访谈·中文精读

马斯克在 Neuralink 2022 Show and Tell 活动回答所有问题

Elon Musk Answers All Your Questions at Neuralink's 2022 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

控制一台电脑,需要多少带宽?Neuralink 给出的数字是每秒 10 比特。

这场 2022 年 Show and Tell 的问答环节里,主持人提到自己曾经是脑机接口比特率纪录的保持者。马斯克接过话,说他们大概把纪录刷新到了 7.4 bit/s。7.4 比特,连一条像样的短信都塞不满,却足够让一个人移动光标、点开应用、敲出字母。

整场问答的信息密度很高。台上坐着做微加工的、做算法的、做机器人的、管动物福利的,问题从电极会不会被疤痕组织淹死,一路问到马斯克敢不敢把设备装进自己孩子脑子里。技术细节大多由团队回答,马斯克负责那些关于规模、时间、风险和终局的部分。

每秒 10 比特就够了,因为拇指是两根慢吞吞的肉棍

很多人对脑机接口的直觉是:通道越多越好,带宽越大越好。马斯克的回答把这个直觉拆了一半。

他提到,控制一台手机或电脑,实际需要的比特率非常低。团队测出来的数字大约在每秒 10 比特量级。他顺手做了一个类比:你用手指在屏幕上敲字,拇指每秒能产生多少信息?"我们的拇指就是两根移动缓慢的肉棍。"(马斯克)

你拿到每秒 10 比特,已经在拼命跑了。这种速率你根本不需要蓝牙,用哔哔声都能发出去。(马斯克)

这句话听起来像玩笑,背后是个重要的工程判断:输出端的瓶颈在无线电之外,卡在人本身。人的运动意图本来就慢,光标移动、点击、选择,都是低速动作。真正吃带宽的是另一件事,把 1024 个通道的原始神经信号完整传出来,做科研分析。这两件事的带宽需求差了几个数量级。

团队给出的现状是这样:蓝牙链路大约 150 KB/s,从植入体发出的压缩流大约 50 KB/s。目前够用。但设备一旦走向 16000 通道,这个数字就撑不住了。

他们的解法分两条路。一条是更好的压缩,芯片实时检测 Spike,只把事件发出去,而不是发整段波形。早期他们做过有线版本,用 USB-C 接口把全部原始信号流出来,还发过论文,那是用来做开发验证的。另一条路更激进:把解码模型直接搬到植入体上跑。现在跑在给猴子打游戏的 MacBook 上的神经网络,要移植到植入体那颗芯片上。团队内部管这个叫解码头(decoder head),有人开玩笑说,这相当于让复杂的神经网络跑在一个车库门开启器的算力上。

无线电本身也在换。他们在看 500 MHz 带宽的超宽带方案,比特率能做到 6 到 10 Mbps,延迟改善明显,另外还在评估 W 波段。马斯克在这里补了一句他惯常的做事顺序:先把该省的都省掉,再去换更快的管道。

大脑是装在椰子里的水气球

被问到过去两年最大的教训,好几个人给了同一个答案:大脑比想象中动得厉害。

做微加工的 Zach 说,很多神经科学研究从小鼠开始,小鼠的脑子几乎不动。换到人,脑组织随呼吸和心跳能移动几百微米。而 Neuralink 的导线和针尖只有几十微米,这个位移在显微镜下看,"就像一英里"。

马斯克的版本更形象。

我学到的是,大脑真的很软。比你以为的软得多。它不像花椰菜或者西兰花,它更像一个水气球,然后在你的颅骨里晃来晃去。所以,一个装在椰子里的软水气球,大概是最接近的比喻。(马斯克)

这个物理事实带来一个技术难题:神经非平稳性。同一个通道,今天记录到的放电率和上个月不一样。有提问者拿出数据图,几百天里波动很大,问这是否意味着电极在漂移。

软件负责人 Bliss 的解释分两层。第一层,你每天记录到的可能根本不是同一个神经元,换了一个细胞,放电率自然变。第二层,他们不认为这是主因,因为即便平均放电率大幅变化,Spike 的波形形状依然稳定。波形稳定说明电极没跑远,大概率还在听同一批细胞。他也承认这个证据不够铁,一部分波动仍可能来自细胞本身的更替。

马斯克补充了机械结构上的原因:电极阵列固定在人颅骨上,中间那段长长的细导线有一段盘绕结构,起到缓冲作用。位置其实相当稳定,波动更多来自生物一侧,机械一侧的问题小得多。

有意思的是,同一个"大脑会变"的事实,在后面被另一位提问者当成优势来问。他说,DJ 最早那场演讲里提到过,脑机接口也许能让人学会复杂技能,比如功夫,因为大脑天生是学习机器。可后面很多技术方案又在努力对抗大脑随时间的变化,这两件事怎么统一?

团队的回答是双向学习。一方面算法要保持稳定,另一方面人会学会怎么用设备,用久了表现会变好。人适应机器,机器也适应人,界面和算法跟着一起更新。另一位工程师补了一句:大脑可塑性对他们是好事,用户在回路里自己会学,他们反而省了功夫。加上通道数足够多,电极均匀铺在一个功能区域上,个别位置漂移可以交给软件处理。

600 天、617 天:寿命的三个瓶颈

寿命是这场问答里被反复追问的主题,团队给的数据也很实在。

Zach 提到,目前最长的动物数据来自一只植入超过 600 天的猴子,期间通道一直有用,信号能支撑脑机接口任务。负责硬件的 Jeremy 把这只猴子的名字点了出来:Pager,617 天,后来升级到了最新设备。当前版本的设备,有一批动物已经接近或达到一年,脑机接口功能完全正常。

加速老化测试的数据更夸张。上一代植入体累计了 8 年等效时间,当前版本已经跑了 4 年等效时间,还在继续。

Jeremy 把寿命拆成三个瓶颈。

第一个是密封。外壳用热塑性聚合物,理论上会有极少量水汽随时间渗入。按这个速率推算,密封本身能撑 20 年以上,目前还没有见过密封失效的案例。

第二个是电池和内部电子。这取决于使用强度。他们现在的目标是在第三年仍保持 80% 的运行时间,换算下来,标称 4 小时的续航,三年后还有 3.5 小时。团队的说法是续航很快翻倍,计划里还有四倍。

第三个是导线,也是真正的瓶颈。电极周围的通道能不能持续记录到信号,决定了设备实际可用的年限。他们同时在推两条材料路线:非晶碳化硅绝缘,目标是远超五年;原子层沉积,用极薄的层保持柔韧性,把寿命再往前推。配套还要建台架测试系统,尽可能模拟体内环境,看通道怎么衰减。

被问到能不能在电极表面载药抑制炎症,回答是正在做生物涂层,方向有两个:减少炎症,以及让表面更滑,方便日后取出。另一条并行的思路是把电极做得更小。小到某个尺度,身体的炎症反应和疤痕组织会变得极其微弱,"身体基本上就忽略了它"。

疤痕组织不碍事,取出来才是麻烦

有位提问者问得很直接:植入后疤痕组织会长在电极周围,记录信号会越来越小,这个问题怎么解?

