代表要插入电极的皮层表面,黑色代表要避开的血管,带编号的井号标记是每一根线程的预定位置。右边是机器人的实际操作,它把线程阵列从硅背衬上一根一根剥下来,插进预先定好的目标点。

第一根插进去的时候,现场鼓掌了。整台手术如果要插 64 根线程,机器人做这件事大约需要 15 分钟。

手术的完整流程是这样的:定位,切开头皮,钻开颅骨(开颅术),切开包在大脑外面那层坚韧的硬脑膜(硬脑膜切开术),插入细软的电极线程,把植入体放进开好的孔里,缝合。到这里结束,皮肤下面就有了一个看不见的植入体。

机器人只负责插线这一段。其余步骤由神经外科医生完成。Neuralink 说得很直白,插线这一段手工做太困难,所以交给机器;而开颅和切开硬脑膜是整个手术里最考验医生技术的两步,目前还得靠人。

最难啃的是那层膜

过去一年 Neuralink 在硬件上做了几件事:提升植入体的耐用性,改进电池和充电性能,优化蓝牙连接的可用性。团队的态度是,每一个新版本都会比上一版明显更好。

但真正卡住手术的,是那一层一毫米左右的膜。

为了让设备的升级和更换像初次安装一样简单,Neuralink 试过很多绕开愈合过程的设计。走得最通的一条路是让手术本身侵入性更小:不直接把大脑表面暴露出来,保留硬脑膜,让身体自带的保护屏障继续工作。

麻烦在于,硬脑膜保护大脑的那些特性,同时也让它很难被穿透。人的硬脑膜可以超过 1 毫米厚。1 毫米听起来不多,但对比 40 微米粗的针,这个差距大到荒谬。Neuralink 给了一个换算:如果把针放大到铅笔那么粗,硬脑膜会超过 4 英寸厚。他们还拍了段镜头,让人看要放大到什么程度才能勉强看见针的轮廓,等针的细节进入画面时,同一帧里已经能数清红细胞了。

插线过程要闯三关。

第一关是抓取。针和包住它的保护套管要一起抓住线程,把它从保护性的硅背衬上剥下来,一直抓着移动到大脑表面,插入时再从套管里释放。

第二关是穿透。硬脑膜下面的大脑非常软,针如果不够锋利,只会把表面压出一个凹陷,怎么都穿不进去;如果针露在外面的自由长度太长,它会在压力下直接弯掉。

第三关是带着东西穿。光把针穿过去没用,线也必须跟着进去。所以要优化的对象是针和线合在一起的整体轮廓,单独把针做细做锐,反而可能让线过不去。

设计的人就坐在产线旁边

Neuralink 提到一个组织方式上的细节:负责设计的工程师同时也在物理制造线上工作,参与搭建和调试。团队说这一条对缩短迭代周期极其重要。

放在植入式医疗器械这个行业里看,这种做法不太常见。传统路径是设计部门出图,制造部门按图生产,两边来回扯皮,一轮迭代几个月就过去了。设计和制造由同一批人负责,问题当天就能改,代价是工程师必须懂产线,产线也必须懂设计。

手术能力也在扩张。Neuralink 在奥斯汀建了自己的手术室,而且是双手术室。团队把它定位成过渡设施,最终目标是建自己的 Neuralink 诊所。

这件事最终要解决什么

把 N1 植入体和 R1 机器人放在一起看,Neuralink 给第一阶段定的目标很具体:帮助因完全性脊髓损伤而瘫痪的人恢复"数字自由",让他们操作设备的能力至少和受伤前一样好,甚至更好。

过去一年,这件事是整家公司的核心焦点。Neuralink 表示一直在和 FDA 密切合作,争取批文,希望在未来六个月内于美国启动第一项人体临床试验。

这场发布会的真实目的:招人

发布会接近尾声时,主讲人说了一句很实在的话:这次更新最主要的目的是招募。

他想传递的信息只有一条。如果你擅长创造手表、手机、电脑这一类先进设备,你的能力在这里会非常有用。

这个说法和整场发布会的内容是对得上的。Neuralink 眼下最缺的,是把东西做出来、做得便宜、做得能批量生产的人。马斯克那句"量产比原型难 100 到 1000 倍",本质上是在解释为什么这家公司要招这么多硬件工程师。

整场活动看下来,真正的信息量在那些具体的数字里。1000 多个通道,40 微米的针,1 毫米的硬脑膜,2 摄氏度的温升上限,15 分钟插 64 根线,100 到 1000 倍的量产难度。这些数字决定了脑机接口什么时候能从演示走进手术室。

