埃隆·马斯克：Neuralink 与人类的未来

The following is a conversation with Elon Musk, DJ Seo, Matthew MacDougall, Bliss Chapman, and Noland Arbaugh about Neuralink and the future of humanity.

以下是与 Elon Musk、DJ Seo、Matthew MacDougall、Bliss Chapman 和 Noland Arbaugh 关于 Neuralink 和人类未来的对话。

00:00 Elon, DJ, Matthew and Bliss are of course part of the amazing Neuralink team, and Noland is the first human to have a Neuralink device implanted in his brain.

Elon、DJ、Matthew 和 Bliss 当然都是令人惊叹的 Neuralink 团队的成员，而 Noland 是第一个在大脑中植入 Neuralink 设备的人。

I speak with each of them individually, so use timestamps to jump around, or as I recommend, go hardcore, and listen to the whole thing.

我会单独与他们每个人交谈，因此请使用时间戳来跳转，或者按照我的建议，进行硬核，并聆听整个内容。

This is the longest podcast I’ve ever done.

这是我做过的最长的播客。

It’s a fascinating, super technical, and wide-ranging conversation, and I loved every minute of it.

这是一次引人入胜、技术含量高、范围广泛的对话，我喜欢其中的每一分钟。

And now, dear friends, here’s Elon Musk, his fifth time on this, the Lex Fridman podcast,

现在，亲爱的朋友们，这是埃隆·马斯克，他第五次参加莱克斯·弗里德曼播客，

Drinking coffee or water?

喝咖啡还是水？

Water.

水。

00:51 I’m so over-caffeinated right now.

我现在咖啡因过量了。

Do you want some caffeine?

你想要一些咖啡因吗？

Sure.

当然。

There’s a Nitro drink.

有一种硝基饮料。

This supposed to keep you up for like tomorrow afternoon, basically.

基本上，这应该能让你在明天下午保持清醒。

Yeah.

是的。

01:08 Yeah.

是的。

I don’t want to [inaudible 00:01:11].

我不想 [听不清 00:01:11]。

So what is Nitro?

那么什么是硝基呢？

01:11 It’s just got a lot of caffeine or something?

只是含有大量咖啡因或其他什么吗？

Don’t ask questions.

不要问问题。

01:13 It’s called Nitro.

它被称为硝基。

Do you need to know anything else?

您还需要了解其他信息吗？

It’s got nitrogen in it.

里面有氮气。

01:17 That’s ridiculous.

这太荒谬了。

What we breathe is 78% nitrogen anyway.

无论如何，我们呼吸的是 78% 的氮气。

What do you need to add more for?

你还需要添加更多做什么？

01:24 Unfortunately, you’re going to eat it.

不幸的是，你会吃掉它。

01:29 Most people think that they’re breathing oxygen and they’re actually breathing 78% nitrogen.

大多数人认为他们呼吸的是氧气，但实际上他们呼吸的是 78% 的氮气。

You need like a milk bar, like from Clockwork Orange.

你需要像发条橙那样的牛奶吧。

Yeah.

是的。

01:41 Yeah.

是的。

Is that the top three Kubrick film for you?

这是您心目中排名前三的库布里克电影吗？

Clockwork Orange?

发条橙？

01:44 It’s pretty good.

非常好。

It’s demented.

这是疯狂的。

Jarring, I’d say.

我想说，很刺耳。

Okay.

好的。

01:49 Okay.

好的。

So, first, let’s step back, and big congrats on getting Neuralink implanted into a human.

所以，首先，让我们退一步，热烈祝贺 Neuralink 被植入人体。

That’s a historic step for Neuralink.

这对 Neuralink 来说是历史性的一步。

Thanks.

谢谢。

Yeah.

是的。

And there’s many more to come.

还有更多的事情即将发生。

Yeah.

是的。

02:07 And we just obviously have our second implant as well.

显然我们也有第二个植入物。

How did that go?

进展如何？

So far, so good.

到目前为止，一切都很好。

02:12 It looks like we’ve got, I think, on the order of 400 electrodes that are providing signals.

我认为，我们似乎拥有大约 400 个提供信号的电极。

Nice.

好的。

Yeah.

是的。

How quickly do you think the number of human participants will scale?

您认为人类参与者的数量会多快扩大？

It depends somewhat on the regulatory approval, the rate at which we get regulatory approvals.

这在一定程度上取决于监管部门的批准，以及我们获得监管部门批准的速度。

02:28 So, we’re hoping to do 10 by the end of this year, total of 10.

所以，我们希望在今年年底前完成 10 个，总共 10 个。

So, eight more.

那么，还有八个。

And with each one, you’re going to be learning a lot of lessons about the neurobiology of the brain, everything.

在每一本书中，你都将学到很多关于大脑神经生物学的课程，一切。

02:42 The whole chain of the Neuralink, the decoding, the signal processing, all that kind of stuff.

Neuralink 的整个链条，解码，信号处理，所有这些东西。

Yeah.

是的。

02:54 Yeah.

是的。

I think it’s obviously going to get better with each one.

我认为显然每一个都会变得更好。

I don’t want to jinx it, but it seems to have gone extremely well with the second implant.

我不想给它带来厄运，但第二次植入似乎进展得非常好。

So, there’s a lot of signal, a lot of electrodes.

所以，有很多信号，很多电极。

It’s working very well.

它运作得很好。

What improvements do you think we’ll see in Neuralink in the coming, let’s say, let’s get crazy, the coming years.

您认为未来几年我们会在 Neuralink 中看到哪些改进，让我们疯狂一下吧。

In years, it’s going to be gigantic, because we’ll increase the number of electrodes dramatically.

几年后，它将会变得巨大，因为我们将大幅增加电极的数量。

03:18 We’ll improve the signal processing.

我们将改进信号处理。

So, even with only roughly, I don’t know, 10, 15% of the electrodes working with Noland, with our first patient, we were able to get to achieve a bit per second.

因此，即使只有大约（我不知道）10%、15% 的电极与 Noland 配合使用，我们的第一个患者也能够达到每秒一点的速度。

That’s twice the world record.

这是世界纪录的两倍。

So, I think we’ll start vastly exceeding the world record by orders of magnitude in the years to come.

因此，我认为我们将在未来几年开始大大超过世界纪录几个数量级。

So, start getting to, I don’t know, 100 bits per second, thousand.

所以，开始达到，我不知道，每秒 100 位，千位。

Maybe if five years from now, we might be at a megabit, faster than any human could possibly communicate by typing, or speaking.

也许五年后，我们的速度可能会达到兆位，比任何人通过打字或说话进行交流的速度都要快。

Yeah.

是的。

04:06 That BPS is an interesting metric to measure.

BPS 是一个有趣的衡量指标。

There might be a big leap in the experience once you reach a certain level of BPS.

一旦您达到一定的 BPS 水平，体验可能会有很大的飞跃。

Yeah.

是的。

Like entire new ways of interacting with a computer might be unlocked.

就像与计算机交互的全新方式可能会被解锁一样。

And with humans.

和人类。

With other humans.

与其他人类。

Provided they have want a Neuralink, too.

前提是他们也想要 Neuralink。

Right.

正确的。

Otherwise they wont be able to absorb the signals fast enough.

否则他们将无法足够快地吸收信号。

Do you think they’ll improve the quality of intellectual discourse?

您认为它们会提高知识分子话语的质量吗？

Well, I think you could think of it, if you were to slow down communication, how do you feel about that?

嗯，我想你可以想一下，如果你放慢沟通速度，你感觉如何？

04:34 If you’d only talk at, let’s say one-tenth of normal speed, you’d be like, “Wow, that’s agonizingly slow.”

如果你只以正常速度的十分之一说话，你会说，“哇，太慢了。”

Yeah.

是的。

So, now imagine you could communicate clearly at 10, or 100, or 1,000 times faster than normal.

所以，现在想象一下，您可以以比平常快 10 倍、100 倍或 1,000 倍的速度进行清晰的沟通。

Listen, I’m pretty sure nobody in their right mind listens to me at 1X.

听着，我很确定没有一个头脑正常的人会在 1X 听我说话。

05:00 they listen at 2X.

他们以 2X 的速度收听。

I can only imagine what 10X would feel like, or I could actually understand it.

我只能想象10X会是什么样的感觉，或者我实际上可以理解它。

I usually default to 1.5X.

我通常默认为1.5X。

05:14 You can do 2X.

你可以做2X。

Well actually, if I’m listening to somebody get to…

好吧，实际上，如果我正在听某人说话……

in 15, 20 minutes, I want to go to sleep, then I’ll do it 1.5X.

15、20 分钟后，我想睡觉，然后我会做 1.5 次。

If I’m paying attention, I’ll do 2X.

如果我注意的话，我会做 2 次。

Right.

正确的。

But actually, if you actually listen to podcasts, or audiobooks or anything at…

但实际上，如果你真的听播客、有声读物或其他任何东西......

05:32 If you get used to doing it at 1.5, then one sounds painfully slow.

如果您习惯以 1.5 进行操作，那么听起来会慢得令人痛苦。

I’m still holding onto one, because I’m afraid, I’m afraid of myself becoming bored with the reality, with the real world, where everyone’s speaking in 1X.

我仍然坚持着一个，因为我害怕，我害怕自己对现实、对现实世界感到厌倦，每个人都在用 1X 说话。

Well, it depends on the person.

嗯，这取决于人。

05:53 You can speak very fast.

你可以说得很快。

Like we can communicate very quickly.

就像我们可以很快沟通一样。

And also, if you use a wide range of…

而且，如果您使用多种……

if your vocabulary is larger, your effective bit rate is higher.

如果您的词汇量越大，您的有效比特率就越高。

That’s a good way to put it.

这是一个很好的表达方式。

Yeah.

是的。

The effective bit rate.

有效比特率。

06:07 That is the question, is how much information is actually compressed in the low bit transfer of language?

这就是问题，在语言的低位传输中实际上压缩了多少信息？

Yeah.

是的。

06:15 If there’s a single word that is able to convey something that would normally require, I don’t know, 10 simple words, then you’ve got maybe a 10X compression on your hands.

如果有一个单词能够传达通常需要（我不知道）10 个简单单词的内容，那么您可能会得到 10 倍的压缩。

And that’s really like with memes.

这真的很像模因。

Memes are like data compression.

模因就像数据压缩。

You’re simultaneously hit with a wide range of symbols that you can interpret, and you get it faster than if it were words, or a simple picture.

你同时会遇到各种各样可以解释的符号，而且你比文字或简单的图片更快地理解它。

And of course, you’re referring to memes broadly like ideas.

当然，你指的是模因，就像想法一样。

Yeah.

是的。

06:52 There’s an entire idea structure that is like an idea template, and then you can add something to that idea template.

有一个完整的想法结构，就像一个想法模板，然后您可以向该想法模板添加一些内容。

But somebody has that pre-existing idea template in their head.

但有些人的头脑中已有预先存在的想法模板。

So, when you add that incremental bit of information, you’re conveying much more than if you just said a few words.

因此，当您添加增量信息时，您传达的信息比只说几句话要多得多。

It’s everything associated with that meme.

这是与该模因相关的一切。

You think there’ll be emergent leaps of capability as you scale the number of electrodes?

您认为随着电极数量的增加，能力会出现突飞猛进吗？

Yeah.

是的。

Do you think there’ll be an actual number where just the human experience will be altered?

你认为会有一个实际数字改变人类的体验吗？

Yes.

是的。

What do you think that number might be?

你认为这个数字可能是多少？

07:27 Whether electrodes, or BPS?

是电极，还是BPS？

We of course, don’t know for sure, but is this 10,000, 100,000?

我们当然不确定，但是这是一万还是十万？

Yeah.

是的。

07:37 Certainly, if you’re anywhere at 10,000 bits per second, that’s vastly faster than any human can communicate right now.

当然，如果您处于每秒 10,000 位的速度，这比任何人类现在的通信速度都要快得多。

If you think what is the average bits per second of a human, it is less than one bit per second over the course of a day.

如果您想一下人类每秒的平均比特数是多少，那么一天中每秒的平均比特数还不到 1 比特。

Because there are 86,400 seconds in a day, and you don’t communicate 86,400 tokens in a day.

因为一天有 86,400 秒，而你一天不交流 86,400 个代币。

Therefore, your bits per second is less than one, averaged over 24 hours.

因此，24 小时内平均每秒的比特数小于 1。

It’s quite slow.

这很慢。

08:04 And now, even if you’re communicating very quickly, and you’re talking to somebody who understands what you’re saying, because in order to communicate, you have to at least to some degree, model the mind state of the person to whom you’re speaking.

现在，即使你的沟通速度非常快，并且你正在与一个理解你所说内容的人交谈，因为为了沟通，你必须至少在某种程度上模拟与你交谈的人的心理状态。

Then take the concept you’re trying to convey, compress that into a small number of syllables, speak them, and hope that the other person decompresses them into a conceptual structure that is as close to what you have in your mind as possible.

然后将你想要传达的概念压缩成几个音节，说出它们，并希望对方将它们解压成一个尽可能接近你脑海中的概念结构。

Yeah.

是的。

08:34 There’s a lot of signal loss there in that process.

在此过程中存在大量信号丢失。

Yeah.

是的。

08:37 Very lossy, compression, and decompression.

非常有损，压缩，解压缩。

And a lot of what your neurons are doing is distilling the concepts down to a small number of symbols, or say syllables that I’m speaking, or keystrokes, whatever the case may be.

你的神经元正在做的很多事情是将概念提炼成少量符号，或者说我正在说的音节，或者击键，无论是什么情况。

So, that’s a lot of what your brain computation is doing.

所以，这就是你的大脑计算正在做的很多事情。

Now, there is an argument that that’s actually a healthy thing to do, or a helpful thing to do because as you try to compress complex concepts, you’re perhaps forced to distill what is most essential in those concepts, as opposed to just all the fluff.

现在，有一种观点认为，这实际上是一件健康的事情，或者是一件有用的事情，因为当你试图压缩复杂的概念时，你可能被迫提炼出这些概念中最本质的东西，而不是所有的废话。

So, in the process of compression, you distill things down to what matters the most, because you can only say a few things.

所以，在压缩的过程中，你将事情提炼为最重要的事情，因为你只能说几件事。

09:27 So that is perhaps helpful.

所以这也许是有帮助的。

I think we’ll probably get…

我想我们可能会得到……

If our data rate increases, it’s highly probable it will become far more verbose.

如果我们的数据速率增加，它很可能会变得更加冗长。

Just like your computer, when computers had…

就像你的电脑一样，当电脑有...

My first computer had 8K of RAM, so you really thought about every byte.

我的第一台计算机有 8K RAM，所以你真的考虑到了每个字节。

And now you’ve got computers with many gigabytes of RAM.

现在您已经拥有了具有许多 GB RAM 的计算机。

So, if you want to do an iPhone app that just says, “Hello world,” it’s probably, I don’t know, several megabytes minimum, a bunch of fluff.

所以，如果你想做一个 iPhone 应用程序，只说“Hello world”，我不知道，它可能至少有几兆字节，一堆废话。

But nonetheless, we still prefer to have the computer with the more memory and more compute.

但尽管如此，我们仍然更喜欢拥有更多内存和更多计算能力的计算机。

10:09 So, the long-term aspiration of Neuralink is to improve the AI human symbiosis by increasing the bandwidth of the communication.

因此，Neuralink 的长期愿望是通过增加通信带宽来改善 AI 人类共生。

Because even if…

因为即使……

In the most benign scenario of AI, you have to consider that the AI is simply going to get bored waiting for you to spit out a few words.