Zach 承认没法彻底解决,但有两个优势:导线柔软、尺寸小,本身就能限制疤痕和损伤。方向是把线继续做细。

更关键的信息来自病理学。做动物研究的负责人说,他们有针对长期植入动物的组织病理终点,结论是导线周围的疤痕组织通常可以忽略,"大脑几乎不对导线作出反应"。真正需要处理的是皮层表面那层新膜,它会填满手术中移除组织留下的空隙。这层膜属于填补被拿掉的组织,谈不上持续性的异物反应,不构成问题。

那么风险在哪?在取出。

这是整场问答里最诚实的一段。团队说,长期植入本身看不到行为缺陷,也看不到认知缺陷,组织病理结果很好。难点在于把设备拿出来的时候,那才是可能造成损伤的环节。所以他们花很大力气做可逆性设计,做经硬脑膜插入,尽量减少取出时的创伤。

好消息是,因为导线极细,脑内的疤痕囊很小,取出相对容易。他们看到过第二次、第三次在同一位置植入后依然有可用信号,也有猴子正在用第二台设备正常工作。

升级流程也在往"换设备跟装设备一样简单"的目标走。他们把旧设备弹出来,在同一位置装新的,位置选在初级运动皮层,这个区域对这类交互最有价值。Pager 就是按老的手术方式换的设备,换完三四天就恢复到接近之前的控制水平。

50 微米的球,和百万像素的视觉

关于刺激精度,做刺激算法的 Dan 给了一组可以抓在手里的数字。

单个电极能激活多少细胞,取决于阻抗、导电垫面积、电流大小、频率,变量很多,正好用来定制光幻视的强度。按他们的估算,当前电极激活的范围大约是直径 50 到 100 微米的球状细胞团。

这个尺度直接决定视觉假体的像素。球越小,光幻视越精细,图像像素越小。更妙的是,电极和像素之间没有一对一的关系。通过控制电极之间的电场,可以动态调整场形,让一个电极对应更多神经元。神经元与电极的比例可以做到 10 到 100 比 1,理论上能拼出百万像素级的视觉。能不能"正常看见",Dan 的回答是,这是可能的结果之一。

这也解释了为什么高通道数在这套设计里如此重要。通道多带来两个好处,记录得多,刺激的自由度也大。

从一只猴子到所有猴子

被问到两年里学到什么,做算法的 Avinash 给了一个产品视角的答案:让一只猴子、一台设备跑通原型,是成功,但相对容易;让所有猴子每天都跑通,就是做产品,这不容易,他们正在学。

做机器人的 Alex 说,他们造的运动系统精度到个位数微米,那就需要比它更可信的验证和测试系统来证明它可靠。把同样的精力投到验证系统上,和产品并行设计,是这两年最重要的收获之一。

还有一条关于实验方法的教训:植入环境是动态的,疤痕会长,组织会填,导线在里面的处境会变。等一个植入位点愈合要几个月,等不起。所以他们花大力气做准确的代理模型(proxy),用几小时而不是几个月,拿到原本要等很久才知道的信息。

验证插入是否到位,用的办法也很工程化。他们做微 CT,把整块代理材料扫一遍,看电极有没有在里面打卷。然后刻意造出两种代理:一种每次都打卷,一种永远不打卷,把真实组织的表现卡在中间。如果最坏情况下都不打卷,就能比较有把握地说,真实手术里也不会。

故障容错的问题也被问到。如果某个通道短路或者阻抗过高怎么办?回答是冗余足够。现在 1000 个通道,未来 16000 个,解码模型实际用到的通道数远少于总数,个别通道坏掉不影响解码质量。他们手上有一个数量级的余量。

打几个孔,十分钟,像 LASIK

手术方式是这场问答里最容易被忽略、却对患者体验影响最大的部分。

现在的做法要处理硬脑膜,那层紧贴颅骨的厚膜。马斯克提到,他们觉得可以不解开硬脑膜,直接在膜上打一排微孔,把电极从孔里送进去。

如果不切开硬脑膜,而是打一堆小孔把电极穿进去,恢复时间会快得离谱。脑脊液几乎不损失。理论上整个过程可以像 LASIK 一样,十分钟搞定。这算不上大动干戈的事,非常快。(马斯克)

这个改动听起来小,实际影响很大。手术时间、麻醉负担、感染风险、恢复周期,全都跟着变。一种需要反复升级的设备,手术能不能做得像换电池一样轻,直接决定它能不能被大规模使用。

深部脑区的问题也顺带被回答了。有人问,现在都是皮层记录,什么时候能进入几厘米深的区域,做神经精神疾病、记忆相关的研究。

马斯克的回答很干脆:颅骨里那部分设备基本不变,它本质上就是一块智能手表,有电池、无线电、感应充电、计算机;变的是导线要更长,机器人的针要更长。这套东西设计成通用输入输出设备,理论上哪都能去。

机器人团队补充了现实约束:现在往下只能看清大约 1 毫米,深部需要新的成像手段,超声波和光声断层成像都能达到厘米级。但还有一条更彻底的路:如果针和线足够细,细到可以直接穿过血管而不造成出血,那就不需要成像避让了。团队对这条路径"略微乐观"。

为了说明现有技术有多粗放,他们拿脑深部刺激做了对比:钻一个 14 毫米的骨孔,把一根 2 毫米的导线盲穿进去 6 到 8 厘米,事先告诉患者有 1% 的概率脑出血且无法控制。Neuralink 的针是 40 微米。这个对比不需要任何修饰。

FDA 问了什么,马斯克怎么回答

关于监管,团队能说的不多,但说得挺具体。

取出安全是 FDA 明确提出的关切之一。在人体上严格证明取出安全,是他们在攻的难题。热安全也被追问:植入体不能因为过热损伤组织,台架测试要足够严格、足够可信,他们正在重新设计测试方法,让它更准确。生物相容性和化学表征同样被反复盘问,监管要确认没有任何有毒化学物质或不兼容材料进入大脑。

还有一个不容易被外界理解的困难:产品的创新性本身抬高了证明门槛。手术机器人和薄膜阵列都太新,找不到足够的文献和同类产品做参照,只能自己设计、自己测试、自己证明。他们明确表示,在第一位人类受试者身上,这一点不打算妥协。

然后有人问了一个把气氛拉紧的问题:你怎么看待在健康人身上降低安全门槛这件事?