  • 把发布会上的时间表打个折扣听。同一场活动里,六个月既指"把设备植入人体",又指"启动临床试验",这两个里程碑差了至少一整年。判断真实节奏,看 FDA 的临床试验备案和器械豁免批准记录,营销话术只当背景音。
  • 评估一家脑机接口公司,先看它有没有自己的手术机器人、自己的产线和自己的手术室。能做出电极的团队不少,能把 64 根电极在 15 分钟内稳定插进运动中的大脑、还能绕开血管的团队很少,这条能力线比通道数更能说明工程成熟度。
  • 记住 2 摄氏度这条线。它是植入式设备安全性的硬约束,也是判断无线充电方案是否靠谱的最快指标。任何一家公司如果对温升避而不谈,基本可以判定它还没走到量产阶段。
  • 关注"可逆"和"可升级"两个承诺有没有落到设计里。设备能不能取出来、能不能换新,决定了早期使用者会不会被锁死在第一代硬件上。这条比"能读多少个神经元"更影响一个人要不要报名做第一批。
  • 如果你是从消费电子、可穿戴设备或精密制造转过来的人,这场发布会的招募信息值得认真读一遍。把手机做到防水、把耳机做到续航三天的那套工程直觉,在脑机接口这个领域目前是稀缺资源。
YouTube 平台字幕(自动或人工),可能有识别错误,人名与数字请以视频为准。已按语句合并成段落,并去掉 um、uh 等口头填充词。 文稿来源:YouTube(自动/人工字幕)。

0:00all right welcome to the neuralink show and tell the overwatching goal of neuralink is to create a ultimately a whole brain interface a generalized input output device that in you know in the long term literally could interface with every aspect of your brain and in the short term can as can interface with any given section of of your brain and and solve a tremendous number of things that that cause debilitating issues for people so so you want to be able to read the signals from the brain you want to be able to to write the signals that you

0:41want to be able to ultimately do that for the entire brain and then also extend that to communicating to the rest of your nervous system if there's a if you have a sort of a seed spinal cord or neck I've often said that prototypes are easy production is hard it's really I'd say 100 to a thousand times harder to go from go from a prototype to a device that is safe reliable Works under a wide range of circumstances is Affordable and done at scale we've submitted I think most of our paperwork to the FDA and we're we're we think probably in about 6 months we should be able to have alost neuralink in a human so before we would even think of putting a

1:32device in an animal we we do everything we possibly can with rigorous Ben benchtop testing so we're not Cavalier in putting devices into animals we're extremely careful and we we always want the device whenever we do the implant if it's in a she sheep or a pig or monkey to be confirmatory

1:57not exploratory let's see since since the pag Dem Mo We've expanded to work with a troop of six monkeys We've we actually upgraded pager they do VAR tasks and we do everything possible to ensure that the things are stable and replicable and that things La that the device lasts for a long time without degradation and here you can see sake it's one of other monkeys typing on a keyboard now he's this is telepathic typing so clear this is the he's he's

2:33not actually using a keyboard he's moving the cursor with his mind to the highlighted key now technically we can't can't actually spell and so I don't want to oversell this thing because that's that's the next version so the but what's really cool here is is sake the monkey is moving the mouse cursor using just his mind moving the cursor around to the highlighted key and then spelling out what we what we want something that could be used for somebody who's who's say quadriplegic or tetrolic human

3:21even before we make the the spinal cord stuff work is being able to control a mouse cursor control a phone and we we're we're confident that you that someone who is has basically no other interface the outside world would be able to control their phone better than someone who has Working Hands and I mentioned upgradeability upgradeability is very important because our first production device will be much like an iPhone 1 and I'm pretty sure you would not want an iPhone one stuck in your head if the iPhone 14 is available so it's going to be it's be able to demonstrate full reversibility and upgrade ability so you can remove a device and replace it with

4:10the latest version or if if it stopped working for any reason replace it it's that's that that's a fundamental requirement for the device at New link I think it's also important to show that sake actually likes doing the demo and it's not like stra to the chair or anything so it's yeah so like our monkeys are pretty happy you know so as you can see a quick decision maker on the fruit front the first two applications we're going to aim for in humans are restoring vision and I think this is like notable in that even if someone has never had Vision ever like they were born blind we believe they can they can we can still restore Vision so because the visual part

5:07of the the visual part of the cortex is still still there and then the the other application being in the motor cortex where we would initially enable someone who has no ability almost no ability to operate their their muscles you know sort of like a sort of Steven Hawking type situation and enable them to operate their phone faster than someone who has H Working Hands but then even obviously even better than that would be to bridge the connection so take

5:45take the out the signals from the motor cortex and let's say somebody's got a broken neck then bridging those signals to neuralink devices located in the spinal cord we're confident there are no there are no physical limitations to enabling full body functionality how do you create a high bandwidth generalized interface to the brain so our first steps along these dimensions for our device is what we called the N1 implant it's a size of of about a quarter and it has over 1,000 channels

6:23that are capable of recording and stimulating it's microfabricated on a flexible thin film array that we call threads it's fully implantable and wireless so no wires and after the surgery the the implant is under the skin and it is invisible it also has a battery that you can charge wirelessly and you can use it at home unlike many consumer electronic devices which can simply offer a physical connector charging a fully implantable device poses several unique challenges first the system must operate over a wide charging volume without relying on magnets for perfect alignment the system must be robust to disturbance and complete quickly so is not to be overly