在人工智能最良性的情况下，你必须考虑到人工智能只是会无聊地等待你吐出几句话。

If the AI can communicate at terabits per second, and you’re communicating at bits per second, it’s like talking to a tree.

如果人工智能能够以每秒太比特的速度进行通信，而你也以每秒比特的速度进行通信，那么就像与一棵树交谈一样。

Well, it is a very interesting question for a super intelligent species, what use are humans?

那么，对于一个超级智慧物种来说，这是一个非常有趣的问题，人类有什么用呢？

I think there is some argument for humans as a source of will.

我认为人类作为意志的来源是有争议的。

Will?

将要？

Will, yeah.

威尔，是的。

11:00 Source of will, or purpose.

意志或目的的来源。

So if you consider the human mind as being…

因此，如果你认为人类的思想是……

Essentially there’s the primitive, limbic elements, which basically even reptiles have.

本质上是原始的边缘元素，基本上连爬行动物也有。

And there’s the cortex, the thinking and planning part of the brain.

还有皮质，大脑的思考和计划部分。

Now, the cortex is much smarter than the limbic system, and yet is largely in service to the limbic system.

现在，皮质比边缘系统聪明得多，但主要为边缘系统服务。

It’s trying to make the limbic system happy.

它试图让边缘系统快乐。

The sheer amount of compute that’s gone into people trying to get laid is insane, without actually seeking procreation.

人们在试图上床时投入的计算量是疯狂的，而不是真正寻求生育。

They’re just literally trying to do this simple motion, and they get a kick out of it.

他们只是在尝试做这个简单的动作，并且从中得到了乐趣。

So, this simple, which in the abstract, rather absurd motion, which is sex, the cortex is putting a massive amount of compute into trying to figure out how to do that.

所以，这个简单的、抽象的、相当荒谬的运动，即性，皮层正在投入大量的计算来试图弄清楚如何做到这一点。

So like 90% of distributed compute of the human species is spent on trying to get laid, probably.

因此，人类 90% 的分布式计算可能都花在了上床上。

11:55 A large percentage.

很大比例。

A massive amount.

数量巨大。

12:00 Yes.

是的。

Yeah.

是的。

Yeah.

是的。

There’s no purpose to most sex except hedonistic.

大多数性行为除了享乐之外没有任何目的。

It’s a sort of joy, or whatever, dopamine release.

这是一种快乐，或者无论如何，多巴胺的释放。

Now, once in a while, it’s procreation, but for modern humans, it’s mostly recreational.

现在，偶尔，它是繁殖，但对于现代人类来说，它主要是娱乐。

And so, your cortex, much smarter than your limbic system, is trying to make the limbic system happy, because the limbic system wants to have sex, or wants some tasty food, or whatever the case may be.

因此，你的大脑皮层比边缘系统聪明得多，它正在努力让边缘系统快乐，因为边缘系统想要做爱，或者想要一些美味的食物，或者无论是什么情况。

12:31 And then that is then further augmented by the tertiary system, which is your phone, your laptop, iPad, whatever, all your computing stuff.

然后，第三系统进一步增强了这一点，第三系统是你的手机、你的笔记本电脑、iPad，等等，你所有的计算设备。

That’s your tertiary layer.

那是你的第三层。

So, you’re actually already a cyborg.

所以，你实际上已经是一个机器人了。

You have this tertiary compute layer, which is in the form of your computer with all the applications, or your compute devices.

您拥有这个第三计算层，它的形式是带有所有应用程序的计算机或计算设备。

And so, in the getting laid front, there’s actually a massive amount of digital compute also trying to get laid, with Tinder and whatever.

因此，在上床方面，实际上有大量的数字计算也试图通过 Tinder 等上床。

Yeah.

是的。

13:04 So, the compute that we humans have built is also participating.

因此，我们人类构建的计算也在参与。

Yeah.

是的。

13:09 There’s like gigawatts of compute going into getting laid, of digital compute.

大约有千兆瓦的数字计算投入使用。

Yeah.

是的。

13:14 What if AGI was-

如果 AGI 是——

This is happening as we speak.

就在我们说话的时候，这一切正在发生。

…

……

13:19 if we merge with AI, it’s just going to expand the compute that we humans use-

如果我们与人工智能合并，它只会扩展我们人类使用的计算能力——

Pretty much.

差不多了。

…

……

13:24 to try to get laid.

尝试上床。

Well, it’s one of the things.

嗯，这是其中之一。

13:25 Certainly, yeah.

当然，是的。

Yeah.

是的。

But what I’m saying is that, yes, is there a use for humans?

但我想说的是，是的，人类有什么用处吗？

13:29 Well, there’s this fundamental question of what’s the meaning of life?

好吧，有一个基本问题：生命的意义是什么？

Why do anything at all?

为什么要做任何事情？

And so, if our simple limbic system provides a source of will to do something, that then goes through our cortex, that then goes to our tertiary compute layer, then I don’t know, it might actually be that the AI, in a benign scenario, is simply trying to make the human limbic system happy.

因此，如果我们简单的边缘系统提供了做某事的意愿来源，然后穿过我们的皮层，然后进入我们的第三计算层，那么我不知道，在良性的情况下，人工智能实际上可能只是试图让人类边缘系统快乐。

Yeah.

是的。

14:03 It seems like the will is not just about the limbic system.

看来意志不仅仅与边缘系统有关。

There’s a lot of interesting, complicated things in there.

里面有很多有趣、复杂的东西。

We also want power.

我们也想要权力。

That’s limbic too, I think.

我认为这也是边缘系统。

But then we also want to, in a kind of cooperative way, alleviate the suffering in the world.

但我们也想以一种合作的方式减轻世界上的苦难。

Not everybody does.

不是每个人都这样。

14:19 But yeah, sure, some people do.

但是，是的，当然，有些人会这样做。

As a group of humans, when we get together, we start to have this kind of collective intelligence that is more complex in its will than the underlying individual descendants of apes, right?

作为一个人类群体，当我们聚集在一起时，我们开始拥有这种集体智慧，其意志比猿类的潜在个体后代更复杂，对吗？

Sure.

当然。

So there’s other motivations, and that could be a really interesting source of an objective function for AGI?

那么还有其他动机，这可能是 AGI 目标函数的一个非常有趣的来源？

Yeah.

是的。

14:45 There are these fairly cerebral, or higher level goals.

有这些相当理智的或更高层次的目标。

For me, it’s like, what’s the meaning of life, or understanding the nature of the universe, is of great interest to me, and hopefully to the AI.

对我来说，生命的意义是什么，或者理解宇宙的本质是什么，对我很感兴趣，希望对人工智能也很感兴趣。

And that’s the mission of xAI and Grok is understand the universe.

这就是 xAI 和 Grok 的使命：了解宇宙。

So do you think people…

那么你认为人们...

15:13 When you have a Neuralink with 10,000, 100,000 channels, most of the use cases will be communication with AI systems?

当你拥有一个具有 10,000、100,000 个通道的 Neuralink 时，大多数用例将是与 AI 系统通信？

Well, assuming that there are not…

好吧，假设没有……

15:27 They’re solving basic neurological issues that people have.

他们正在解决人们所面临的基本神经问题。

If they’ve got damaged neurons in their spinal cord, or neck, as is the case with our first two patients, then obviously the first order of business is solving fundamental neuron damage in a spinal cord, neck, or in the brain itself.

如果他们的脊髓或颈部神经元受损，就像我们前两名患者的情况一样，那么显然首要任务是解决脊髓、颈部或大脑本身的基本神经元损伤。

So, our second product is called Blindsight, which is to enable people who are completely blind, lost both eyes, or optic nerve, or just can’t see at all, to be able to see by directly triggering the neurons in the visual cortex.

所以，我们的第二个产品叫Blindsight，它是让那些完全失明、失去双眼、失去视神经、或者根本看不见的人，能够通过直接触发视觉皮层的神经元来看到东西。

16:18 So we’re just starting at the basics here, so it’s the simple stuff, relatively speaking, is solving neuron damage.

所以我们只是从基础开始，所以相对而言，解决神经元损伤是简单的事情。

It can also solve I think probably schizophrenia, if people have seizures of some kind, it could probably solve that.

我认为它还可以解决精神分裂症，如果人们有某种癫痫发作，它可能可以解决这个问题。

It could help with memory.

它可以帮助记忆。

So, there’s kind of a tech tree, if you will.

所以，如果你愿意的话，有一种科技树。

You’ve got the basics.

你已经掌握了基础知识。

You need literacy before you can have Lord of the Rings.

在拥有《指环王》之前，你需要识字。

Got it.

知道了。

So, do you have letters and the alphabet?

那么，你有字母和字母表吗？

17:02 Okay, great.

好的，太好了。

Words?

字？

And then eventually you get sagas.

然后最终你会得到传奇。

So, I think there may be some things to worry about in the future, but the first several years are really just solving basic neurological damage, like for people who have essentially complete or near complete loss from the brain to the body, like Stephen Hawking would be an example, the Neuralink would be incredibly profound, because you can imagine if Stephen Hawking could communicate as fast as we’re communicating, perhaps faster.

所以，我认为未来可能会有一些事情需要担心，但最初的几年实际上只是解决基本的神经损伤，就像对于那些从大脑到身体基本上完全或接近完全丧失的人来说，像史蒂芬·霍金就是一个例子，Neuralink 将是非常深刻的，因为你可以想象如果史蒂芬·霍金能够像我们一样快速地沟通，甚至更快。

And that’s certainly possible.

这当然是可能的。

Probable, in fact.

事实上，有可能。

Likely, I’d say.

我想说，有可能。

So there’s a kind of dual track of medical and non-medical, meaning so everything you’ve talked about could be applied to people who are non-disabled in the future?

所以说是一种医疗和非医疗的双轨，也就是说你所说的一切都可以应用到未来的非残疾人身上？

The logical thing to do is…

合乎逻辑的事情是……

17:58 Sensible thing to do is to start off solving basic neuron damage issues.

明智的做法是从解决基本的神经元损伤问题开始。

Yes.

是的。

Because there’s obviously some risk with a new device.

因为新设备显然存在一些风险。

18:11 You can’t get the risk down to zero, it’s not possible.

你无法将风险降至零，这是不可能的。

So, you want to have the highest possible reward, given there’s a certain irreducible risk.

因此，考虑到存在一定的不可降低的风险，您希望获得尽可能高的回报。

And if somebody’s able to have a profound improvement in their communication, that’s worth the risk.

如果某人能够在沟通方面取得深刻的进步，那么冒险是值得的。

As you get the risk down.

当你降低风险时。

Yeah.

是的。

18:36 As you get the risk down.

当你降低风险时。

And once the risk is down to…

一旦风险降至……

If you have thousands of people that have been using it for years and the risk is minimal, then perhaps at that point you could consider saying, “Okay, let’s aim for augmentation.” Now, I think we’re actually going to aim for augmentation with people who have neuron damage.

如果你有成千上万的人多年来一直在使用它，并且风险很小，那么也许那时你可以考虑说：“好吧，让我们以增强为目标。”现在，我认为我们实际上的目标是对神经元受损的人进行增强。

So we’re not just aiming to give people the communication data rate equivalent to normal humans.

因此，我们的目标不仅仅是为人们提供与正常人相当的通信数据速率。

We’re aiming to give people who have…

我们的目标是为那些拥有……的人提供

A quadriplegic, or maybe have complete loss of the connection to the brain and body, a communication data rate that exceeds normal humans.

四肢瘫痪，或者可能完全失去了大脑和身体的连接，通信数据速率超过了正常人。

While we’re in there, why not?

当我们在那里时，为什么不呢？

Let’s give people superpowers.

让我们赋予人们超能力。

And the same for vision.

视力也是如此。

19:20 As you restore vision, there could be aspects of that restoration that are superhuman.

当您恢复视力时，恢复的某些方面可能是超人的。

Yeah.

是的。

19:27 At first, the vision restoration will be low res, because you have to say, “How many neurons can you put in there, and trigger?” And you can do things where you adjust the electric field.

首先，视力恢复将是低分辨率的，因为你必须说，“你可以在那里放置多少个神经元并触发？”你可以做一些调整电场的事情。

So, even if you’ve got, say 10,000 neurons, it’s not just 10,000 pixels, because you can adjust the field between the neurons, and do them in patterns in order to have say, 10,000 electrodes, effectively give you, I don’t know, maybe like having a megapixel, or a 10 megapixel situation.

所以，即使你有，比如说 10,000 个神经元，它也不仅仅是 10,000 个像素，因为你可以调整神经元之间的场，并以某种模式进行它们，以便有说，10,000 个电极，有效地给你，我不知道，也许就像拥有一个百万像素或一个 10 兆像素的情况。

And then over time, I think you get to higher resolution than human eyes.

随着时间的推移，我认为你会获得比人眼更高的分辨率。

And you could also see in different wavelengths.

你还可以看到不同波长的东西。

So, like Geordi La Forge from Star Trek, he had the thing.

所以，就像《星际迷航》中的乔迪·拉·福吉（Geordi La Forge）一样，他拥有这样的东西。

Do you want to see it in radar?

你想在雷达上看到它吗？

No problem.

没问题。

You could see ultraviolet, infrared, eagle vision, whatever you want.

你可以看到紫外线、红外线、鹰眼视觉，任何你想要的东西。

Do you think there’ll be…

你觉得会有吗……

20:28 let me ask a Joe Rogan question.

让我问乔·罗根一个问题。

Do you think there’ll be…

你觉得会有吗……

I just recently have taken ayahuasca.

我最近刚刚服用死藤水。

Is that a serious question?

这是一个严肃的问题吗？

No.

不。

20:38 Well, yes.

嗯，是的。

Well, I guess technically it is.

嗯，我想从技术上来说是这样的。

Yeah.

是的。

Yeah.

是的。

Ever try DMT bro?

兄弟试过DMT吗？

Yeah, is this DMT in there, or something?

是啊，里面有DMT吗？

Love you, Joe.

爱你，乔。

20:42 Okay.

好的。

Wait, wait.

等等，等等。

20:48 Have you said much about it, the ayahuasca stuff?

你有没有说过很多有关死藤水的事？

I have not.

我没有。

I have not.

我没有。

I have not.

我没有。

Okay.

好的。

20:53 Well, why don’t you spill the beans?

好吧，你为什么不透露秘密呢？

It is a truly incredible experience.

这真是一次令人难以置信的经历。

Let me turn the tables on you.

让我来扭转局面。

Well, yeah.

嗯，是的。

You’re in the jungle.

你在丛林里。

Yeah, amongst the trees, myself and a shaman.

是的，在树林里，我和一个萨满。

Yeah.

是的。

21:02 It must’ve been crazy.

这一定是疯了。

Yeah, yeah, yeah.

是啊，是啊，是啊。

21:05 With the insects, with the animals all around you, the jungle as far as the eye can see, there’s no…

昆虫、动物在你身边，一望无际的丛林，没有……

That’s the way to do it.

这就是方法。

Things are going to look pretty wild.

事情看起来会非常疯狂。

Yeah, pretty wild.

是啊，相当狂野。

21:14 I took an extremely high dose.

我服用了极高的剂量。

Just don’t go hugging an Anaconda or something.

只是不要去拥抱蟒蛇之类的东西。

You haven’t lived unless you made love to an Anaconda.

除非你和蟒蛇做爱，否则你就不算活着。

21:24 I’m sorry, but…

抱歉，但是……

Snakes and Ladders.