马斯克没有绕。

打个比方,如果我的一个孩子脖子断了,我现在会不会放心给他装这个设备。我的回答是,以现在的状态,在我看来,它不会造成危险。(马斯克)

这句话的重量在于场景。他举的例子落在颈椎损伤、四肢瘫痪这类明确适应症上,跟健康人做增强是两回事。在这个前提下,一个尚未获批的设备,在他眼里已经跨过了风险门槛。这也解释了他后面那句关于进度的判断:面向人的进展一开始会慢得让人难受,但产能建设在同步推进,先做几千台,再做几万台,最终做到几百万台,理论上进度会变成指数级。

想炒股打败兄弟的人

有一位提问者问,你们和运动障碍群体有合作吗,他们提过什么想法?

回答里有一个细节,值得单独说。

Neuralink 有一个消费者顾问委员会,成员包括四肢瘫痪等不同状况的人,定期给公司提建议。半年前有个人来办公室,被问到最想用设备做什么,他说了两件事:一是能自己炒股,好赢过他兄弟;二是能玩射击游戏。

"最让我震动的是这件事有多平常。"(现场工程师)

这个回答比任何宏大叙事都更能说明需求落在哪里。外部世界把脑机接口当前沿科技,使用者眼里它首先意味着恢复一件普通人每天都在做的、琐碎而具体的事。能自己下单,能在游戏里赢,这就够了。

动物福利的部分也在问答里占了不小篇幅。负责人说,训练团队里配了行为分析师,主要工作是移除训练中的强迫和挫败感,以正强化为主。评估框架是 3R:优化、替代、减少。最优先要移除的是身体约束。所以你看到猴子自己走到测试位,在生活区里自愿玩笔记本,用的是全植入、全无线的设备,配感应充电。这些技术选择看着琐碎,直接决定了动物能不能不被绑着完成任务。

职业路径的问题也有人问。马斯克的回答很宽:芯片、材料、软件、动物照护,所有在招的岗位都在官网列着。团队里另一位工程师的说法更彻底,大学专业目录翻一遍,几乎每个专业都能在这家公司找到对应的人。他给高中生的建议是,找到你有热情或者有天赋的那件事,尽可能往深里做,最后你会在造神经接口的地方找到位置。

通用 IO,和给大脑造的示波器

最后几个问题回到了终局。

有人问,你们长期想实现的是和高级人工智能的高带宽通信,而高级 AI 要理解人类最复杂的思想和情感,这本身就是神经科学的任务,你们有没有超越神经工程的野心?

马斯克的回答回到了平台逻辑:他们做输入输出设备和配套的软件接口,并且打算尽可能开源,让别人在这个基础上继续建。他拿微处理器打比方,有人造计算机,就会有人为计算机写软件;如果连计算机都没有,软件就无处运行。他们造的是那台计算机。

谈到开源,前面还有一段更具体的承诺。有人问工具和数据会不会开放给神经科学界,马斯克说会,等设备进入量产、拿到 FDA 批准之后,把手术机器人和设备提供给大学的神经科学实验室和医院是合理的事。数据集也可以直接放在网站上,谁想用就用。他甚至提到,这些数据对做 AI 研究的人可能很有价值,可以用来训练大脑的基础模型。

至于记忆记录这类话题,他自己先把话挑明了:"这就进入《黑镜》的领域了。"(马斯克)

团队里做研究的 Ian 补了一段更朴素的定位:Neuralink 并非凭空出现,医学和学术界几十年的电极研究铺了路。更好的工具能让整个领域跑得更快,因为你能看见大脑里正在发生什么,还能无缝地干预它。他引用了同事的比喻,说这有点像给大脑造一台示波器。

马斯克把这句话接到了最大:

现在我们对大脑里发生的事,很多是靠猜。有了直接的输入输出,就不用猜了。这种设备被广泛使用之后,我们对大脑的理解会比现在多出好几个数量级,文字几乎没法表达这个差距。(马斯克)

关于更高带宽的输入方式,团队承认还没有答案。手写解码是从斯坦福起步的工作,他们在这里继续推进;不同键盘布局、滑动输入这类界面侧的优化也在试。另一条路是换脑区,比如语言和言语中枢,如果目标就是输出文字,从那里取信号可能更自然。这些都要等第一位人类受试者之后才好验证。

整场问答听下来,比那些数字更值得记住的,是这群人谈论约束时的坦诚程度。疤痕解决不了,就把它做小;带宽不够,就把计算搬到植入体上;深部看不清,就把针做细到不需要看清;取出有风险,就把可逆性当成设计目标。每一个回答背后都是同一个动作:先把问题缩小到能测、能验、能改的尺度。

对读者来说,这次问答里有几件事可以直接拿去用:

  • 如果你在做神经数据相关的 AI 研究,盯住他们承诺开源的数据集和工具。1024 通道、数百天、带行为标签的长期记录,是训练大脑基础模型稀缺的原料,比再刷一个视觉 benchmark 有价值得多。
  • 评估任何脑机接口方案时,先问三个瓶颈:密封、电池、导线。前两个已有 20 年和 3 年的量化答案,第三个才是真战场,非晶碳化硅和原子层沉积是这个战场上目前最具体的材料路线。
  • 判断通道数的意义,别只盯着记录。刺激端的电场塑形能做到 10 到 100 比 1 的神经元电极比,通道数同时决定了输出精度,视觉假体的像素上限就卡在这里。
  • 做深部介入的团队,把超声波和光声断层成像列进技术雷达。1 毫米的成像深度是当前手术机器人的硬约束,厘米级成像会直接打开神经精神疾病和记忆研究的入口。
  • 设计任何需要长期植入的器械,把取出当成一等公民来设计。病理数据再好,取出那一刻才是风险集中点,可逆性和经膜插入应该写进需求文档的第一页。
  • 和患者群体沟通时,先问他们想做什么具体的事。答案可能是炒股赢兄弟、玩射击游戏,这些需求朴素、明确、可验证,比任何宏大愿景都更适合当产品定义。
  • 如果你是学生或刚入行,别被"神经科学"四个字挡住。材料、机械、射频、嵌入式、动物行为、临床事务,每一个方向在这类项目里都有位置,深度比跨界标签更重要。
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)。

0:11this is obviously amazing and has clear therapeutic potential it would also be great for the scientific Neuroscience Community to access some of these tools do you have any plans to make these available to neuroscientists yes yes we do so no it's a great question is I think there's probably a lot that could be figured out if we provide the surgical robot and devices to Neuroscience research departments and universities and hospitals so I think at the point of which we have we need to be in production with the machines and obviously have the FDA approvals but I think it would make a lot of sense to provide this to research universities and hospitals