7:08burdensome however most important is safety in contact with brain tissue the outer surface of the implant must not rise more than 2 de C our current production charger which charges our current generation of implants is implemented in an aluminum battery base which also includes the drive circuitry a remote coil four times the size of of our original device also disconnectable this this remote coil has a increased switching frequency driving improved coil coupling I'd like to show you one of these applications here with a device we call our simple charger and the coil has been embedded into the habitat with the addition of one new

7:52outer control Loop plus a banana smoothie pump The Troop has been trained to charge themselves on the right we're streaming realtime diagnostics from pagers N1 when he climbs up and sits below the coil you can see the charger automatically detect his presence and transition from searching to charging we see the regulated power output on a scale of 0 to one and the current driven into his battery so similarly for implanting our device safely into the brain we built the surgical robot that we call the R1 robot it's capable of uing these tiny threats they're only on the order of few red blood cells wide and inserting them reliably into a moving brain while avoiding

8:40bascur it's it's quite good at doing this reliably and in fact because we've never shown an endtoend insertion of a robot in action we're going to do a live demo of the robot doing surgery in our brain proxy so who wants to see some insertions so here here it is that's our R1 robot with our patient Alpha who's lying comfortably on the patient bed this is what we call the targeting view so what you're seeing is this is a picture of our brain proxy and the pink represents the cortical surface that we want to insert our electrodes into and the black represents the vascular that we want to avoid and what you're seeing is these hasch Mark with numbers that represents where we intend

9:25to put each of our threads so this is another view real quick on the left is the view of the insertion area and on the right what the robot's going to do is it's going to peel the array the threads one by one from it silicon backing and insert it into the targets that we predetermined in the targeting view

9:52so there you go that's the first insertion [Applause] so we're going to see a couple more insertions the whole process of inserting about 64 threads in our first product is going to be around 15 minutes for this robot to get an N1 device it's essentially these steps targeting and the incision drill the craniectomy remove the tough outer menal layer called the dura then insert the thin flexible threads of electrodes place the implant in into the hole we created and then that's it you've got an implant Under the Skin the surgical robot does the thread insertion part of the surgery this is because it would be very difficult to do manually the rest

10:40of the surgery is done by the neurosurgeon there's still a lot for us to do to get to that procedure where we reduce the role of the neurosurgeon and make it affordable and accessible the primary the two elements of the surgery that demand the most skills from the neurosurgeon are the craniectomy and the dctom you've gotten to hear about the advancements we made over the past year we've improved implant robustness battery and charging performance Bluetooth usability realistically every new device version is going to be significantly better we need to keep this new technology accessible for our early adopters this means that we need a solution to make device upgrade or

11:19replacement just as easy as it is to initially install we've explored many different avenues for Designing around this healing process and finding a solution to make device upgrade seamless our best successes have come from making the procedure less invasive instead of directly exposing the brain surface we instead keep the dura in place maintaining the body's natural protective barrier the same properties of the dura that make it a good protector of the brain also make it really difficult for us to insert the threads into in humans the dura can be over a millimeter in thickness which doesn't sound like a lot but compared to our 40 Micron needles it actually is a

11:56lot for example if you scaled up the needles to the the size of a pencil the durwood scale to over 4 in in thickness take a look at how far you have to zoom in to even see it by the time the features of the needle come into frame you could see individual red blood cells in the same frame One Challenge is that we have to use the needle and the protective canula that it sits in to grab onto the thread and to hold it while we peel it from the protective silicon backing and then we have to keep holding it while we bring it over to the surface and then release it from the canula during insertions and another challenge is that the brain is really soft beneath the

12:32tough Dura and so if the needle isn't sharp enough it'll just keep dimpling the surface without puncturing and if this free length gets too long it can actually just Buckle the needle like this another challenge is that we don't just have to get the needle through we have to get the thread through as well so we really have to focus on optimizing the combined profile of the needle and thread together it's very typical for us to have our Engineers who design also work on the physical manufacturing line to to build and debug and this has been extremely extremely critical in reducing our iteration cycle time and we've also scaled up our surgery so we now have

13:12a dedicated our own o in fact a double O in Austin and this is just a stepping stone before we eventually build our own neuralink clinic so with this product N1 and R1 our initial goal is to help people with paralysis from complete spinal cord injury regain their digital Freedom by enabling them to use their devices as good as if not better than they could before the injury and as Elon mentioned over the last year this has been the central focus of the company and we've been working very closely with the FDA to get approval and to launch our first inhuman clinical trial in the US hopefully in the six in next six months

14:03primary purpose of this update is recruiting so this is if there's one message I want to convey it is that if you have expertise in creating Advanced devices like watches and phones computers then your your capabilities would be of great use in solving these important problems

资源与出处

相关视频

同频道、同主题或相邻年份