蛇和梯子。

Yeah.

是的。

21:33 I took a extremely high dose.

我服用了极高的剂量。

Okay.

好的。

Nine cups.

九杯。

Damn.

该死。

21:39 Okay.

好的。

That sounds like a lot.

听起来很多。

Is normal to just one cup?

只喝一杯正常吗？

Or…

或者……

One or two.

一两个。

21:42 Usually one.

通常是一个。

Okay.

好的。

21:46 Wait.

等待。

Like right off the bat, or did you work your way up to it?

是一开始就这样，还是你自己努力做到的？

Did you just jump in at the deep end?

你刚刚跳进了深渊吗？

Across two days, because the first day, I took two, and it was a ride, but it wasn’t quite like a…

跨越两天，因为第一天，我坐了两次，这是一次骑行，但不太像……

It wasn’t like a revelation.

这不像是一个启示。

It wasn’t into deep space type of ride.

它不是深空类型的骑行。

22:01 It was just like a little airplane ride.

这就像坐一次小飞机一样。

And I [inaudible 00:22:07] saw some trees, and some visuals, and just saw a dragon and all that kind of stuff.

我 [听不清 00:22:07] 看到了一些树和一些视觉效果，还看到了一条龙和所有类似的东西。

But…

但是……

Nine cups, you went to Pluto, I think.

九杯，我想你去了冥王星。

Pluto.

冥王星。

22:15 Yeah.

是的。

No, Deep space.

不，深空。

Deep space.

深空。

One of the interesting aspects of my experience is I thought I would have some demons, some stuff to work through.

我的经历中一个有趣的方面是我认为我会有一些恶魔，一些需要解决的问题。

That’s what people [inaudible 00:22:26].

这就是人们的想法 [听不清 00:22:26]。

That’s what everyone says.

大家都是这么说的。

That’s what everyone says.

大家都是这么说的。

22:27 Yeah, exactly.

是的，完全正确。

I had nothing.

我什么都没有。

22:29 I had all positive.

我的一切都是积极的。

I just…

我只是……

So full-

这么饱满——

Just a pure soul.

只是一个纯粹的灵魂。

I don’t think so.

我不这么认为。

22:32 I don’t know.

我不知道。

But I kept thinking about, I had extremely high resolution thoughts about the people I know in my life.

但我一直在想，我对我生命中认识的人有非常清晰的想法。

You were there, and it is just not from my relationship with that person, but just as the person themselves.

你在那里，这不是因为我和那个人的关系，而是因为那个人本身。

I had just this deep gratitude of who they are.

我对他们深表感激。

That’s cool.

这很酷。

It was just like this exploration, like Sims, or whatever.

就像这次探索，就像模拟人生，或者其他什么。

22:53 You get to watch them.

你可以观察他们。

I got to watch people, and just be in awe of how amazing they are.

我必须观察人们，并对他们的惊人之处感到敬畏。

That sounds awesome.

听起来棒极了。

Yeah, it was great.

是的，太棒了。

23:02 I was waiting for-

我在等待——

When’s the demon coming?

恶魔什么时候来？

Exactly.

确切地。

23:07 Maybe I’ll have some negative thoughts.

也许我会有一些消极的想法。

Nothing.

没有什么。

Nothing.

没有什么。

Just extreme gratitude for them.

只是对他们表示万分的感谢。

And also a lot of space travel.

还有很多太空旅行。

Space travel to where?

太空旅行去哪里？

So here’s what it was.

事情就是这样。

23:20 It was people, the human beings that I know, they had this kind of…

是人，我所认识的人类，他们有这样的……

The best way I could describe it is they had a glow to them.

我能描述的最好的方式就是他们散发着光芒。

Okay.

好的。

And then I kept flying out from them to see earth, to see our solar system, to see our galaxy.

然后我不断地从他们那里飞出去，看看地球，看看我们的太阳系，看看我们的银河系。

23:30 And I saw that light, that glow all across the universe, whatever that form is, whatever that…

我看到了那道光，那光芒遍布整个宇宙，无论那是什么形式，无论那是什么……

Did you go past the Milky Way?

你穿过银河系了吗？

Yeah.

是的。

Okay.

好的。

23:53 You’re like intergalactic.

你就像星际一样。

Yeah, intergalactic.

是的，星际。

Okay.

好的。

23:55 Dang.

当当。

But always pointing in, yeah.

但总是指向，是的。

23:56 Past the Milky Way, past…

过去了银河系，过去了……

I mean, I saw a huge number of galaxies, intergalactic, and all of it was glowing, but I couldn’t control that travel, because I would actually explore near distances to the solar system, see if there’s aliens, or any of that kind of stuff.

我的意思是，我看到了大量的星系，星系间的星系，所有这些都在发光，但我无法控制这种旅行，因为我实际上会探索太阳系附近的距离，看看是否有外星人，或者任何类似的东西。

Sure.

当然。

Did you see an alien?

你看到外星人了吗？

No.

不。

24:14 I didn’t, no.

我没有，没有。

Zero aliens?

零外星人？

Implication of aliens, because they were glowing.

暗示外星人，因为他们会发光。

24:16 They were glowing in the same way that humans were glowing.

它们的发光方式与人类发光的方式相同。

That life force that I was seeing, the thing that made humans amazing was there throughout the universe.

我所看到的生命力，让人类惊叹的东西存在于整个宇宙中。

There was these glowing dots.

有这些发光的点。

So, I don’t know.

所以，我不知道。

It made me feel like there is life…

这让我觉得还有生命……

No, not life, but something, whatever makes humans amazing all throughout the universe.

不，不是生命，而是某种东西，任何让人类在整个宇宙中令人惊叹的东西。

Sounds good.

听起来不错。

Yeah, it was amazing.

是的，太棒了。

24:42 No demons.

没有恶魔。

No demons.

没有恶魔。

I looked for the demons.

我寻找恶魔。

There’s no demons.

没有恶魔。

There were dragons, and they’re pretty awesome.

有龙，而且它们非常可怕。

So the thing about trees-

那么关于树的事情-

Was there anything scary at all?

有什么可怕的吗？

Dragons.

龙。

24:54 But they weren’t scary.

但他们并不可怕。

They were friends.

他们是朋友。

They were protective.

他们是保护性的。

So, the thing is-

所以，事情是——

Sure.

当然。

24:57 Like Puff the Magic Dragon.

比如魔龙帕夫。

No, it was more like a Game of Thrones kind of dragons.

不，它更像是权力的游戏中的龙。

24:58 They weren’t very friendly.

他们不太友好。

They were very big.

它们很大。

So the thing is that bought giant trees, at night, which is where I was-

所以事情是在晚上买了参天大树，这就是我所在的地方——

Yeah.

是的。

25:09 I mean, the jungle’s kind of scary.

我的意思是，丛林有点可怕。

Yeah.

是的。

25:10 The trees started to look like dragons, and they were all looking at me.

树木开始看起来像龙，它们都在看着我。

Sure.

当然。

25:15 Okay.

好的。

And it didn’t seem scary.

而且看起来并不可怕。

25:17 They seemed like they were protecting me.

他们看起来就像是在保护我。

And the shaman and the people didn’t speak any English, by the way, which made it even scarier, because we’re not even…

顺便说一下，萨满和人们不会说任何英语，这让它更可怕，因为我们甚至不……

We’re worlds apart in many ways, but yeah, they talk about the mother of the forest protecting you, and that’s what I felt like.

我们在很多方面都有天壤之别，但是，是的，他们谈论森林之母保护你，这就是我的感受。

And you were way out in the jungle.

而你却身处丛林之中。

Way out.

出路。

25:40 This is not like a tourist retreat.

这不像是一个旅游胜地。

Like 10 miles outside of Rio or something.

比如距离里约热内卢 10 英里之类的地方。

No, we went…

不，我们去了……

25:47 No, this is not a-

不，这不是——

You’re in deep Amazon.

你在亚马逊深处。

Me and this guy named Paul Rosolie, who basically is a Tarzan, he lives in the jungle, we went out deep and we just went crazy.

我和这个叫保罗·罗索利（Paul Rosolie）的人，他基本上是泰山，他住在丛林里，我们深入深处，我们简直疯了。

Wow.

哇。

25:59 Cool.

凉爽的。

Yeah.

是的。

26:01 So anyway.

所以无论如何。

Can I get that same experience in a Neuralink?

我可以在 Neuralink 中获得同样的体验吗？

Probably.

大概。

26:04 Yeah.

是的。

I guess that is the question for non-disabled people.

我想这就是非残疾人的问题。

26:05 Do you think that there’s a lot in our perception, in our experience of the world that could be explored, that could be played with, using Neuralink?

您是否认为我们的感知、我们对世界的体验中有很多东西可以使用 Neuralink 来探索、玩转？

Yeah, I mean, Neuralink, it’s really a generalized input-output device.

是的，我的意思是，Neuralink，它确实是一个通用的输入输出设备。

26:18 It’s reading electrical signals, and generating electrical signals, and I mean, everything that you’ve ever experienced in your whole life, smell, emotions, all of those are electrical signals.

它正在读取电信号，并生成电信号，我的意思是，你一生中经历过的一切，气味，情绪，所有这些都是电信号。

So, it’s kind of weird to think that your entire life experience is distilled down to electrical signals for neurons, but that is in fact the case.

因此，认为你的整个生活经历都被归结为神经元的电信号有点奇怪，但事实确实如此。

Or I mean, that’s at least what all the evidence points to.

或者我的意思是，至少所有证据都表明了这一点。

So, I mean, if you trigger the right neuron, you could trigger a particular scent.

所以，我的意思是，如果你触发正确的神经元，你就可以触发一种特定的气味。

You could certainly make things glow.

你当然可以让事物发光。

I mean, do pretty much anything.

我的意思是，做几乎任何事情。

I mean, really, you can think of the brain as a biological computer.

我的意思是，实际上，你可以将大脑视为一台生物计算机。

So, if there are certain say, chips or elements of that biological computer that are broken, let’s say your ability to…

所以，如果生物计算机的某些芯片或元件损坏了，比如说你有能力……

If you’ve got a stroke, that if you’ve had a stroke, that means some part of your brain is damaged.

如果你中风了，那就意味着你大脑的某些部分受损了。

Let’s say it’s speech generation, or the ability to move your left hand.

假设它是语音生成，或者移动左手的能力。

That’s the kind of thing that a Neuralink could solve.

这就是 Neuralink 可以解决的问题。

27:33 If you’ve got a massive amount of memory loss that’s just gone, well, we can’t get the memories back.

如果你的记忆大量丧失，而且刚刚消失，那么我们就无法恢复记忆。

We could restore your ability to make memories, but we can’t restore memories that are fully gone.

我们可以恢复你创造记忆的能力，但我们无法恢复完全消失的记忆。

Now, I should say, maybe if part of the memory is there, and the means of accessing the memory is the part that’s broken, then we could re-enable the ability to access the memory.

现在，我应该说，也许如果部分内存还存在，并且访问内存的方式是损坏的部分，那么我们可以重新启用访问内存的能力。

But you can think of it like ram in a computer, if the ram is destroyed, or your SD card is destroyed, we can’t get that back.

但你可以把它想象成电脑中的内存，如果内存被破坏，或者你的SD卡被破坏，我们就无法恢复它。

But if the connection to the SD card is destroyed, we can fix that.

但如果与 SD 卡的连接被破坏，我们可以修复它。

If it is fixable physically, then it can be fixed.

如果物理上可以修复，那么就可以修复。

Of course, with AI, just like you can repair photographs, and fill in missing parts of photographs, maybe you can do the same, just like [inaudible 00:28:31] parts.

当然，有了人工智能，就像你可以修复照片，填补照片中缺失的部分一样，也许你也可以做同样的事情，就像[听不清 00:28:31] 的部分。

Yeah, you could say like, create the most probable set of memories based on all the information you have about that person.

是的，你可以说，根据你拥有的关于那个人的所有信息创建最可能的一组记忆。

28:30 You could then…

那么你可以……

It would be probabilistic restoration of memory.

这将是记忆的概率恢复。

Now, we’re getting pretty esoteric here.

现在，我们在这里变得非常深奥。

But that is one of the most beautiful aspects of the human experience is remembering the good memories.

但人类经历中最美丽的方面之一就是记住美好的回忆。

Sure.

当然。

We live most of our life, as Danny Kahneman has talked about, in our memories, not in the actual moment.

正如丹尼·卡尼曼所说，我们一生的大部分时间都生活在我们的记忆中，而不是现实的时刻。

28:53 We’re collecting memories and we kind of relive them in our head.

我们正在收集记忆，并在脑海中重温它们。

And that’s the good times.

那就是美好时光。

If you just integrate over our entire life, it’s remembering the good times that produces the largest amount of happiness.

如果你把我们的一生融为一体，记住美好时光就会产生最大的幸福。

Yeah.

是的。

29:11 Well, I mean, what are we but our memories?

好吧，我的意思是，除了我们的记忆之外，我们还算什么呢？

And what is death?

什么是死亡？

But the loss of memory, loss of information?

但失去记忆、失去信息呢？

If you could say, well, if you could run a thought experiment, what if you were disintegrated painlessly, and then reintegrated a moment later, like teleportation, I guess?

如果你可以说，好吧，如果你可以进行一个思想实验，如果你被无痛地解体，然后片刻后重新整合，我猜，就像瞬间传送一样，会怎么样？

Provided there’s no information loss, the fact that your one body was disintegrated is irrelevant.

只要没有信息丢失，你的身体被解体这一事实就无关紧要了。

And memories is just such a huge part of that.

记忆只是其中很大一部分。

Death is fundamentally the loss of information, the loss of memory.

死亡从根本上来说就是信息的丢失、记忆的丢失。

So, if we can store them as accurately as possible, we basically achieve a kind of immortality.

所以，如果我们能够尽可能准确地存储它们，我们基本上就实现了一种永生。

Yeah.

是的。

You’ve talked about the threats, the safety concerns of AI.

您谈到了人工智能的威胁和安全问题。

29:57 Let’s look at long-term visions.

让我们看看长期愿景。

Do you think Neuralink is, in your view, the best current approach we have for AI safety?

在您看来，Neuralink 是我们当前实现人工智能安全的最佳方法吗？

It’s an idea that may help with AI safety.

这个想法可能有助于人工智能安全。

30:13 Certainly, I wouldn’t want to claim it’s some panacea, or that it’s a sure thing, but I mean, many years ago I was thinking like, “Well, what would inhibit alignment of collective human will with artificial intelligence?” And the low data rate of humans, especially our slow output rate would necessarily, just because the communication is so slow, would diminish the link between humans and computers.

当然，我不想声称它是某种灵丹妙药，或者这是肯定的事情，但我的意思是，很多年前我就在想，“那么，什么会阻碍人类集体意志与人工智能的结合呢？”人类的低数据速率，尤其是我们缓慢的输出速率，仅仅因为通信太慢，必然会削弱人类和计算机之间的联系。

The more you are a tree, the less you know what the tree is.

你越是一棵树，你就越不知道树是什么。

Let’s say you look at this plant or whatever, and hey, I’d really like to make that plant happy, but it’s not saying a lot.

假设你看着这种植物或其他什么，嘿，我真的很想让那种植物快乐，但这并没有说明太多。

So the more we increase the data rate that humans can intake and output, then that means the better, the higher the chance we have in a world full of AGI’s.

因此，我们提高人类接收和输出的数据速率越多，就意味着我们在一个充满 AGI 的世界中拥有的机会就越好，机会就越大。

Yeah.