1:20the question is of the data that we of the data sets that you've collected are there any that you plan to open source for the scientific community yeah I think that would be that would be fine I think yeah sure absolutely because I think it could be really interesting for people working in AI research to build upon that in Mill and build Foundation models for the brain yeah it's a good point yeah like to actually get no problem with just publishing it on our website you can use it if you want looking forward great

2:06thank you so thank you for the very wonderful presentation so I have one question so as we will know for implantable electoral either for stimulation or recording after we implant the electrode Scar Tissue will grow around the electrode and especially for recording the signal we get will become smaller and smaller after long-term implant how do you solve this

2:33issue go figure so for context I'm Zach I lead the microfabrication team on brain interfaces I don't think we can solve it specifically but one thing one advantage we have is both the flexibility and the small size of our threads to try to limit that scar tissue and that damage and some future work that we have started working on that we'll continue working on is pushing the size of the threads down just to try to limit the immune response and really limit that scar tissue growth I actually have I want to follow up so do you think it will be helpful to actually load some drug on on the surface of your electrode or some other way well I think like maybe the just the

3:21question is like what what sort of signal degradation have we seen over time and you know basically just does it does it work a year later does it work two years later it does so yeah yeah so that's a good point so in terms of thread longevity specifically really the gold standard that we can use to assess is the data we have from our animal participants and so for that I'm not sure if it was mentioned before but the longest data we have right now is for an animal participant who has 600 600 days with useful functioning channels where we were doing something useful with the the signals for BCI and then with the newest version of our device we have sort of

4:06a collection of participants who are at or near one year of data and completely useful functioning BCI from that as well thank you if I may add one more thing so you mentioned potentially having drugs to kind of reduce inflammation so one of the things that we are actually actively working on is having some sort of biological coding to either reduce inflammation or make them slippery so you know you mentioned you heard from the presentation that one of the challenges that we have is removing the threats from these neo-membrane tissues that are formed after implantation so there are programs like that where we're really looking at kind of incorporating some of the learnings

4:46from biology and these Coatings into our thread so that we can hope reduce inflammation as well as make it easier to extract we're also continuing to reduce the size of the the electrode so as when the electrode gets really small there's the sort of inflammation response are Scar Tissue becomes minuscule so it's like a very very tiny electrode the body basically ignores this is really impressive congrats to the whole team so as as you of course know one of the problems with current electrodes is they're rigid and they move around so you have these neural nonstationarities and I think many of us had hoped that with these very thin threads they would

5:29maybe move more with the brain and you wouldn't see that but from the data you showed over many hundreds of days there was a lot of variability so can you speak to how much do they move and do you have any idea of like why does it move can you stop it from moving is it how stable are the signals hour to hour and day to day

5:49hi I'm Bliss I'm one of the leads of the software groups in the brand interfaces team in the particular plot you're mentioning before what we were showing was the average firing rate recorded per day on a particular Channel it's as you well know pretty complicated to understand if you're recording from the exact same neuron day after day after day it could be for example that you're actually picking up a different neuron day to day and that's why you get the change in firing rate we don't think this is at least the majority cause of the situation here the reason is that if you look at sort of the spike shapes day to day even when the average firing rate is shifting a

6:20lot you still see sort of stable Spike shapes that's obviously not a fully bulletproof story but at least gives some confidence that it's not actually different neurons you're picking up however there's still is very much a chance that that could be the case in at least some part of the of the robustness now stationary story cool thanks yep thanks question yeah let's be clear that like the Electoral position is actually fairly stable because you've got these very tiny basically very tiny wires with with that and then there's some play in the like you've got you've got the advice attached to the skull ritually but then you've got this this long so tiny wire

7:00with kind of a a coiled section so it's a it does tend to basically stay in the same place questions on nerling's Twitter so we'll be interleaving some of those supe wants to ask what could knurling help people with that most people don't realize well I mean so once you're in there you know

7:27there's a lot you could do so you can obviously measure temperature so you could do very early detection of a fever you could not measure pressure I think you probably detect that at the very early the very beginnings of a stroke because you can see sort of like electrical signals starting to go sort of hair wire so there's actually probably a lot of just General Health monitoring that you could do once you're in there you know and and with with very simple sensors

8:11hi you guys all did a great job of distilling a lot of complex engineering and Science and making it wonderfully clear so great job I wanted to ask a little bit about the stimulation I guess for the phosphines and for the the evoked movement how are you thinking is this more like local stimulation is it is it juxta cellular are you steering current around how many cells are you activating how much current are you using I'm just curious what the scale of this is and whether you have a lot of precision or a lot of you know you have pretty profound behavioral effects too hi yeah I'm Dan and how many cells you stimulate with a single electrode is dependent on the

8:51impedance of the electrode size of the conductive pad how much current you deliver frequency all these factors so there's a great deal of variability that we can use to customize the the shape of a phosphine or the or the shape necessarily but maybe the the intensity of the phosphine we think with our current electrodes at least in code back of the envelope calculation would be something like about a 50 to 100 Micron diameter sphere of cells are being stimulated in a visual system the smaller that sphere the smaller and more specific you can make a particular phosphine basically the smaller the pixel in in the image you can produce so there's

9:37plenty of scope for customization of that there's actually also it's possible to get to a much higher like effective pixel count by controlling the the the field electric field between the electrodes so it's not necessarily it's not a one-to-one relationship you could actually dynamically adjust the the field and simulate Farm have a have a very high neuron to electrode ratio so try your like because you get like you know maybe 10 to 100 to 1 potentially so it's a megapixel plug type basically

10:21can you see it normally but I think people would want to know that I think that is one of the one of the possible outcomes pylon this is amazing oh can you talk about the longevity of the implant itself also how would the material of the implant would react with the brain tissue or density of the bone or bone structure thank you

10:42yeah happy to talk about this I'm Jeremy engineer on the brand interfaces team and I think it's good to start with data so like Zach mentioned we have an implant that was you know a monkey who's performing BCI for 617 days that was pager before being upgraded to the latest device for our current version of the device it's lasted for almost a year and then for Accelerated lifetime tester that Josh kind of talked about we have data from our implants from the previous version eight years of accelerated time and from the current version four years of accelerated time and Counting so that's kind of starting with the data those devices are still lasting and still going

11:15theoretically there are kind of three fundamental factors that contribute to the longevity of the device one is going to be the seal the Hermetic enclosure of the device two is going to be the battery and internal electronics and then three is going to be the threads that Zach talked about a little bit and the channels being able to functionally record signals from the brain the seal we think will far Outlast the other two in terms of the bottlenecks so the seal just theoretically I think Josh mentioned that it is a thermoplastic polymer material so there's going to be a a very small amount of moisture that diffuses through it over time and we