是的。

31:21 We could better align collective human will with AI if the output rate especially was dramatically increased.

如果产出率大幅提高，我们就可以更好地将人类集体意志与人工智能结合起来。

And I think there’s potential to increase the output rate by, I don’t know, three, maybe six, maybe more orders of magnitude.

我认为有可能将产出率提高三个，也许六个，也许更多数量级，我不知道。

So, it’s better than the current situation.

所以，比现在的情况要好一些。

And that output rate would be by increasing the number of electrodes, number of channels, and also maybe implanting multiple Neuralinks?

输出率是通过增加电极数量、通道数量，也可能是植入多个 Neuralink 来实现？

Yeah.

是的。

Do you think there’ll be a world in the next couple of decades where it’s hundreds of millions of people have Neuralinks?

您认为未来几十年内会有一个数亿人拥有 Neuralink 的世界吗？

Yeah, I do.

是的，我愿意。

You think when people just when they see the capabilities, the superhuman capabilities that are possible, and then the safety is demonstrated.

你认为，当人们看到能力、可能的超人能力时，然后就会证明安全性。

Yeah.

是的。

32:11 If it’s extremely safe, and you can have superhuman abilities, and let’s say you can upload your memories, so you wouldn’t lose memories, then I think probably a lot of people would choose to have it.

如果它非常安全，而且你可以拥有超人的能力，比如说你可以上传你的记忆，这样你就不会丢失记忆，那么我想可能很多人都会选择拥有它。

It would supersede the cell phone, for example.

例如，它将取代手机。

I mean, the biggest problem that say, a phone has, is trying to figure out what you want.

我的意思是，手机最大的问题是试图弄清楚你想要什么。

That’s why you’ve got auto complete, and you’ve got output, which is all the pixels on the screen, but from the perspective of the human, the output is so frigging slow.

这就是为什么你有自动完成功能，并且有输出，即屏幕上的所有像素，但从人类的角度来看，输出太慢了。

Desktop or phone is desperately just trying to understand what you want.

台式电脑或手机正在拼命地试图了解您想要什么。

And there’s an eternity between every keystroke from a computer standpoint.

从计算机的角度来看，每次击键之间都有一个永恒的时间。

Yeah.

是的。

33:06 Yeah.

是的。

The computer’s talking to a tree, that slow moving tree that’s trying to swipe.

计算机正在与一棵树对话，这棵树移动缓慢，试图滑动。

Yeah.

是的。

33:12 So, if you had computers that are doing trillions of instructions per second, and a whole second went by, I mean, that’s a trillion things it could have done.

所以，如果你的计算机每秒执行数万亿条指令，并且一整秒过去了，我的意思是，它可以完成万亿次的事情。

Yeah.

是的。

33:24 I think it’s exciting, and scary for people, because once you have a very high bit rate, it changes the human experience in a way that’s very hard to imagine.

我认为这对人们来说既令人兴奋又令人恐惧，因为一旦拥有非常高的比特率，它就会以一种难以想象的方式改变人类的体验。

Yeah.

是的。

33:35 We would be something different.

我们会有所不同。

I mean, some sort of futuristic cyborg, I mean, we’re obviously talking about, by the way, it’s not like around the corner.

我的意思是，某种未来派的机器人，我的意思是，我们显然正在谈论，顺便说一句，它并不是指日可待。

You asked me what the distant future is.

你问我遥远的未来是什么。

Maybe this is…

也许这就是……

It’s not super far away, but 10, 15 years, that kind of thing.

并不是很遥远，但是10年、15年之类的事情。

When can I get one?

我什么时候可以得到一个？

33:58 10 years?

10年？

Probably less than 10 years.

大概不到10年。

34:02 It depends on what you want to do.

这取决于你想做什么。

Hey, if I can get a thousand BPS?

嘿，如果我能得到一千个 BPS 吗？

A thousand BPS, wow.

一千 BPS，哇。

And it’s safe, and I can just interact with a computer while laying back and eating Cheetos.

而且它很安全，我可以一边躺着吃奇多一边与电脑互动。

34:12 I don’t eat Cheetos.

我不吃奇多。

There’s certain aspects of human computer interaction when done more efficiently, and more enjoyably, are worth it.

当人机交互的某些方面做得更高效、更愉快时，是值得的。

Well, we feel pretty confident that I think maybe within the next year or two, that someone with a Neuralink implant will be able to outperform a pro gamer.

好吧，我们非常有信心，我认为也许在未来一两年内，植入 Neuralink 的人将能够超越职业玩家。

Nice.

好的。

Because the reaction time would be faster.

因为反应时间会更快。

I got to visit Memphis.

我得去孟菲斯看看。

Yeah.

是的。

34:46 Yeah.

是的。

You’re going big on compute.

您正在大力发展计算。

Yeah.

是的。

And you’ve also said, “Play to win, or don’t play at all.”

你也说过，“为了胜利而打球，或者干脆不打球。”

Yeah.

是的。

So what does it take to win?

那么，怎样才能获胜呢？

For AI, that means you’ve got to have the most powerful training compute, and the rate of improvement of training compute has to be-

对于人工智能来说，这意味着你必须拥有最强大的训练计算，并且训练计算的改进率必须是-

35:00 And the rate of improvement of training compute has to be faster than everyone else, or you will not win.

而且训练计算的提升速度必须比其他人快，否则你就赢不了。

Your AI will be worse.

你的人工智能会更糟。

So how can Grok, let’s say 3…

那么 Grok 怎样才能做到，比方说 3……

35:10 That might be available, what, next year?

那可能是可用的，什么，明年？

Well, hopefully end of this year.

嗯，希望今年年底。

Grok 3.

格罗克 3.

If we’re lucky.

如果我们幸运的话。

35:17 Yeah.

是的。

How can that be the best LLM, the best AI system available in the world?

这怎么可能是世界上最好的法学硕士、最好的人工智能系统呢？

35:20 How much of it is compute?

其中有多少是计算的？

How much of it is data?

其中有多少是数据？

How much of it is post-training?

其中有多少是后期培训的？

How much of it is the product that you package it up in, all that kind of stuff?

诸如此类的东西，其中有多少是你包装的产品？

I mean, they all matter.

我的意思是，它们都很重要。

35:35 It’s sort of like saying, let’s say it’s a Formula 1 race, what matters more, the car or the driver?

这有点像在说，假设这是一场一级方程式比赛，赛车和车手哪个更重要？

I mean, they both matter.

我的意思是，它们都很重要。

If a car is not fast, then if, let’s say, it’s half the horsepower of your competitors, the best driver will still lose.

如果一辆汽车速度不快，那么如果它的马力只有竞争对手的一半，那么最好的车手仍然会失败。

If it’s twice the horsepower, then probably even a mediocre driver will still win.

如果马力是两倍，那么即使是平庸的车手也可能会获胜。

So, the training compute is kind of like the engine, this horsepower of the engine.

所以，训练计算有点像引擎，引擎的马力。

So, really, you want to try to do the best on that.

所以，真的，你想尽力做到最好。

And then, it’s how efficiently do you use that training compute, and how efficiently do you do the inference, the use of the AI?

然后，问题是你使用训练计算的效率如何，以及你进行推理、人工智能的使用的效率如何？

So, obviously, that comes down to human talent.

所以，显然，这取决于人才。

And then, what unique access to data do you have?

然后，您对数据有什么独特的访问权限？

That also plays a role.

这也发挥了作用。

Do you think Twitter data will be useful?

您认为 Twitter 数据有用吗？

Yeah.

是的。

36:31 I mean, I think most of the leading AI companies have already scraped all the Twitter data.

我的意思是，我认为大多数领先的人工智能公司已经抓取了所有 Twitter 数据。

Not I think.

我认为不是。

They have.

他们有。

So, on a go forward basis, what’s useful is the fact that it’s up to the second, because that’s hard for them to scrape in real time.

因此，从长远来看，有用的是它取决于第二个事实，因为这对他们来说很难实时抓取。

So, there’s an immediacy advantage that Grok has already.

因此，Grok 已经具备了即时优势。

I think with Tesla and the real time video coming from several million cars, ultimately tens of millions of cars with Optimus, there might be hundreds of millions of Optimus robots, maybe billions, learning a tremendous amount from the real world.

我认为，有了特斯拉以及来自数百万辆汽车（最终是数千万辆配备 Optimus 的汽车）的实时视频，可能会有数亿甚至数十亿个 Optimus 机器人从现实世界中学习大量知识。

That’s the biggest source of data, I think, ultimately, is Optimus, probably.

我认为，最终，最大的数据来源可能是 Optimus。

Optimus is going to be the biggest source of data.

Optimus 将成为最大的数据源。

Because it’s able to-

因为它能够——

Because reality scales.

因为现实是有尺度的。

37:22 Reality scales to the scale of reality.

现实规模与现实规模成正比。

It’s actually humbling to see how little data humans have actually been able to accumulate.

看到人类实际上能够积累的数据如此之少，实际上令人感到羞愧。

Really, if you say how many trillions of usable tokens have humans generated, where on a non-duplicative…

真的，如果你说人类生成了多少万亿个可用代币，在非重复的情况下......

Discounting spam and repetitive stuff, it’s not a huge number.

扣除垃圾邮件和重复的内容，这个数字并不是很大。

You run out pretty quickly.

你很快就用完了。

And Optimus can go…

擎天柱可以去……

37:54 So, Tesla cars, unfortunately, have to stay on the road.

因此，不幸的是，特斯拉汽车必须继续上路。

Right.

正确的。

Optimus robot can go anywhere.

擎天柱机器人可以去任何地方。

38:01 And there’s more reality off the road.

还有更多的现实正在发生。

And go off-road.

并越野。

Yeah.

是的。

38:06 I mean, the Optimus robot can pick up the cup and see, did it pick up the cup in the right way?

我的意思是，擎天柱机器人可以拿起杯子，看看它是否以正确的方式拿起杯子？

Did it, say, go pour water in the cup?

它是说，往杯子里倒水吗？

Did the water go in the cup or not go in the cups?

水进了杯子还是没进杯子？

Did it spill water or not?

水洒了还是没洒？

Simple stuff like that.

诸如此类的简单事情。

But it can do that at scale times a billion, so generate useful data from reality, so cause and effect stuff.

但它可以在十亿倍的规模上做到这一点，因此可以从现实中生成有用的数据，以及因果关系。

What do you think it takes to get to mass production of humanoid robots like that?

您认为如何才能大规模生产这样的人形机器人？

It’s the same as cars, really.

这真的和汽车一样。

38:40 I mean, global capacity for vehicles is about 100 million a year, and it could be higher.

我的意思是，全球汽车产能每年约为 1 亿辆，而且可能会更高。

It’s just that the demand is on the order of 100 million a year.

只是每年的需求量在亿个量级。

And then, there’s roughly two billion vehicles that are in use in some way, which makes sense because the life of a vehicle is about 20 years.

然后，大约有 20 亿辆汽车以某种方式在使用，这是有道理的，因为汽车的使用寿命约为 20 年。

So, at steady state, you can have 100 million vehicles produced a year with a two billion vehicle fleet, roughly.

因此，在稳定状态下，每年可以生产 1 亿辆汽车，拥有大约 20 亿辆汽车。

Now, for humanoid robots, the utility is much greater.

现在，对于人形机器人来说，效用要大得多。

So, my guess is humanoid robots are more like at a billion plus per year.

所以，我的猜测是，人形机器人每年的数量可能超过十亿。

But until you came along and started building Optimus, it was thought to be an extremely difficult problem.

但直到你出现并开始构建擎天柱之前，这被认为是一个极其困难的问题。

Well, I think it is.

嗯，我想是的。

I mean, it still is an extremely difficult problem.

我的意思是，这仍然是一个极其困难的问题。

Yes.

是的。

39:28 So, a walk in the park.

于是，去公园散步。

I mean, Optimus, currently, would struggle to walk in the park.

我的意思是，擎天柱目前在公园里走路很困难。

I mean, it can walk in a park.

我的意思是，它可以在公园里散步。

The park is not too difficult, but it will be able to walk over a wide range of terrain.

公园难度不算太大，但是能走的地形范围很广。

Yeah.

是的。

39:43 And pick up objects.

并捡起物体。

Yeah, yeah.

是啊是啊。

39:45 It can already do that.

它已经可以做到这一点。

But all kinds of objects.

但各种物体。

Yeah, yeah.

是啊是啊。

All foreign objects.

所有异物。

39:50 I mean, pouring water in a cup is not trivial, because then if you don’t know anything about the container, it could be all kinds of containers.

我的意思是，将水倒入杯子中并不是一件小事，因为如果你对容器一无所知，那么它可能是各种容器。

Yeah, there’s going to be an immense amount of engineering just going into the hand.

是的，将会有大量的工程工作投入到手中。

39:59 The hand, it might be close to half of all the engineering in Optimus.

这只手，可能接近擎天柱所有工程的一半。

From an electromechanical standpoint, the hand is probably roughly half of the engineering.

从机电的角度来看，手可能大约占工程的一半。

But so much of the intelligence of humans goes into what we do with our hands.

但人类的大部分智慧都体现在我们用手所做的事情上。

Yeah.

是的。

It’s the manipulation of the world, manipulation of objects in the world.

这是对世界的操纵，对世界上物体的操纵。

40:22 Intelligent, safe manipulation of objects in the world.

智能、安全地操控世界上的物体。

Yeah.

是的。

Yeah.

是的。

40:28 I mean, you start really thinking about your hand and how it works.

我的意思是，你开始真正思考你的手以及它是如何运作的。

I do all the time.

我一直都这样做。

The sensory control homunculus is where you have humongous hands.

感觉控制小人就是你拥有巨大双手的地方。

40:35 So I mean, your hands, the actuators, the muscles of your hand are almost overwhelmingly in your forearm.

所以我的意思是，你的手、执行器、你手上的肌肉几乎绝大多数都在你的前臂上。

So, your forearm has the muscles that actually control your hand.

所以，你的前臂有真正控制你的手的肌肉。

There’s a few small muscles in the hand itself, but your hand is really like a skeleton meat puppet and with cables.

手本身有一些小肌肉，但你的手真的就像一个骷髅肉傀儡，有电缆。

So, the muscles that control your fingers are in your forearm, and they go through the carpal tunnel, which is that you’ve got a little collection of bones and a tiny tunnel that these cables, the tendons go through, and those tendons are mostly what move your hands.

所以，控制你手指的肌肉在你的前臂，它们穿过腕管，腕管里有一些骨头和一条小隧道，这些电缆、肌腱穿过这些隧道，而这些肌腱主要是移动你的手的。

And something like those tendons has to be re-engineered into the Optimus in order to do all that kind of stuff.

像这些肌腱这样的东西必须重新设计到擎天柱中才能完成所有这些事情。

Yeah.

是的。

41:26 So the current Optimus, we tried putting the actuators in the hand itself.

所以目前的擎天柱，我们尝试将执行器放在手上。

Then you sort of end up having these-

然后你最终会得到这些-

Giant hands?

巨手？

…

……

41:34 yeah, giant hands that look weird.

是的，巨大的手看起来很奇怪。

And then, they don’t actually have enough degrees of freedom or enough strength.

然后，他们实际上没有足够的自由度或足够的力量。

So then you realize, “Oh, okay, that’s why you got to put the actuators in the forearm.” And just like a human, you’ve got to run cables through a narrow tunnel to operate the fingers.