11:47think that that will last you know 20 plus years easily in terms of just that property and like I said we have not seen our seals fail with our current version of the device yet so we haven't really pushed the limits here for the battery and internal Electronics that's really based on usage and how much runtime you want and we are working currently on getting data to project out even farther but right now we believe that we can you know achieve 80 run time at the three year time point which would be about you know three and a half hours for a four hour run time but we're like Avinash mentioned we're we're doubling that very soon and quadrupling is what

12:20we have plans to do that so the internal Electronics really aren't the bottleneck either and so really we're attacking the threads themselves and longevity of those channels that Zach Zach can kind of talk about some of the improvements that we're doing to increase that longevity cool thanks yeah so as sort of mentioned before we don't necessarily have an end point as Jeremy said for the testing of the threads that being said we are focusing on longevity because we think this is an important issue to solve so one thing that we're doing in parallel with the current device is aggressively pursuing silicon amorphous silicon carbide insulation of the threads which we

12:58believe will take us well beyond five years of longevity but of course still to be tested and in parallel with that we're just starting to look at Atomic layer deposition which we think could even push longevity of the threads much further and deposit very thin layers to keep the flexibility of the threads and that Advantage there so along with that we're also of course having to design and validate very robust bench top testing to model really in Vivo conditions and look at Channel degradation so that's what we're looking at for longevity of the threads and then I think you asked about biochem pump and I think for biocomp essentially all the materials we're using right now

13:41I can say or at least biostable and we send out testing for biocompatibility very often and essentially what we're doing is we're using in many cases known materials from literature that academic Labs have already started to look at and sort of jumping on that and using that as a starting point thanks for answering that

14:02cool so we have another question from Twitter this is from David and he asked the team what are the biggest lessons you learned since the previous presentation it's been about two years I'm sure there was a lot of engineering done so yeah anyone want to answer what we learned in the last two years so one thing that we've learned in the last couple years is just how much the brain moves on the human scale compared to you know when you start small when you make brain proxies and a lot of research starts with rodents the brain does not move that much and you get a human and the Brain can move like hundreds of microns or more and when our threads and

14:43needles are so small that motion when you zoom in looks like a mile I think to add to that one thing is how I guess Dynamic the implant environment actually is so we've talked about like when those implant site Hills SCAR or new tissue might grow and fill in the space and that'll affect like how our threads might interact in that space so that's why we've emphasized so heavily the importance of Designing accurate proxies so instead of having to wait months for an implant site to heal you can hopefully learn that information in hours I'm Alex on the robotics team I think one of the things we've definitely learned within the engineering teams is

15:30the importance of really continuous validation and testing where we're building say motion systems that are precise to single digit microns we need validation and test systems that we trust even more than that to prove that they work reliably and putting just as much focus into those validation and test systems and designing those alongside our products I think is one thing we've definitely learned

15:58thing and I don't think that we learn I think as part of BCI or the pain control and algorithm is that again building a prototype and making it work with only one monkey one pager was a great maybe a success but also relatively easy to making it work every day for all the other monkeys so actually making it a product is something that it's not easy but we are learning how to do it

16:27I mean I've learned that the brain is really squishy like way squishier than you think it's not like you know cauliflower or broccoli or something like that it's more like water balloon and then it's moving in your skull like a lot so got a squishy water balloon in a coconut is maybe a good way to think of it

16:54hello given Bluetooth bandwidth limitations have you considered other Technologies for wireless communication foreign hey yeah I can take the first part of this question and then I'll let Matt to the second part of it it's a great question especially as you think about how to increase and scale the number of channels that we want to record from this becomes increasingly a bottleneck for the kind of work that we want to do we're thinking about this in a couple ways one is just directly improving the underlying radio interfaces and now let Matt talk about that in a second the other way we're thinking about this is how can you be more efficient with

17:30the data you send off the implant and I think the first version of that is compression so just taking your data looking at the characteristics of it find out a way to represent it more efficiently and just send off that compressed Stream So for reference right now our Bluetooth bandwidth is around 150 kilobytes per second the compressed stream of data that we send off the implant is around 50 kilobytes per second so we're doing fairly well there so far but when you start thinking about 16 000 Channel devices that won't get you all the way there so some other things that can help on the compression slider to actually just send out the output of the machine learning model

18:00rather than the input required to actually run it so one thing we've been trying in the background here is called decodon head which is essentially taking the machine learning models right now we're running on MacBooks that our monkeys are gaming on and moving those to actually run on the implant and this is like a super cool engineering problem if you want to talk about how to make complex neural networks run on what is the equivalent of a garage door opener come talk to me it's fun yeah so that's another way to solve this problem is basically do the computationally Intensive work to just get the raw signal that you actually care to use to control something and

18:33then send that thing out of the implant on the radio side I'll hand it over to Matt yes so to answer your question we are looking at other radio Technologies one in particular is 500 megahertz bandwidth Ultra wideband at a couple different frequencies so this has an advantage in terms of the bit rate that you can achieve it's on the order of six to eight to ten megabit there's also a latency Improvement that's quite substantial and there's also a another wireless technology that we're looking at at w band

19:14hi thank you all for really clear and compelling presentations something that struck me in one of the earliest talks I think it might have been DJs was this vision for the ability to acquire new complex skills via these bcis like the ability to perform Kung Fu and that reflects the fact that the brain is fundamentally a learning machine and yet the many of the Technical Solutions presented later framed were framed in such a way as to try to correct for the way the brain changes over time over longer time skills drift over the course of days or the way that the tissue might heal over time I was curious what your vision collectively was for developing out this

20:03technology that interfaces with a fundamentally plastic system that changes in complex ways over a variety of time skills days months years yeah okay so tough question yeah I think it would be kind of maybe bi-directional learning in some way in the sometimes skills that we will might fix our algorithms and we prefer to have like more stable kind of performance but of course if the over time the person in the brain will learn how to use better to be safe we'll need to update our models so they will be kind of in an Interactive kind of a relationship in some way to learn even new tasks this probably yeah will be something we'll over time will need to

20:56learn what the person kind of learn how to interact with the computer and then build the appropriate interface the ux and also the UI and build algorithms that will help him to control what we want just one thing I'd like to add on to when you're said yeah it actually is an advantage in some ways that the brain is plastic and learns and that can help us because we actually have to do less work in the the human in the loop we'll actually learn how to use our device better but one of the advantages of our particular approach and device is that we are trying to do an extremely high Channel count device so we can you know uniformly distribute electrodes over a

21:31functional region and then it doesn't matter so much whether things move or shift over time we can offload that to software and so we can build algorithms that change over time as well and so both those things are actually I think advantages to our particular approach we have another question from Twitter Juan wants to know what career path do you suggest for somebody that is just getting out of high school if they want to work at neuralink in the future well it's really any of the skills that we described so I mean we're developing new Chips you this