然后你意识到，“哦，好吧，这就是为什么你必须把执行器放在前臂上。”就像人类一样，你必须将电缆穿过狭窄的隧道来操作手指。

And then, there’s also a reason for not having all the fingers the same length.

此外，所有手指的长度不一样也是有原因的。

So, it wouldn’t be expensive from an energy or evolutionary standpoint to have all your fingers be the same length.

因此，从能量或进化的角度来看，让所有手指长度相同并不昂贵。

So, why not do the same length?

那么，为什么不做同样的长度呢？

Yeah, why not?

是啊，为什么不呢？

Because it’s actually better to have different lengths.

因为实际上有不同的长度更好。

42:04 Your dexterity is better if you’ve got fingers that are different lengths.

如果您的手指长度不同，您的灵活性会更好。

There are more things you can do and your dexterity is actually better if your fingers are a different length.

如果您的手指长度不同，您可以做的事情会更多，而且您的灵活性实际上会更好。

There’s a reason we’ve got a little finger.

我们有小手指是有原因的。

Why not have a little finger that’s bigger?

为什么没有一个更大的小指呢？

Yeah.

是的。

Because it helps you with fine motor skills.

因为它可以帮助您培养精细运动技能。

This little finger helps?

这个小指有用吗？

It does.

确实如此。

42:28 But if you lost your little finger, you’d have noticeably less dexterity.

但如果你失去了小指，你的灵活性就会明显下降。

So, as you’re figuring out this problem, you have to also figure out a way to do it so you can mass manufacture it, so as to be as simple as possible.

所以，当你解决这个问题时，你还必须找到一种方法来实现它，这样你就可以批量生产它，从而尽可能简单。

It’s actually going to be quite complicated.

实际上会非常复杂。

42:42 The as possible part is it’s quite a high bar.

尽可能的部分是它的门槛相当高。

If you want to have a humanoid robot that can do things that a human can do, actually, it’s a very high bar.

如果你想拥有一个能做人类能做的事情的人形机器人，实际上这是一个非常高的门槛。

So, our new arm has 22 degrees of freedom instead of 11 and has, like I said, the actuators in the forearm.

因此，我们的新手臂有 22 个自由度，而不是 11 个自由度，并且正如我所说，执行器位于前臂中。

And all the actuators are designed from scratch, from physics first principles.

所有执行器都是根据物理学第一原理从头开始设计的。

The sensors are all designed from scratch.

传感器都是从头开始设计的。

And we’ll continue to put a tremendous amount of engineering effort into improving the hand.

我们将继续投入大量的工程努力来改进手部。

By hand, I mean the entire forearm, from elbow forward, is really the hand.

我所说的手，是指从肘部向前的整个前臂，实际上就是手。

So, that’s incredibly difficult engineering, actually.

所以，这实际上是一项非常困难的工程。

And so, the simplest possible version of a humanoid robot that can do even most, perhaps not all, of what a human can do is actually still very complicated.

因此，最简单的人形机器人版本实际上仍然非常复杂，它甚至可以完成人类所能做的大部分（也许不是全部）工作。

It’s not simple.

这并不简单。

It’s very difficult.

这是非常困难的。

Can you just speak to what it takes for a great engineering team for you?

您能谈谈对您来说优秀的工程团队需要什么吗？

43:47 What I saw in Memphis, the supercomputer cluster, is just this intense drive towards simplifying the process, understanding the process, constantly improving it, constantly iterating it.

我在孟菲斯超级计算机集群中看到的正是这种简化流程、理解流程、不断改进、不断迭代的强烈动力。

Well, it’s easy to say ‘simplify,’ and it’s very difficult to do it.

“简化”说起来容易，做起来却很难。

44:08 I have this very basic first principles algorithm that I run kind of as a mantra, which is to first question the requirements, make the requirements less dumb.

我有一个非常基本的第一原理算法，我将其作为口头禅运行，即首先质疑需求，使需求不那么愚蠢。

The requirements are always dumb to some degree.

在某种程度上，要求总是愚蠢的。

So, you want to start off by reducing the number of requirements, and no matter how smart the person is who gave you those requirements, they’re still dumb to some degree.

因此，你想从减少要求的数量开始，无论提出这些要求的人多么聪明，他们在某种程度上仍然是愚蠢的。

You have to start there, because, otherwise, you could get the perfect answer to the wrong question.

你必须从那里开始，因为否则你可能会得到错误问题的完美答案。

So, try to make the question the least wrong possible.

所以，尽量让问题尽可能少出错。

That’s what question the requirements means.

这就是需求的意义所在。

44:53 And then, the second thing is try to delete whatever the step is, the part or the process step.

然后，第二件事是尝试删除任何步骤、零件或流程步骤。

It sounds very obvious, but people often forget to try deleting it entirely.

这听起来很明显，但人们经常忘记尝试完全删除它。

And if you’re not forced to put back at least 10% of what you delete, you’re not deleting enough.

如果您没有被迫恢复至少 10% 的删除内容，那么您删除的内容还不够。

Somewhat illogically, people often, most of the time, feel as though they’ve succeeded if they’ve not been forced to put things back in.

有点不合逻辑的是，大多数时候，人们常常觉得，如果他们没有被迫把事情放回去，他们就已经成功了。

But, actually, they haven’t because they’ve been overly conservative and have left things in there that shouldn’t be.

但实际上，他们没有这样做，因为他们过于保守，留下了不该留下的东西。

And only the third thing is try to optimize it or simplify it.

只有第三件事是尝试优化它或简化它。

Again, these all sound, I think, very obvious when I say them, but the number of times I’ve made these mistakes is more than I care to remember.

再说一次，我认为，当我说这些话时，这些听起来都非常明显，但我犯这些错误的次数多得我记不清了。

That’s why I have this mantra.

这就是为什么我有这个咒语。

So in fact, I’d say the most common mistake of smart engineers is to optimize a thing that should not exist.

所以事实上，我想说聪明的工程师最常犯的错误就是优化了一个不应该存在的东西。

Right.

正确的。

46:01 So, like you say, you run through the algorithm and basically show up to a problem, show up to the supercomputer cluster, and see the process, and ask, “Can this be deleted?”

所以，就像你说的，你运行算法，基本上会出现一个问题，出现在超级计算机集群上，看到进程，然后问，“这个可以删除吗？”

Yeah.

是的。

46:14 First try to delete it.

首先尝试删除它。

Yeah.

是的。

Yeah.

是的。

46:18 That’s not easy to do.

这并不容易做到。

No.

不。

46:20 Actually, what generally makes people uneasy is that at least some of the things that you delete, you will put back in.

事实上，通常让人不安的是，至少有一些你删除的东西，你又会放回去。

But going back to sort of where our limbic system can steer us wrong is that we tend to remember, with sometimes a jarring level of pain, where we deleted something that we subsequently needed.

但回到我们的边缘系统可能引导我们错误的地方，我们往往会记住，有时会感到非常痛苦，我们删除了我们随后需要的东西。

And so, people will remember that one time they forgot to put in this thing three years ago, and that caused them trouble.

所以，人们会记得三年前有一次他们忘记放这个东西，这给他们带来了麻烦。

And so, they overcorrect, and then they put too much stuff in there and overcomplicate things.

因此，他们矫枉过正，然后他们放了太多东西，让事情变得过于复杂。

So, you actually have to say, “Look, we’re deliberately going to delete more than we should.” At least one in 10 things, we’re going to add back in.

所以，你实际上必须说，“看，我们故意删除比我们应该删除的更多的内容。”至少十分之一的事情，我们会重新添加。

I’ve seen you suggest just that, that something should be deleted, and you can kind of see the pain.

我看到你建议应该删除一些东西，你可以看到其中的痛苦。

Oh, yeah.

哦，是的。

47:18 Absolutely.

绝对地。

Everybody feels a little bit of the pain.

每个人都感到一点点痛苦。

Absolutely.

绝对地。

47:21 And I tell them in advance, “Yeah, some of the things that we delete, we’re going to put back in.” People get a little shook by that, but it makes sense because if you’re so conservative as to never have to put anything back in, you obviously have a lot of stuff that isn’t needed.

我提前告诉他们，“是的，我们删除的一些东西，我们会放回去。”人们对此感到有点震惊，但这是有道理的，因为如果你如此保守以至于永远不需要放回任何东西，那么你显然有很多不需要的东西。

So, you got to overcorrect.

所以，你必须矫枉过正。

This is, I would say, like a cortical override to a limbic instinct.

我想说，这就像皮质超越边缘本能。

One of many that probably leads us astray.

其中之一可能会让我们误入歧途。

Yeah.

是的。

47:50 There’s a step four as well, which is any given thing can be sped up.

还有第四步，即任何给定的事情都可以加快。

However fast you think it can be done, whatever the speed it’s being done, it can be done faster.

无论你认为它能以多快的速度完成，无论它的完成速度如何，它都可以做得更快。

But you shouldn’t speed things up until you’ve tried to delete it and optimize.

但在尝试删除它并优化之前，您不应该加快速度。

Although, you’re speeding up something that…

虽然，你正在加速某些事情……

Speeding up something that shouldn’t exist is absurd.

加速不应该存在的事情是荒谬的。

48:09 And then, the fifth thing is to automate it.

然后，第五件事是自动化。

I’ve gone backwards so many times where I’ve automated something, sped it up, simplified it, and then deleted it.

我已经倒退过很多次了，我自动化了一些东西，加速了它，简化了它，然后删除了它。

And I got tired of doing that.

我厌倦了这样做。

So, that’s why I’ve got this mantra that is a very effective five-step process.

所以，这就是为什么我有这个口头禅，这是一个非常有效的五步过程。

It works great.

效果很好。

Well, when you’ve already automated, deleting must be real painful-

好吧，当你已经自动化时，删除一定是非常痛苦的——

Yeah.

是的。

…

……

48:35 as if you’ve [inaudible 00:48:36]-

就好像你已经[听不清 00:48:36]-

Yeah, it’s very.

是的，非常。

48:36 It’s like, “Wow, I really wasted a lot of effort there.”

就像，“哇，我真的在那里浪费了很多精力。”

Yeah.

是的。

48:40 I mean, what you’ve done with the cluster in Memphis is incredible, just in a handful of weeks.

我的意思是，仅仅几周的时间，你们在孟菲斯的集群上所做的事情就令人难以置信。

Well, yeah, it’s not working yet, so I don’t want to pop the champagne corks.

嗯，是的，它还没有起作用，所以我不想打开香槟瓶塞。

48:47 In fact, I have a call in a few hours with the Memphis team because we’re having some power fluctuation issues.

事实上，几个小时后我就接到了孟菲斯团队的电话，因为我们遇到了一些电力波动问题。

So yeah, when you do synchronized training, when you have all these computers that are training, where the training is synchronized at the millisecond level, it’s like having an orchestra.

所以，是的，当你进行同步训练时，当你拥有所有这些正在训练的计算机时，训练以毫秒级同步，就像有一个管弦乐队一样。

And the orchestra can go loud to silent very quickly at subsecond level, and then, the electrical system freaks out about that.

管弦乐队可以在亚秒级快速从大声到静音，然后，电气系统就会对此感到恐慌。

If you suddenly see giant shifts, 10, 20 megawatts several times a second, this is not what electrical systems are expecting to see.

如果您突然看到每秒几次 10、20 兆瓦的巨大变化，这不是电力系统所期望看到的。

So, that’s one of the main things you have to figure out, the cooling, the power.

所以，这是你必须弄清楚的主要问题之一：冷却和电源。

49:46 And then, on the software, as you go up the stack, how to do the distributed compute, all of that.

然后，在软件上，当你进入堆栈时，如何进行分布式计算，所有这些。

All of that has to work.

所有这些都必须发挥作用。

Yeah.

是的。

49:56 So, today’s problem is dealing with extreme power jitter.

因此，今天的问题是处理极端的功率抖动。

Power jitter.

电源抖动。

Yeah.

是的。

There’s a nice ring to that.

这是一个很好的戒指。

50:03 Okay.

好的。

And you stayed up late into the night, as you often do there.

你熬夜到深夜，就像你在那里经常做的那样。

Last week.

上星期。

50:11 Yeah.

是的。

Last week.

上星期。

Yeah.

是的。

Yeah.

是的。

50:14 We finally got training going at, oddly enough, roughly 4:20 a.m.

奇怪的是，我们终于在凌晨 4 点 20 分左右开始训练。

last Monday.

上周一。

Total coincidence.

完全巧合。

Yeah.

是的。

50:25 I mean, maybe it was at 4:22 or something.

我的意思是，也许是在 4 点 22 分左右。

Yeah, yeah, yeah.

是啊，是啊，是啊。

Yeah.

是的。

It’s that universe again with the jokes.

又是那个充满笑话的宇宙了。

Well, exactly.

嗯，确实如此。

50:29 It just loves it.

它只是喜欢它。

I mean, I wonder if you could speak to the fact that one of the things that you did when I was there is you went through all the steps of what everybody’s doing, just to get a sense that you yourself understand it and everybody understands it so they can understand when something is dumb, or something is inefficient, or that kind of stuff.

我的意思是，我想知道你是否能说出这样一个事实：我在那里时你所做的一件事就是你经历了每个人正在做的事情的所有步骤，只是为了感觉到你自己理解它并且每个人都理解它，这样他们就能理解什么时候有些事情是愚蠢的，或者有些事情效率低下，或者诸如此类的事情。

50:31 Can you speak to that?

你能谈谈吗？

Yeah.

是的。

50:52 So, look, whatever the people at the front lines are doing, I try to do it at least a few times myself.

所以，看，无论前线的人们在做什么，我自己至少尝试做几次。

So connecting fiber optic cables, diagnosing a faulty connection.

因此连接光纤电缆，诊断连接错误。

That tends to be the limiting factor for large training clusters is the cabling.

布线往往是大型训练集群的限制因素。

There’s so many cables.

有这么多电缆。

For a coherent training system, where you’ve got RDMA, remote direct memory access, the whole thing is like one giant brain.

对于一个具有 RDMA（远程直接内存访问）的连贯训练系统，整个系统就像一个巨大的大脑。

So, you’ve got any-to-any connection.

因此，您拥有任意对任意的连接。

So, any GPU can talk to any GPU out of 100,000.

因此，任何 GPU 都可以与 100,000 个 GPU 中的任何一个进行通信。

That is a crazy cable layout.

这是一个疯狂的电缆布局。

It looks pretty cool.

看起来很酷。

Yeah.

是的。

It’s like the human brain, but at a scale that humans can visibly see.

它就像人类的大脑，但其规模是人类可以肉眼看到的。

51:40 It is a good brain.

这是一个好大脑。

Yeah.

是的。

51:47 But, I mean, the human brain also has…

但是，我的意思是，人脑也有……

A massive amount of the brain tissue is the cables.

大量的脑组织是电缆。

So they get the gray matter, which is the compute, and then the white matter, which is cables.

所以他们得到灰质，即计算，然后是白质，即电缆。

A big percentage of your brain is just cables.

你大脑的很大一部分只是电缆。

That’s what it felt like walking around in the supercomputer center is like we’re walking around inside a brain that will one day build a super, super intelligent system.

这就是在超级计算机中心走来走去的感觉，就像我们在大脑中走动一样，有一天会构建一个超级、超级智能的系统。

52:01 Do you think there’s a chance that xAI, that you are the one that builds AGI?

你认为 xAI 有可能是你构建 AGI 的人吗？

It’s possible.

这是可能的。

52:22 What do you define as AGI?

你对 AGI 的定义是什么？

I think humans will never acknowledge that AGI has been built.