22:06material science you know as software obviously animal care it's a really all the things they were listed in the in the neural link slash careers okay that'd be a good guide yeah I'm actually very fond of saying you know when you flip through any college like booklets and look through all the majors I think you can point to every single one of those majors and there's someone at this company who either is an expert or you know have majored in that so it really is truly truly multi-disciplinary Endeavor and I think you know just focus on whatever your you know passionate about or whatever you're talented at and then just you know pursue that as deeply as you can

22:56and then there's definitely going to be a place for you in you will end up building neural interfaces hey we got to see the monkeys doing telepathy but could you say a little bit more about the animal behavioral training kind of their lives and day-to-day processes sure I'm Autumn I am head of Research Services which includes our Animal Care Program and as an animal welfare scientist this is a topic that I'm deeply interested in so our training

23:30program is outfitted mostly with behavior analysts who help us think about how to remove any of the potential aversives or frustrations from our training we think about conditioning as the primary which includes positive reinforcement as the primary way to train let's see what else can I share with you

24:06yeah yeah I mean that may not be part of the behavioral training itself but we think of Animal Welfare assessment in the framework of the three R's which is refinement replacement and reduction and so when we think about refinement behavioral training does apply in in that way and where we want to remove specifically in research restraint is one of the things we make a very top goal to remove so you saw a lot of videos today where animals were walking up to their stations because we worked really hard to remove any requirement to restrain the animal

24:51anything else yeah I can just yeah well just on top of the last point you said just as an engineer here one of the things that is really inspiring and really cool about this place is that we do get to work on a lot of technological innovations that directly translate to Greater independence with animals when they're engaging in these tasks so as you saw you know monkeys charge just by voluntarily walking up to a branch they play games in their home habitat with a laptop computer voluntarily and the fully implantable fully wireless device the inductive charger all these things enable that kind of experience and so this is one of the very cool Parts about

25:22working here is we do get to innovate on things like that definitely helps to work with a group of Engineers who can like really make cool stuff for for monkeys to be able to do easier behavioral training yeah so I guess to answer the previous question about what you can study to be part of neuraling I guess monkey engineering you can add to add to that Monkey Business hello my question is on upgrade upgradability which you guys mentioned quite a bit so in that procedure I imagine there's some kind of expand procedure and then you're going to putting a new set of implants so could you talk about the damage possible if any tissue damage from the X-Men

26:12procedure how long you have to wait do you imply in the same areas and what's your like brain scanning for the implant procedure in terms of upgrading it I don't know how many questions I can ask so I can start to speak to some of those so I I work a lot on upgradability in those explant processes and designing those to be better the the the goal that we're working towards is that as I mentioned in the presentation it's really just as easy to upgrade an implant as it is to initially install we didn't we didn't show many of those explant examples today but we've come pretty close to just popping out an implant and reinstalling another one in the exact

27:08same location definitely definitely the goal it we are installing the implant in primary motor cortex which is a valuable area for interacting with a device like this and so we the goal is to implant in the same location maybe if you expand out to other applications then you'd be interested in moving somewhere else but we definitely want to be able to insert into the same area in terms of damage the I think that the damage that we care most about is damage within the brain and what we've found and we we talked about that that challenge of the tissue layer on top of the brain and we're I think we're well on our way towards figuring that out but because of the thread small size

27:53the the sort of Scar capsule within the brain is so minimal that they are actually removed quite easily and so we see useful signals even on the second or third time that you've placed an implant and I think some of our BCI folks can probably speak to that we we do have monkey participants working with their second devices and really

28:21making use of those so one two questions one was somebody that asked the question about the brain being plastic have you noticed any plasticity from a behavior perspective from any of the monkeys or is it too soon to tell or there haven't been any observations from the monkey Behavior we see that it takes them a while to learn how to of course to turn on the task but also when they are implanted and it's relatively quickly for them to ramp up and get to a high performance of brain control with pager for example after a few days he was able to like three days already able to learn very quickly to use the device he was trained on the task of first one with his previous implant but

29:17with the new one he was after three four days he was able to control to a close to Performance of he he held with the previous implant but have you noticed anything on the adverse side which means the brain has outpaced the neural network that you're running it's hard to say no not really okay so I have a another question which is more about the electrical side so you talked about 1024 channels being recording are you transmitting the raw signal or was it only the three Spike events that you were talking about the low mid and the high or is it the raw entire raw film waveform that you transmit yeah hi I'm Julian I can speak a bit about this and maybe Avinash wants to

30:06contribute but our chips see the real signals but then when we transmit out typically spikes and we detect those spikes in real time on the chip this massively compresses the data I guess yeah we're making improvements to that but we can request we can request rule samples sometimes we also process particular statistics or other data directly on the chip and then send out the calculated values so there are many ways to sort of play with the data yeah so at least with the current and one system that we have which relies on ble radio there is a bandwidth limitations so you can't actually stream raw data from all 1024 channels but just kind of to give you a little bit of a

30:48history of how our compression algorithm the spike detection algorithm was developed we did have sort of a wire system there was a paper that we published with the USB C connector that you know streams all those signals through a hype bandwidth wire connection so we did have kind of those development platforms to be able to see the raw signals and know which set of information that we want to extract that are you know going to fit within the bandwidth of the radio as well as is useful for BCI control and you know also just sending data wirelessly does cause a lot of energy so there's any opportunities we have to reduce that burden you know we try to do

31:30basically have all that compression closer to where the electrodes are as possible maybe one thing that isn't obvious is that the the actual bit rate that you need to control a phone or a computer is actually very very low so I think we might have the record for bitrate is that correct we think we do maybe so on the order of 10 bits per second so that's super slow

32:01but if you think like if when you're inputting data into a phone like how fast your thumb is moving how many thumbs what's your thumb Taps per second pretty pretty low and I mean like basically our thumbs are like two slow-moving meat sticks that we you know do this and it's like this is really a load it's like a low bar that's what I'm saying so for at least for output it's it's a you get 10 bits per second you're you're hauling ass so and that's you don't need your Bluetooth anything for you so you could practically send it out with beeps and Buffs you know so it's not if you go if you're going like a high bandwidth visual now you're you know maybe going to

32:45megabit plus but it's it's so it's really well within Bluetooth or wait but anyway it's just we're that is what I'm saying is that's not that's not a constraint that the data rate one of the sort of like maybe notable item which we talked about in the presentation but we we think we can probably solve for doing the implant without cutting the the dura we can just do basically a bunch of holes through the door which is like the door is like the big thick iron dry D thing that contains the that's up against the skull if you don't pierce the dura you know if you don't cut the door away and instead you have a bunch of tiny holes and insert the electrodes through the tiny

33:30holes into the rain and then the recovery time is ridiculously fast you know you're not really losing much in the way of cerebral spinal fluid it's it's you couldn't Theory we I mean this could be like a the whole thing could be a 10 minute operation like LASIK like it's fast it's not like a big laborious thing it's super fast

34:00just going back to the long-term use I'm wondering if you have any pathology looking at scar tissue from many animals that have had long-term implants and along that lines it seems like there might be a little bit of a gap between use in medical conditions and helping individuals from a safety perspective

34:30so I didn't quite catch the last question but I Heard the first one I'll ask you to repeat the the second one so the first one is do we have pathology from from long-term use animals we absolutely do we don't have any pathology from our monkeys which we upgrade and you know are still going we have other studies that are primarily to determine safety and so we do have histopathologic endpoints that we we determine the scar tissue formation around the threads themselves in the brain is typically negligible like it barely reacts to the threats at all so that that's very promising in terms of the scar tissue formation over the the cortex so this neo-membrane

35:03growth that fills in the areas that Elon and Alex were mentioning that we remove with our current operation those we we do you know evaluate that scar tissue but it isn't it doesn't pose a problem in any way it's not a continuous reaction to a foreign body it's just filling in tissue that was removed and if you could repeat the the second question I didn't hear that yeah the second question really following up on that seems like there might be a little bit of a gap in use and healthy and individuals from a safety perspective you know I think people mentioned that they might be interested in trying prototypes but just wondering what your perspective is

35:39on trying to lower the safety risks yeah it's a great question so in terms of really it's about the long-term use of the device so you know we have devices that have been implanted like I said in monkeys where you know for many years where we see no behavioral deficit at all so this first is a question of how you evaluate safety so you have histopathologic endpoints you can evaluate but we're also looking for cognitive deficits or behavioral deficits as well and we don't see any of those in our animals which is you know an important point in terms of the histopathologic endpoints they look really really great the challenge is one of explanating the device which is why

36:13we're putting so much effort into the reversibility efforts and our through Dura insertions so when removing the device that's when you potentially could cause damage and so we are doing we have a lot of ongoing studies right now to really minimize the risk of that but we don't think it's a substantial risk with our current approach and like I said pager was upgraded with the previous surgical approach and is doing great so clearly it is you know can be perfectly safe but proving that beyond a shadow of a doubt for humans is something that we're still working to do rigorously to answer your question yeah thank you

36:46so thank you for a very deep dive on many of the different aspects of the device and the system it's very impressive to see all the engineering work that's gone into it you just mentioned about bitrate as the prior bit rate holder I can confirm you have indeed shattered my record so congratulations on I think I saw a peak of 7.4 bits per second well done my question is actually around the clinical trials and the FDA to the extent that you can share I gather that that device removal or maybe electrode removal is one of the concerns that the FDA highlighted is there anything else you can tell us about what the FDA was concerned about or had questions about

37:28with respect to their ID submission yeah I mean we can probably talk a little bit I mean it's these are really challenges you know that we have broadly so explanation safety proving that right rigorously for humans is something that we it definitely is one Challenge and was something that that the FDA commented on other things that they do ask some really great questions so other things that involve you know things like the thermal bench shop testing of our implants so obviously it's important that our implant doesn't damage the tissue by overheating so having really rigorous and valid bench shop testing for that is very important it's actually

38:02Something That We're redesign to be even more accurate now it's also the case that you know they ask a lot of very hard questions on biocompatibility chemical characterization so we've done very rigorous testing for that but you know they they do ask a lot of questions about getting into the weeds of the data and making sure that there really is no chance for any toxic chemicals or bio incompatible materials to be in the brain so these are all things we're working with you know to to just prove again above and beyond Beyond a shot of a doubt one thing that's maybe worth mentioning here is that it can be difficult to appreciate the novelty of

38:34our product so the surgical robot and the thin film array in particular are quite new and unlike existing devices and this means that we can't rely heavily on literature to support the safety and efficacy of the device so we do spend a ton of effort in designing and Performing testing on our devices so that we can rigorously prove the safety of them and we can't rely just on another product or on some paper and that's something that we're not willing to compromise for our first human participant working very hard to do I think if you ask a question like like in my opinion like would I be comfortable in planning this in someone you know one of my kids or something

39:10like that if at this point like if if they're in a serious like let's say they if they broke their neck what I feel comfortable right now doing it I would I would say we're at the point where I at least in my opinion it would not be dangerous hi thank you for the presentation so I have a non-technical question are you collaborating with people with motor disabilities and if so like have they shared any ideas of applications that they would be excited about

39:51production I can take the first part of this I'm not the best person to speak to this to be honest but there is a consumer Advisory Board we have made up of a number of people that have various conditions including tetraplegia and they give advice to us on a number of topics there's just as an antidote someone came to the office maybe six months ago and they were telling me what they most wanted to do with their neuralink device and there were two things that they said one was they wanted to be able to trade stocks day to day to be able to beat their brother and the second one was they want to be able to play shooter games so I think what was most shocking to me

40:25about that encounter was the normalcy of that and I found that conversation truly inspiring so you know who you are the person who came and talked with me have a great day yeah yeah what something that's we've talked about but it's maybe I should be re-emphasized we are doing we're we're building up a production system for the devices so we're we're building up breaking out the production line making large numbers of devices we want to make thousands ultimately tens of thousands then millions of devices so the the progress at first particularly as it applies to humans will seem perhaps agonizingly slow but we're doing all of the things necessary to bring it

41:13to scale in parallel so Theory it should progress should be exponential okay so thank you that was a very cool presentation so one of the stated goals was recording from everywhere in the brain being able to record from and perturb any location so it seems like currently it's all cortical and I'm curious with the current device is it is there any sort of long-term goal or idea as to extending it into going deeper in the brain I mean for Neuropsychiatric disorders for memory all these things are much deeper several centimeters so I'm wondering what's the time scale if you were to give a very rough estimate of when I can expect to see a knurling product that

41:56goes that deep yeah so I mean the the fundamentals of the device in the skull will stay essentially the same because the as I said earlier the the device this call is very much like a like a smart watch essentially it's got It's a battery radio inductor charger computer and and then you've got the the little wires and so you need to make the wires longer and you'd have to have a deeper insertion needle for the robot but it this really is intended to be a generalized i o device so apart from the tiny wires being longer and the surgical robot needing a longer needle in theory you should be able to go anywhere because it seems to me that part of