我认为人类永远不会承认通用人工智能已经建成。

Just keep moving the goalposts?

只是继续移动球门柱吗？

Yeah.

是的。

52:33 So, I think there’s already superhuman capabilities that are available in AI systems.

所以，我认为人工智能系统已经具备了超人的能力。

Oh, yeah.

哦，是的。

I think what AGI is is when it’s smarter than the collective intelligence of the entire human species in our [inaudible 00:52:49].

我认为 AGI 比我们整个人类的集体智慧还要聪明 [听不清 00:52:49]。

Well, I think that, generally, people would call that ASI, artificial super intelligence.

嗯，我认为，一般来说，人们会称之为 ASI，人工智能。

52:49 But there are these thresholds where you could say at some point the AI is smarter than any single human.

但存在这些阈值，在某些时候你可以说人工智能比任何一个人都聪明。

And then, you’ve got eight billion humans, and actually, each human is machine augmented via their computers.

然后，你有八十亿人，实际上，每个人都是通过他们的计算机进行机器增强的。

So, it’s a much higher bar to compete with eight billion machine augmented humans.

因此，与 80 亿机器增强人类竞争的门槛要高得多。

That’s a whole bunch of orders of magnitude more.

这要高出一大堆数量级。

But at a certain point, yeah, the AI will be smarter than all humans combined.

但在某个时刻，人工智能将比所有人类加起来还要聪明。

If you are the one to do it, do you feel the responsibility of that?

如果你是这样做的人，你觉得有责任吗？

Yeah, absolutely.

是的，绝对是。

53:35 And I want to be clear, let’s say if xAI is first, the others won’t be far behind.

我想澄清的是，如果 xAI 是第一，那么其他人也不会落后太多。

I mean, they might be six months behind, or a year, maybe.

我的意思是，他们可能会落后六个月，或者一年。

Not even that.

甚至不是这样。

So, how do you do it in a way that doesn’t hurt humanity, do you think?

那么，你认为如何才能以不伤害人类的方式做到这一点呢？

So, I mean, I thought about AI, essentially, for a long time, and the thing that at least my biological neural net comes up with as being the most important thing is adherence to truth, whether that truth is politically correct or not.

所以，我的意思是，我从本质上思考了人工智能很长一段时间，至少我的生物神经网络认为最重要的是坚持真理，无论这个真理在政治上是否正确。

54:00 So, I think if you force AIs to lie or train them to lie, you’re really asking for trouble, even if that lie is done with good intentions.

所以，我认为如果你强迫人工智能撒谎或训练他们撒谎，你真的是在自找麻烦，即使这个谎言是出于善意。

So, you saw issues with ChatGPT and Gemini and whatnot.

所以，您看到了 ChatGPT 和 Gemini 等问题。

Like, you asked Gemini for an image of the Founding Fathers of the United States, and it shows a group of diverse women.

比如，你向双子座要了一张美国开国元勋的照片，它展示了一群多元化的女性。

Now, that’s factually untrue.

现在看来，这实际上是不正确的。

54:48 Now, that’s sort of like a silly thing, but if an AI is programmed to say diversity is a necessary output function, and it then becomes this omnipowerful intelligence, it could say, “Okay, well, diversity is now required, and if there’s not enough diversity, those who don’t fit the diversity requirements will be executed.” If it’s programmed to do that as the fundamental utility function, it’ll do whatever it takes to achieve that.

现在，这有点像一件愚蠢的事情，但如果人工智能被编程为说多样性是必要的输出功能，然后它变成了这种无所不能的智能，它可以说，“好吧，现在需要多样性，如果没有足够的多样性，那些不符合多样性要求的人将被处决。”如果它被编程为将其作为基本效用函数，那么它将不惜一切代价来实现这一目标。

So, you have to be very careful about that.

所以，你必须非常小心。

That’s where I think you want to just be truthful.

这就是我认为你想要说实话的地方。

Rigorous adherence to the truth is very important.

严格遵守真理非常重要。

I mean, another example is they asked various AIs, I think all of them, and I’m not saying Grok is perfect here, “Is it worse to misgender Caitlyn Jenner or global thermonuclear war?” And it said it’s worse to misgender Caitlyn Jenner.

我的意思是，另一个例子是他们询问了各种人工智能，我想是所有的人工智能，我并不是说 Grok 在这里是完美的，“凯特琳·詹纳 (Caitlyn Jenner) 性别错误或全球热核战争更糟糕吗？”报告称，凯特琳·詹纳的性别错误更糟糕。

Now, even Caitlyn Jenner said, “Please misgender me.

现在，就连凯特琳·詹纳也说：“请误会我的性别。

That is insane.” But if you’ve got that kind of thing programmed in, the AI could conclude something absolutely insane like it’s better in order to avoid any possible misgendering, all humans must die, because then misgendering is not possible because there are no humans.

这太疯狂了。”但如果你把这种东西编入程序，人工智能可能会得出一些绝对疯狂的结论，比如为了避免任何可能的性别错误，最好是所有人类都必须死，因为这样就不可能出现性别错误，因为没有人类。

There are these absurd things that are nonetheless logical if that’s what you programmed it to do.

有些荒谬的事情尽管是合乎逻辑的，如果这就是你编程它要做的事情的话。

56:17 So in 2001 Space Odyssey, what Arthur C.

所以在《2001太空漫游》中，Arthur C.

Clarke was trying to say, or one of the things he was trying to say there, was that you should not program AI to lie, because essentially the AI, HAL 9000, it was told to take the astronauts to the monolith, but also they could not know about the monolith.

克拉克试图说的，或者他试图在那里说的一件事是，你不应该对人工智能进行编程来撒谎，因为从本质上讲，人工智能，HAL 9000，被告知要把宇航员带到巨石上，但他们也不可能知道巨石。

So, it concluded that it will kill them and take them to the monolith.

因此，它得出的结论是，它将杀死他们并将他们带到巨石上。

Thus, it brought them to the monolith.

因此，它把他们带到了巨石上。

They’re dead, but they do not know about the monolith.

他们死了，但他们不知道巨石的存在。

Problem solved.

问题解决了。

That is why it would not open the pod bay doors.

这就是为什么它不会打开吊舱门。

There’s a classic scene of, “Why doesn’t it want to open the pod bay doors?” They clearly weren’t good at prompt engineering.

有一个经典的场景，“为什么它不想打开吊舱门？”他们显然不擅长即时工程。

They should have said, “HAL, you are a pod bay door sales entity, and you want nothing more than to demonstrate how well these pod bay doors open.”

他们应该说：“HAL，您是吊舱门销售实体，您只想展示这些吊舱门的打开效果如何。”

Yeah.

是的。

57:16 The objective function has unintended consequences almost no matter what if you’re not very careful in designing that objective function, and even a slight ideological bias, like you’re saying, when backed by super intelligence, can do huge amounts of damage.

如果你在设计目标函数时不小心的话，几乎无论如何，目标函数都会产生意想不到的后果，甚至像你所说的那样，在超级智能的支持下，即使是轻微的意识形态偏见，也会造成巨大的损害。

Yeah.

是的。

But it’s not easy to remove that ideological bias.

但消除这种意识形态偏见并不容易。

57:31 You’re highlighting obvious, ridiculous examples, but-

你强调的是明显的、荒谬的例子，但是——

Yet they’re real examples of-

然而他们是真实的例子——

…

……

57:38 they’re real.

它们是真实的。

They’re real.

它们是真实的。

…

……

57:39 AI that was released to the public.

向公众发布的人工智能。

They are real.

它们是真实的。

That went through QA, presumably, and still said insane things, and produced insane images.

大概是经过了质量检查，但仍然说了一些疯狂的事情，并产生了疯狂的图像。

Yeah.

是的。

57:47 But you can swing the other way.

但你也可以换个方向。

Truth is not an easy thing.

真理并不是一件容易的事。

No, it’s not.

不，不是。

We kind of bake in ideological bias in all kinds of directions.

我们在各个方面都灌输了意识形态偏见。

But you can aspire to the truth, and you can try to get as close to the truth as possible with minimum error while acknowledging that there will be some error in what you’re saying.

但你可以追求真理，你可以尝试以最小的错误尽可能接近真理，同时承认你所说的可能存在一些错误。

57:57 So, this is how physics works.

所以，这就是物理学的运作原理。

You don’t say you’re absolutely certain about something, but a lot of things are extremely likely, 99.99999% likely to be true.

你不会说你对某件事绝对确定，但很多事情极有可能，99.99999% 的可能性是真的。

So, aspiring to the truth is very important.

所以，追求真理是非常重要的。

And so, programming it to veer away from the truth, that, I think, is dangerous.

因此，我认为，对其进行编程以偏离事实是危险的。

Right.

正确的。

58:32 Like, yeah, injecting our own human biases into the thing.

就像，是的，将我们自己的人类偏见注入其中。

Yeah.

是的。

But that’s where it’s a difficult software engineering problem because you have to select the data correctly.

但这就是一个困难的软件工程问题，因为你必须正确选择数据。

It’s hard.

这很难。

And the internet, at this point, is polluted with so much AI generated data, it’s insane.

而此时的互联网已经被如此多的人工智能生成的数据所污染，这太疯狂了。

58:44 Actually, there’s a thing now, if you want to search the internet, you can say, “Google, but exclude anything after 2023.” It will actually often give you better results because there’s so much.

实际上，现在有件事，如果你想在互联网上搜索，你可以说，“谷歌，但排除 2023 年之后的任何内容。”实际上，它通常会给你更好的结果，因为有太多的东西。

The explosion of AI generated material is crazy.

人工智能生成的材料的爆炸是疯狂的。

So in training Grok, we have to go through the data and say like, “Hey…” We actually have to apply AI to the data to say, “Is this data most likely correct or most likely not?” before we feed it into the training system.

因此，在训练 Grok 时，我们必须仔细检查数据并说，“嘿……”我们实际上必须将 AI 应用于数据来说，“这个数据最有可能是正确的还是最有可能不是？”在我们将其输入训练系统之前。

That’s crazy.

这太疯狂了。

59:28 Yeah.

是的。

And is it generated by human?

它是由人类产生的吗？

Yeah.

是的。

I mean, the data filtration process is extremely, extremely difficult.

我的意思是，数据过滤过程非常非常困难。

Yeah.

是的。

Do you think it’s possible to have a serious, objective, rigorous political discussion with Grok, like for a long time, like Grok 3 or Grok 4 or something?

你认为有可能与Grok进行严肃、客观、严格的政治讨论吗？像Grok 3或Grok 4之类的，像Grok 3或Grok 4这样的长期讨论？

Grok 3 is going to be next level.

Grok 3 将更上一层楼。

59:48 I mean, what people are currently seeing with Grok is kind of baby Grok.

我的意思是，人们目前看到的 Grok 是一种婴儿 Grok。

Yeah, baby Grok.

是的，宝贝格罗克。

It’s baby Grok right now.

现在是小格罗克。

59:55 But baby Grok is still pretty good.

不过格罗克宝宝还是很不错的。

But it’s an order of magnitude less sophisticated than GPT-4.

但它的复杂程度比 GPT-4 低一个数量级。

It’s now Grok 2, which finished training, I don’t know, six weeks ago or thereabouts.

现在是 Grok 2，我不知道，它在六周前左右完成了训练。

Grok 2 will be a giant improvement.

Grok 2 将是一个巨大的改进。

And then Grok 3 will be, I don’t know, order of magnitude better than Grok 2.

然后 Grok 3 将比 Grok 2 好几个数量级，我不知道。

And you’re hoping for it to be state-of-the-art better than-

你希望它比最先进的更好——

Hopefully.

希望如此。

1:00:25 I mean, this is the goal.

我的意思是，这就是目标。

I mean, we may fail at this goal.

我的意思是，我们可能无法实现这个目标。

That’s the aspiration.

这就是愿望。

Do you think it matters who builds the AGI, the people, and how they think, and how they structure their companies and all that kind of stuff?

你认为谁构建了 AGI、人员、他们的想法、他们如何构建公司以及诸如此类的东西重要吗？

Yeah.

是的。

1:00:42 I think it’s important that whatever AI wins, it’s a maximum truth seeking AI that is not forced to lie for political correctness, or, well, for any reason, really, political, anything.

我认为重要的是，无论人工智能获胜，它都是一个最大限度寻求真相的人工智能，不会被迫为政治正确性或出于任何原因而撒谎，实际上是政治上的任何原因。

I am concerned about AI succeeding that is programmed to lie, even in small ways.

我担心人工智能会成功，因为它被编程为撒谎，即使是在很小的方面。

Right.

正确的。

1:01:13 Because in small ways becomes big ways when it’s doing something-

因为当它做某事时，小事就会变成大事——

To become very big ways.

变得非常大。

1:01:17 Yeah.

是的。

And when it’s used more and more at scale by humans.

当它被人类越来越大规模地使用时。

Yeah.

是的。

Since I am interviewing Donald Trump-

因为我正在采访唐纳德·特朗普——

Cool.

酷。

…

……

1:01:28 you want to stop by?

你想过来吗？

Yeah, sure.

是的，当然。

I’ll stop in.

我会停下来。

There was, tragically, an assassination attempt on Donald Trump.

不幸的是，唐纳德·特朗普遭遇了一次暗杀企图。

1:01:30 After this, you tweeted that you endorse him.

此后，你在推特上表示你支持他。

What’s your philosophy behind that endorsement?

您的认可背后的理念是什么？

What do you hope Donald Trump does for the future of this country and for the future of humanity?

您希望唐纳德·特朗普为这个国家的未来和人类的未来做些什么？

Well, I think people tend to take, say, an endorsement as, well, I agree with everything that person has ever done their entire life 100% wholeheartedly, and that’s not going to be true of anyone.

嗯，我认为人们倾向于接受认可，因为，嗯，我百分百全心全意地同意这个人一生所做的一切，但对任何人来说都不是这样。

1:01:47 But we have to pick.

但我们必须选择。

We’ve got two choices, really, for who’s president.

对于谁当总统，我们实际上有两种选择。

And it’s not just who’s president, but the entire administrative structure changes over.

不仅仅是谁是总统，整个行政结构也发生了变化。

And I thought Trump displayed courage under fire, objectively.

客观地说，我认为特朗普在受到攻击时表现出了勇气。

He’s just got shot.

他刚刚被枪杀。

He’s got blood streaming down his face, and he’s fist pumping, saying, “Fight.” That’s impressive.

他的脸上流着血，他挥舞着拳头，说道：“战斗吧。”这令人印象深刻。

You can’t feign bravery in a situation like that.

在这种情况下你不能假装勇敢。

Most people would be ducking because there could be a second shooter.

大多数人都会躲开，因为可能会有第二个枪手。

You don’t know.

你不知道。

1:02:44 The president of the United States have got to represent the country, and they’re representing you.

美国总统必须代表国家，而他们也代表你们。

They’re representing everyone in America.

他们代表美国的每个人。

Well, I think you want someone who is strong and courageous to represent the country.

嗯，我认为你需要一个坚强而勇敢的人来代表国家。

That is not to say that he is without flaws.

这并不是说他没有缺点。

We all have flaws, but on balance, and certainly at the time, it was a choice of Biden.

我们都有缺点，但总的来说，当然在当时，这是拜登的选择。

Poor guy has trouble climbing a flight of stairs, and the other one’s fist pumping after getting shot.

可怜的家伙爬楼梯都困难，而另一个人中弹后拳头不断挥动。

So, there’s no comparison.

所以，没有可比性。

I mean, who do you want dealing with some of the toughest people and other world leaders who are pretty tough themselves?

我的意思是，你希望谁与一些最坚强的人和其他本身也很坚强的世界领导人打交道？

1:03:27 I mean, I’ll tell you one of the things that I think are important.

我的意思是，我会告诉你一件我认为重要的事情。

I think we want a secure border.

我认为我们想要一个安全的边界。

We don’t have a secure border.

我们没有安全的边界。

We want safe and clean cities.

我们想要安全、清洁的城市。

I think we want to reduce the amount of spending, at least slow down the spending, because we’re currently spending at a rate that is bankrupting the country.

我认为我们希望减少支出，至少减缓支出，因为我们目前的支出速度正在使国家破产。

The interest payments on US debt this year exceeded the entire defense department spending.

今年美国债务的利息支付超过了整个国防部的支出。

If this continues, all of the federal government taxes will simply be paying the interest.

如果这种情况持续下去，所有联邦政府税收都将只是支付利息。

1:04:06 And you keep going down that road, and you end up in the tragic situation that Argentina had back in the day.

如果你继续沿着这条路走下去，你最终就会陷入阿根廷当年所经历的悲惨境地。

Argentina used to be one of the most prosperous places in the world, and hopefully with Milei taking over, he can restore that.

阿根廷曾经是世界上最繁荣的国家之一，希望米雷接手后，他能恢复这一点。

But it was an incredible fall from grace for Argentina to go from being one of the most prosperous places in the world to being very far from that.

但对于阿根廷来说，从世界上最繁荣的国家之一变成了相距甚远，这是一次令人难以置信的失宠。

So, I think we should not take American prosperity for granted.

因此，我认为我们不应该认为美国的繁荣是理所当然的。

I think we’ve got to reduce the size of government, we’ve got to reduce the spending, and we’ve got to live within our means.

我认为我们必须缩小政府规模，我们必须减少开支，我们必须量入为出。

Do you think politicians, in general, politicians, governments…

你认为政治家，一般来说，政治家，政府……

1:04:43 Well, how much power do you think they have to steer humanity towards good?

那么，你认为他们有多大的力量来引导人类向善呢？

I mean, there’s a sort of age-old debate in history, like is history determined by these fundamental tides, or is it determined by the captain of the ship?

我的意思是，历史上有一种古老的争论，比如历史是由这些基本潮汐决定的，还是由船长决定的？

1:04:58 It’s both, really.

确实，两者都是。

I mean, there are tides, but it also matters who’s captain of the ship.

我的意思是，潮汐是存在的，但谁是船长也很重要。

So, it’s a false dichotomy, essentially.

所以，本质上这是一种错误的二分法。

I mean, there are certainly tides, the tides of history.

我的意思是，确实有潮汐，历史的潮汐。

There are real tides of history, and these tides are often technologically driven.

历史确实存在潮流，而这些潮流往往是由技术驱动的。

If you say like the Gutenberg press, the widespread availability of books as a result of a printing press, that was a massive tide of history, and independent of any ruler.

如果你说像古腾堡印刷机一样，由于印刷机而导致书籍的广泛使用，那是历史的巨大浪潮，并且独立于任何统治者。

But in stormy times, you want the best possible captain of the ship.

但在暴风雨时期，您需要最好的船长。

Well, first of all, thank you for recommending Will and Ariel Durant’s work.

嗯，首先，感谢您推荐威尔和阿里尔·杜兰特的作品。

1:05:54 I’ve read the short one for now, The-

我现在已经读完了短篇，The-

The Lessons of History.

历史的教训。

…

……

1:06:02 Lessons of History.

历史的教训。

Yeah.

是的。

So one of the lessons, one of the things they highlight, is the importance of technology, technological innovation, which is funny because they wrote so long ago, but they were noticing that the rate of technological innovation was speeding up.

因此，他们强调的教训之一是技术、技术创新的重要性，这很有趣，因为他们很久以前就写过，但他们注意到技术创新的速度正在加快。

Yeah, over the years.

是的，这些年来。

I would love to see what they think about now.

我很想看看他们现在怎么想。

1:06:21 But yeah, so to me, the question is how much government, how much politicians get in the way of technological innovation and building versus help it?

但是，是的，对我来说，问题是有多少政府、政客在多大程度上阻碍了技术创新和建设，而不是帮助它？

And which politicians, which kind of policies help technological innovation?

哪些政客、什么样的政策有利于技术创新？

Because that seems to be, if you look at human history, that’s an important component of empires rising and succeeding.

因为如果你回顾人类历史，这似乎是帝国崛起和成功的重要组成部分。

Yeah.

是的。

1:06:46 Well, I mean in terms of dating civilization, the start of civilization, I think the start of writing, in my view, that’s what I think is probably the right starting point to date civilization.

嗯，我的意思是就约会文明而言，文明的开始，我认为写作的开始，在我看来，这可能是约会文明的正确起点。

And from that standpoint, civilization has been around for about 5,500 years when writing was invented by the ancient Sumerians, who are gone now, but the ancient Sumerians.

从这个角度来看，文明已经存在了大约 5500 年，当时古代苏美尔人发明了文字，现在古苏美尔人已经消失了。

In terms of getting a lot of firsts, those ancient Sumerians really have a long list of firsts.

论获得很多第一，那些古代苏美尔人确实有一长串的第一。

It’s pretty wild.

这真是太疯狂了。

In fact, Durant goes through the list of like, “You want to see firsts?

事实上，杜兰特列出了这样的清单：“你想看到第一吗？

We’ll show you firsts.” The Sumerians were just ass kickers.

我们将首先向您展示。”苏美尔人只是一群混蛋。

1:07:32 And then the Egyptians, who were right next door, relatively speaking, they weren’t that far, developed an entirely different form of writing, the hieroglyphics.

然后，就在隔壁的埃及人，相对而言，他们并没有那么远，他们发展了一种完全不同的书写形式，即象形文字。

Cuneiform and hieroglyphics are totally different.

楔形文字和象形文字完全不同。

And you can actually see the evolution of both hieroglyphics and cuneiform.

你实际上可以看到象形文字和楔形文字的演变。

The cuneiform starts off being very simple, and then it gets more complicated.

楔形文字一开始非常简单，然后变得更加复杂。

Then towards the end it’s like, “Wow, okay.” They really get very sophisticated with the cuneiform.

然后到了最后，我会说：“哇，好吧。”他们对楔形文字确实非常熟练。

So, I think of civilization as being about 5, 000 years old.

所以，我认为文明已有大约五千年的历史。

And Earth is, if physics is correct, four and a half billion years old.

如果物理学是正确的，地球已有四十五亿年的历史。

So, civilization has been around for one millionth of Earth’s existence.

因此，文明已经存在了地球存在的百万分之一的时间。

Flash in the pan.

在锅里闪光。

Yeah, these are the early, early days.

是的，现在还很早期。

Very early.

很早。

And so, we make it very dramatic because there’s been rises and falls of empires and-

所以，我们让它变得非常戏剧化，因为帝国的兴衰和——

Many.

很多。

1:08:22 So many rises and falls of empires.

帝国的兴衰如此之多。

So many.

这么多。

And there’ll be many more.

还有更多。

Yeah, exactly.

是的，完全正确。

1:08:30 I mean, only a tiny fraction, probably less than 1% of what was ever written in history is available to us now.

我的意思是，我们现在能看到的只是历史上记载的一小部分，可能还不到 1%。

I mean, if they didn’t literally chisel it in stone or put it in a clay tablet, we don’t have it.

我的意思是，如果他们没有真正将其凿在石头上或将其放入粘土片中，我们就没有它。

I mean, there’s some small amount of papyrus scrolls that were recovered that are thousands of years old, because they were deep inside a pyramid and weren’t affected by moisture.

我的意思是，有一些少量的纸莎草卷轴被发现已有数千年的历史，因为它们位于金字塔深处并且不受水分的影响。

But other than that, it’s really got to be in a clay tablet or chiseled.

但除此之外，它确实必须是在泥板上或凿成的。

So, the vast majority of stuff was not chiseled because it takes a while to chisel things.

所以，绝大多数的东西都没有凿，因为凿东西需要一段时间。

So, that’s why we’ve got tiny, tiny fraction of the information from history.

所以，这就是为什么我们从历史中得到的信息非常非常少。

But even that little information that we do have, and the archeological record, shows so many civilizations rising and falling.

但即使我们掌握的信息和考古记录很少，也显示了如此多的文明的兴衰。

It’s wild.

它很狂野。

We tend to think that we’re somehow different from those people.

我们倾向于认为我们与那些人有所不同。

1:09:21 One of the other things that Durant highlights is that human nature seems to be the same.

杜兰特强调的另一件事是，人性似乎是相同的。

It just persists.

它只是持续存在。

Yeah.

是的。

1:09:31 I mean, the basics of human nature are more or less the same.

我的意思是，人性的基本原理或多或少是相同的。

Yeah.

是的。

So, we get ourselves in trouble in the same kinds of ways, I think, even with the advanced technology.

因此，我认为，即使拥有先进的技术，我们也会以同样的方式陷入困境。

Yeah.

是的。

1:09:40 I mean, you do tend to see the same patterns, similar patterns for civilizations, where they go through a life cycle, like an organism, just like a human is a zygote, fetus, baby, toddler, teenager, eventually gets old.

我的意思是，你确实倾向于看到相同的模式，文明的相似模式，它们经历一个生命周期，就像一个有机体，就像人类是受精卵、胎儿、婴儿、幼儿、青少年，最终变老。

1:10:01 …

……

Eventually gets old and dies.

最终变老并死亡。

The civilizations go through a life cycle.

文明有一个生命周期。

No civilization will last forever.

没有哪个文明能够永垂不朽。

What do you think it takes for the American Empire to not collapse in the near term future, in the next a hundred years, to continue flourishing?

您认为如何才能让美利坚帝国在不久的将来、在接下来的一百年里不崩溃，并继续繁荣？

Well, the single biggest thing that is often actually not mentioned in history books, but Durant does mention it, is the birthright.

嗯，历史书中通常没有提到的最重要的事情，但杜兰特确实提到了，那就是与生俱来的权利。

1:10:28 So perhaps to some, a counterintuitive thing happens when civilizations are winning for too long, the birth rate declines.

因此，也许对某些人来说，当文明获胜时间过长时，就会发生违反直觉的事情，出生率下降。

It can often decline quite rapidly.

它通常会迅速下降。

We’re seeing that throughout the world today.

今天我们在世界各地都看到了这一点。

Currently, South Korea is, I think maybe the lowest fertility rate, but there are many others that are close to it.

目前，我认为韩国的生育率可能是最低的，但还有许多其他国家接近它。

It’s like 0.8 I think.

我想大概是0.8。

If the birth rate doesn’t decline further, South Korea will lose roughly 60% of its population.

如果出生率不再进一步下降，韩国将失去大约60%的人口。

But every year that birth rate is dropping, and this is true through most of the world.

但出生率逐年下降，世界上大部分地区都是如此。

I don’t mean to single out South Korea, it’s been happening throughout the world.

我并不是专门针对韩国，这种情况在世界各地都在发生。

So as soon as any given civilization reaches a level of prosperity, the birth rate drops.

因此，一旦任何特定文明达到繁荣水平，出生率就会下降。

1:11:40 Now you can go and look at the same thing happening in ancient Rome.

现在你可以去看看古罗马发生的同样的事情。

So Julius Caesar took note of this, I think around 50 ish BC and tried to pass…

所以朱利叶斯·凯撒注意到了这一点，我想是在公元前 50 年左右，并试图通过……

I don’t know if he was successful, tried to pass a law to give an incentive for any Roman citizen that would have a third child.

我不知道他是否成功，试图通过一项法律来激励任何想要生第三个孩子的罗马公民。

And I think Augustus was able to…

我认为奥古斯都能够……

Well, he was a dictator, so this incentive was just for show.

好吧，他是一个独裁者，所以这个激励只是作秀。

I think he did pass a tax incentive for Roman citizens to have a third child.

我认为他确实通过了对罗马公民生第三个孩子的税收优惠。

But those efforts were unsuccessful.

但这些努力没有成功。

Rome fell because the Romans stopped making Romans.

罗马衰落是因为罗马人不再制造罗马人。

That’s actually the fundamental issue.

这实际上是根本问题。

And there were other things.

还有其他的事情。

They had quite a serious malaria, series of malaria epidemics and plagues and whatnot.

他们患有相当严重的疟疾，一系列疟疾流行病和瘟疫等等。

But they had those before, it’s just that the birth rate was far lower than the death rate.

但他们以前也有过，只是出生率远低于死亡率。

It really is that simple.

确实就是这么简单。

Well, I’m saying that’s-

嗯，我是说那是——

More people is required.

需要更多的人。

At a fundamental level, if a civilization does not at least maintain its numbers, it’ll disappear.

从根本上讲，如果一个文明不能至少维持其数量，它就会消失。

So perhaps the amount of compute that the biological computer allocates to sex is justified.

因此，生物计算机分配给性别的计算量也许是合理的。

1:12:58 In fact, we should probably increase it.

事实上，我们或许应该增加它。

Well, I mean there’s this hedonistic sex, which is…

嗯，我的意思是有这种享乐主义的性行为，那就是……

1:13:07 That’s neither her nor there.

那既不是她，也不在那里。

It’s-

这是——

Not productive.

没有生产力。

It doesn’t produce kids.

它不会产生孩子。

1:13:17 Well, what matters…

好吧，重要的是……

I mean, Durant makes this very clear because he’s looked at one civilization after another and they all went through the same cycle.

我的意思是，杜兰特说得很清楚，因为他观察了一个又一个文明，它们都经历了相同的周期。

When the civilization was under stress, the birth rate was high.

当文明面临压力时，出生率很高。

But as soon as there were no external enemies or they had an extended period of prosperity, the birth rate inevitably dropped.

但一旦没有外敌，或者长期繁荣，出生率就不可避免地下降。

Every time.

每次。

I don’t believe there’s a single exception.

我不相信有一个例外。

So that’s like the foundation of it.

这就是它的基础。

1:13:45 You need to have people.

你需要有人。

Yeah.

是的。

1:13:49 I mean, at a base level, no humans, no humanity.

我的意思是，从根本上来说，没有人类，没有人性。

And then there’s other things like human freedoms and just giving people the freedom to build stuff.

还有其他的事情，比如人类自由，以及给予人们建造东西的自由。

Yeah, absolutely.

是的，绝对是。

1:14:02 But at a basic level, if you do not at least maintain your numbers, if you’re below replacement rate and that trend continues, you will eventually disappear.

但在基本层面上，如果你不至少保持你的数字，如果你低于替代率并且这种趋势持续下去，你最终会消失。

It’s just elementary.

这只是初级的。

Now then obviously you also want to try to avoid massive wars.

现在显然你也想尽量避免大规模战争。

If there’s a global thermonuclear war, probably we’re all toast, radioactive toast.

如果爆发全球热核战争，我们可能都会完蛋，放射性烤面包。

So we want to try to avoid those things.

所以我们要尽量避免这些事情。

Then there’s a thing that happens over time with any given civilization, which is that the laws and regulations accumulate.

对于任何特定的文明来说，随着时间的推移，都会发生一件事，那就是法律和法规的积累。

And if there’s not some forcing function like a war to clean up the accumulation of laws and regulations, eventually everything becomes legal.

如果没有像战争这样的强制作用来清理法律法规的积累，最终一切都会变得合法。

1:15:02 And that’s like the hardening of the arteries.