42:50the robot is trying to detect where the blood vessels are and then avoid them correct would that be possible at that scale I mean certainly not just visually but maybe there's some other way of detecting it is that is that like is that a current goal and do you expect that within I suppose the next decade oh definitely yes I'm eating I run the Robotics and surgery engineering team here like of the three axes that DJ mentioned one of them is you know hack system Warriors of the brain so the robot team thinks about this a ton in terms of what sensors do you need to essentially go past the surface and so in this case you're right that right now we can really only see down

43:27the maximum about a millimeter I think within the team there's questions of what's best to use next but like ultrasound and photo acoustic tomography are two that come to mind as things that can get centimeters deep essentially but it's a super interesting problem you sort of need Deep Imaging and some ability to steer to at least avoid large vasculature deep down

43:51yeah or if our if we can make our needles and threads small enough in a way that we can still be precise and accurate at a deep depth then maybe you don't cause a bleed if you hit a vessel yeah I think that's that's really the ideal situation if the threads are really tiny they can actually go through a blood vessel and it's and it's okay if they're tiny enough so so we wouldn't need the blood vessel Imaging in that case I I'm I actually am slightly optimistic that that is achievable and Matt could probably speak more into this but what DBS currently is kind of just like send it yeah the current approach involves a wire that you blindly pass in that's

44:38massive compared to our threads orders of magnitude bigger and so that's a low bar for us to clear as well because people don't realize like before the Deep Red simulation just how big the hole is it's a I mean what is it like I mean how basically in current deep brain stimulation how much of a borehole is drilled in the brain yeah you're drilling a 14 millimeter Burr hole and then passing a two millimeter wire you know six centimeters eight centimeters deep into the brain so all blindly hoping that you don't hit a blood vessel telling the patient up front this might be good for you and there's a one percent chance you're gonna your brain is gonna bleed in a way

45:24we can't control so that is current technology that is happening right now so doing better than that is we can we can definitely do way better than that there's no problem our needle is 40 microns yeah exactly thanks again for the phenomenal presentation I thought it was fascinating how rapidly you could test all of these electrodes but it begs the question about like what your fault tolerance is if you run these Diagnostics and it comes back that you have something that's either shorted or high Z how many of those before you get degraded performance the second question is when you're actually inserting this device we saw examples of the electrode going in and then like

46:06looping back on itself but it looked like that was something that was assessed basically by slicing the synthetic material I'm curious what you're doing to validate the insertion of all these electrodes sort of in Vivo how do we know that that's not happening on an actual patient yeah I can enter the SEC the second one so like I mentioned we can so we weren't actually sectioning in that case we have a really cool micro CT so I mean it's essentially like a CT scanner so that's just an intact proxy that we put in this machine and we can you know take a picture all the way through it and like I mentioned before like we can make a proxy where it happens you

46:45know that that looping back happens every single time and then we can make one where it never happens and we've pinpointed roughly now where actual tissue Falls in there and so our current plan for you know validating and confirming that is making proxies where it you know happens really easily much worse case than any you know any tissue could possibly be and then designing it such that it never happens in that scenario and that'll give us the doing that enough times and with a weak enough proxy that'll give us the confidence that this isn't actually happening

47:18yeah and yeah and this is the next gen you we don't see this problem at all with the the current generation and I'll take a stab at your first question so to clarify you asking what happens if there's a fault on a particular Channel or something yeah that's correct yeah so the nominal scenario is that basically the impedance will stabilize pretty quickly within the brain and even at that level we can record great signals we see lots of spikes and we can use that for BCI because we have so many channels like a thousand now 16 000 later we can actually run our models with far less channels than we actually have so it doesn't matter if like one channel dies

47:55here or there like we can still do really good decode I'm not sure if we have official numbers on how many channels we need but it's like we have an order of magnitude more and the more we have about it but we can already do a lot with what we have all right maybe just one or two more questions yeah I have a question about your very very long-term inspiration to have this high band with communication with Advanced AIS so it seems like you know the advanced AI would need to understand the humans most complex thoughts and emotions and that's what neuroscientists are trying to do so do you have any Ambitions to tackle Neuroscience Beyond neuro engineering

48:44well I mean I think we're going to make the input output device and the software interface with it and I think probably you know pose suggestion earlier we'll try to open source as much as possible so people can take a look at it and I think there will be a lot of others that that build upon the work that we're doing you know the same way that if you make a if you make a microprocessors or CPU or computer that people will write lots of software that runs on that computer so but if you don't have the computer this is the software's moved so we're making the input output device with the computer and and then I think probably there there'll probably be a lot of other organizations

49:28companies that that built that built upon that Foundation so yeah I mean one of the things that's I would sometimes wonder is that if you do have a whole brain interface and you can record memories and really getting into Black Mirror stuff here but this could be one of them I I also think it's worth mentioning an important point which is that you're only didn't come out of nothing there's decades and Decades of research in the medical academic field that has really set the foundation for what is possible by putting these electrodes in parts of the brain and being able to read those signals decode it for mapping it through some application and

50:23you know being being in Academia before before coming to neuralink you know I do think that there's a lot of opportunities for kind of the field to advance at a much rapid rate by having just better tools for observing the Dynamics that are happening and then engaging with it in a seamless way and I think it was Ian who sort of mentioned that you know it's almost as if like we're kind of building an oscilloscope for the brain which I think is like kind of a beautiful analogy of just giving us a bit more abilities into peering into the Dynamics and using those information learned that to I don't know hopefully understand like what makes us and how

51:00the brain works and know the whole champagne hello the presentation covered keyboard and handwriting based input methods how do you plan to develop an input model that will achieve much higher bandwidths for complex tasks in humans yeah this is a tough question and we start to exploring this with monkeys as you saw we have like a multiple we train many monkeys on very different tasks it's still an open question that we're after I think hopefully once we get to our first participant it will be easier to investigate one of the options we are exploring as we showed is to the code handwriting directly this is a one a work that started Stanford and we are

51:44exploring here and trying to expand there's also a different in addition to just decoding different things from the brain we also try to provide the user different maybe user like interfaces for example we show different type of keyboards maybe also swipe and other things that can help increase the communication rate so we are kind of attacking those in two dimensions yeah just one other thing to add in that direction as pointed out by many people here so far this is a general i o system that you can sort of plug and play in different places in the brain there's other areas of the brain that can help increase bandwidth for example language wrist speech centers that can

52:22help you much more seamlessly communicate for example text if that's your main thing that you're trying to do yeah just I think just having this General input output device will just so gigantically improve our understanding of the brain as far as is hard to the words can barely Express like you know right now we're just guessing a lot of what's going on in the brain but if you have direct i o it's not no more guessing what would learn about the what we will learn about the brain with such a device in in wide use is absolutely or many orders of magnitude more than we currently understand so I guess on that on that note thank you for coming and thank you for watching online

资源与出处

相关视频

同频道、同主题或相邻年份