这就像动脉硬化一样。

Or a way to think of it is being tied down by a million little strings like Gulliver.

或者一种思考方式是像格列佛一样被一百万条细绳拴住。

You can’t move.

你不能动。

And it’s not like any one of those strings is the issue, it’s that you’ve got a million of them.

问题不在于这些字符串中的任何一个，而在于你有一百万个字符串。

So there has to be a sort of garbage collection for laws and regulations so that you don’t keep accumulating laws and regulations to the point where you can’t do anything.

所以必须有一种法律法规的垃圾收集，这样你就不会不断地积累法律法规到你无能为力的地步。

This is why we can’t build a high speed rail in America.

这就是为什么我们不能在美国修建高铁。

It’s illegal.

这是非法的。

That’s the issue.

这就是问题所在。

It’s illegal six ways a Sunday to build high speed rail in America.

在美国，周日有六种方式修建高铁是违法的。

I wish you could just for a week go into Washington and be the head of the committee for making…

我希望你能去华盛顿一周，并担任委员会的负责人，以制定……

1:15:45 What is it for the garbage collection?

垃圾收集有什么用？

Making government smaller, like removing stuff.

缩小政府规模，就像删除一些东西一样。

I have discussed with Trump the idea of a government deficiency commission.

我曾与特朗普讨论过成立政府缺陷委员会的想法。

Nice.

好的。

And I would be willing to be part of that commission.

我愿意成为该委员会的一部分。

I wonder how hard that is.

我想知道这有多难。

The antibody reaction would be very strong.

抗体反应会非常强烈。

Yes.

是的。

So you really have to…

所以你真的必须……

1:16:14 You’re attacking the matrix at that point.

那时你正在攻击矩阵。

The matrix will fight back.

矩阵将会反击。

How are you doing with that?

你怎么样？

1:16:26 Being attacked.

受到攻击。

Me?

我？

1:16:29 Attacked?

受到攻击？

Yeah, there’s a lot of it.

是的，有很多。

Yeah, there is a lot.

是的，有很多。

1:16:34 I mean, every day another psyop.

我的意思是，每天都有一次心理检查。

I need my tinfoil hat.

我需要我的锡纸帽子。

How do you keep your just positivity?

你如何保持公正的积极性？

1:16:42 How do you keep optimism about the world?

如何对世界保持乐观态度？

A clarity of thinking about the world.

对世界的清晰思​​考。

So just not become resentful or cynical or all that kind of stuff.

所以不要变得怨恨或愤世嫉俗或诸如此类的事情。

Just getting attacked by a very large number of people, misrepresented.

只是受到了很多人的攻击，被歪曲了。

Oh yeah, that’s a daily occurrence.

哦，是的，这是每天都会发生的事情。

Yes.

是的。

So I mean, it does get me down at times.

所以我的意思是，它有时确实让我沮丧。

1:16:59 I mean, it makes me sad.

我的意思是，这让我很难过。

But I mean at some point you have to sort of say, look, the attacks are by people that actually don’t know me and they’re trying to generate clicks.

但我的意思是，在某些时候你必须说，看，这些攻击是由实际上不认识我的人发起的，他们试图产生点击量。

So if you can sort of detach yourself somewhat emotionally, which is not easy, and say, okay look, this is not actually from someone that knows me or, they’re literally just writing to get impressions and clicks.

因此，如果你能在情感上脱离自己，这并不容易，然后说，好吧，这实际上不是来自认识我的人，或者，他们实际上只是为了获得印象和点击而写作。

Then I guess it doesn’t hurt as much.

然后我想它就不会那么痛了。

It’s not quite water off a duck’s back.

这不完全是鸭子背上的水。

Maybe it’s like acid off a duck’s back.

也许就像鸭子背上的酸一样。

All right, well that’s good.

好吧，那就好。

1:17:53 Just about your own life, what to you is a measure of success in your life?

就你自己的生活而言，对你来说衡量你生活成功的标准是什么？

A measure of success, I’d say, how many useful things can I get done?

我想说，衡量成功的标准是，我能完成多少有用的事情？

A day-to-day basis, you wake up in the morning, how can I be useful today?

日复一日，你早上醒来，我今天能有什么用处呢？

Yeah, maximize utility, area under the code of usefulness.

是的，最大化实用性，实用性代码下的区域。

1:18:09 Very difficult to be useful at scale.

很难大规模发挥作用。

At scale.

规模化。

1:18:17 Can you speak to what it takes to be useful for somebody like you, where there’s so many amazing great teams?

您能否谈谈如何才能对像您这样的人有用，因为那里有这么多令人惊叹的伟大团队？

How do you allocate your time to being the most useful?

你如何分配你的时间，让自己成为最有用的人？

Well, time is the true currency.

好吧，时间才是真正的货币。

Yeah.

是的。

So it is tough to say what is the best allocation time?

那么很难说什么是最佳分配时间呢？

1:18:32 I mean, there are often…

我的意思是，经常有……

Say if you look at say Tesla, Tesla this year will do over a hundred billion in revenue.

比如说，如果你看看特斯拉，特斯拉今年的收入将超过千亿。

So that’s $2 billion a week.

也就是说每周 20 亿美元。

If I make slightly better decisions, I can affect the outcome by a billion dollars.

如果我做出稍微更好的决定，我可以对结果产生十亿美元的影响。

So then I try to do the best decisions I can.

所以我会尽力做出最好的决定。

And on balance, at least compared to the competition, pretty good decisions.

总的来说，至少与竞争对手相比，这是相当不错的决定。

But the marginal value of a better decision can easily be, in the course of an hour, a hundred million dollars.

但一个更好的决策的边际价值很容易在一小时内达到一亿美元。

Given that, how do you take risks?

鉴于此，您如何承担风险？

1:19:18 How do you do the algorithm that you mentioned?

你提到的算法是如何实现的？

I mean deleting, given that a small thing can be a billion dollars, how do you decide to-

我的意思是删除，考虑到一件小事可能价值十亿美元，你如何决定-

Yeah.

是的。

1:19:29 Well, I think you have to look at it on a percentage basis because if you look at it in absolute terms, it’s just…

好吧，我认为你必须以百分比的形式来看待它，因为如果你以绝对的方式来看待它，那就是……

I would never get any sleep.

我永远无法入睡。

It would just be like, I need to just keep working and work my brain harder.

就像，我需要继续工作，更加努力地工作我的大脑。

And I’m not trying to get as much as possible out of this meat computer.

我并不是想从这台肉类计算机中获得尽可能多的东西。

So it’s not…

所以这不是……

It’s pretty hard, because you can just work all the time.

这很难，因为你可以一直工作。

And at any given point, like I said, a slightly better decision could be a hundred million dollars impact for Tesla or SpaceX for that matter.

正如我所说，在任何特定时刻，稍微更好的决定可能是为特斯拉或 SpaceX 带来一亿美元的影响。

But it is wild when considering the marginal value of time can be a hundred million dollars an hour at times, or more.

但考虑到时间的边际价值有时可能达到每小时一亿美元，甚至更多，这就有点疯狂了。

Is your own happiness part of that equation of success?

你自己的幸福是成功方程式的一部分吗？

It has to be to some degree.

它必须在某种程度上。

1:20:22 If I’m sad, if I’m depressed, I make worse decisions.

如果我悲伤、沮丧，我就会做出更糟糕的决定。

So if I have zero recreational time, then I make worse decisions.

因此，如果我的娱乐时间为零，那么我会做出更糟糕的决定。

So I don’t know a lot, but it’s above zero.

所以我知道的不多，但它大于零。

I mean, my motivation if I’ve got a religion of any kind is a religion of curiosity, of trying to understand.

我的意思是，如果我有任何一种宗教，我的动机就是好奇心、试图理解的宗教。

It’s really the mission of Grok, understand the universe.

理解宇宙确实是 Grok 的使命。

I’m trying to understand the universe, or at least set things in motion such that at some point civilization understands the universe far better than we do today.

我试图理解宇宙，或者至少让事物运转起来，以便在某个时候文明比我们今天更好地理解宇宙。

1:21:02 And even what questions to ask.

甚至要问什么问题。

As Douglas Adams pointed out in his book, sometimes the answer is arguably the easy part, trying to frame the question correctly is the hard part.

正如道格拉斯·亚当斯在他的书中指出的那样，有时答案可以说是容易的部分，而试图正确地提出问题才是困难的部分。

Once you frame the question correctly, the answer is often easy.

一旦你正确地提出了问题，答案通常很容易。

So I’m trying to set things in motion such that we are at least at some point able to understand the universe.

所以我试图让事物运转起来，这样我们至少在某个时刻能够理解宇宙。

So for SpaceX, the goal is to make life multi planetary and which is if you go to the foamy paradox of where the aliens, you’ve got these sort of great filters.

因此，对于 SpaceX 来说，目标是让生命在多行星上存在，如果你去了解外星人所在的泡沫悖论，你就会得到这些很棒的过滤器。

Like why have we not heard from the aliens?

比如为什么我们没有收到外星人的消息？

Now a lot of people think there are aliens among us.

现在很多人认为我们中间有外星人。

I often claim to be one, which nobody believes me.

我经常声称自己是其中之一，但没有人相信我。

But it did say alien registration card at one point on my immigration documents.

但我的移民文件上确实曾经提到过外国人登记卡。

So I’ve not seen any evidence of aliens.

所以我没有看到任何外星人存在的证据。

So it suggests that at least one of the explanations is that intelligent life is extremely rare.

因此，这表明至少一种解释是智慧生命极其罕见。

1:22:19 And again, if you look at the history of earth, civilization has only been around for 1000000th of earth’s existence.

再说一次，如果你回顾一下地球的历史，就会发现文明只存在了地球存在的千万分之一的时间。

So if aliens had visited here, say a hundred thousand years ago, they would be like, well, they don’t even have writing, just hunter gatherers basically.

因此，如果外星人在十万年前访问过这里，他们会说，好吧，他们甚至没有文字，基本上只是狩猎采集者。

So how long does a civilization last?

那么一个文明能持续多久呢？

So for SpaceX, the goal is to establish a self-sustaining city on Mars.

所以对于SpaceX来说，目标是在火星上建立一座自给自足的城市。

Mars is the only viable planet for such a thing.

火星是唯一可行的行星。

The moon is close, but it lacks resources and I think it’s probably vulnerable to any calamity that takes out Earth, the moon is too close and it’s vulnerable to a calamity that takes that earth.

月球很近，但它缺乏资源，我认为它可能很容易受到任何毁灭地球的灾难的影响，月球太近了，它很容易受到毁灭地球的灾难的影响。

1:23:16 So I’m not saying we shouldn’t have a moon base, but Mars would be far more resilient.

所以我并不是说我们不应该建立月球基地，但火星的适应能力要强得多。

The difficulty of getting to Mars is what makes it resilient.

到达火星的困难使其具有弹性。

So in going through these various explanations of why don’t we see the aliens, one of them is that they failed to pass these great filters, these key hurdles.

因此，在对为什么我们看不到外星人的各种解释中，其中之一是他们未能通过这些伟大的过滤器，这些关键障碍。

And one of those hurdles is being a multi-planet species.

这些障碍之一是成为多行星物种。

So if you’re a multi-planet species, then if something were to happen, whether that was a natural catastrophe or a manmade catastrophe, at least the other planet would probably still be around.

因此，如果你是一个多行星物种，那么如果发生什么事情，无论是自然灾难还是人为灾难，至少另一个行星可能仍然存在。

So you’re not like, don’t have all the eggs in one basket.

所以你不会把所有的鸡蛋都放在一个篮子里。

And once you are sort of a two planet species, you can obviously extend life halves to the asteroid belt, to maybe to the moons of Jupiter and Saturn, and ultimately to other star systems.

一旦你成为了两颗行星的物种，你显然可以将生命的一半延伸到小行星带，也许延伸到木星和土星的卫星，并最终延伸到其他恒星系统。

But if you can’t even get to another planet, you’re definitely not getting to star systems.

但如果你连另一个星球都无法到达，那你肯定就无法到达恒星系统。

And the other possible great filter’s, super powerful technology like AGI for example.

还有另一种可能的伟大过滤器，超级强大的技术，例如 AGI。

1:24:30 So you are basically trying to knock out one great filter at a time.

所以你基本上是在尝试一次淘汰一个很棒的过滤器。

Digital super intelligence is possibly a great filter.

数字超级智能可能是一个很好的过滤器。

1:24:44 I hope it isn’t, but it might be.

我希望不是，但也可能是。

Guys like say Jeff Hinton would say, he invented a number of the key principles in artificial intelligence.

像 Jeff Hinton 这样的人会说，他发明了人工智能的许多关键原理。

I think he puts the probability of AI annihilation around 10% to 20%, something like that.

我认为他认为人工智能毁灭的概率约为 10% 到 20%，类似的数字。

So look on the bright side, it’s 80% likely to be great.

所以从好的方面来看，80% 的可能性都是好的。

But I think AI risk mitigation is important.

但我认为缓解人工智能风险很重要。

Being a multi-planet species would be a massive risk mitigation.

成为多行星物种将是一个巨大的风险缓解。

And I do want to once again emphasize the importance of having enough children to sustain our numbers, and not plummet into population collapse, which is currently happening.

我确实想再次强调拥有足够的孩子来维持我们的人口数量的重要性，而不是陷入目前正在发生的人口崩溃。

Population collapse is a real and current thing.

人口崩溃是一个真实的、当前的事情。

1:25:51 So the only reason it’s not being reflected in the total population numbers as much is because people are living longer.

因此，它没有在总人口中反映出来的唯一原因是人们的寿命更长。

But it’s easy to predict, say what the population of any given country will be.

但预测任何特定国家的人口数量是很容易的。

Just take the birth rate last year, how many babies were born, multiply that by life expectancy and that’s what the population will be, steady state, if the birth rate continues to that level.

只需计算去年的出生率，即出生的婴儿数量，乘以预期寿命，这就是如果出生率继续保持在这一水平的话，稳定状态下的人口数量。

But if it keeps declining, it will be even less and eventually dwindle to nothing.

但如果继续下降，就会更少，最终化为乌有。

So I keep banging on the baby drum here, for a reason, because it has been the source of civilizational collapse over and over again throughout history.

所以我一直在这里敲小鼓，这是有原因的，因为它在历史上一次又一次地成为文明崩溃的根源。

And so why don’t we just not try to stave off that day?

那么我们为什么不尝试推迟这一天的到来呢？

Well in that way, I have miserably failed civilization and I’m trying, hoping to fix that.

就这样，我的文明彻底失败了，我正在努力，希望能够解决这个问题。

1:26:41 I would love to have many kids.

我很想有很多孩子。

Great.

伟大的。

1:26:49 Hope you do.

希望你能做到。

No time like the present.

没有比现在更好的时间了。

Yeah, I got to allocate more compute to the whole process, but apparently it’s not that difficult.

是的，我必须为整个过程分配更多的计算，但显然这并不那么困难。

No, it’s like unskilled labor.

不，这就像非熟练劳动力。

Well, one of the things you do for me, for the world, is to inspire us with what the future could be.

嗯，你为我、为世界所做的事情之一，就是激励我们了解未来。

1:27:06 And so some of the things we’ve talked about, some of the things you’re building, alleviating human suffering with Neuralink and expanding the capabilities of the human mind, trying to build a colony on Mars.

因此，我们讨论过的一些事情，你们正在建造的一些事情，通过 Neuralink 减轻人类痛苦并扩展人类思维的能力，试图在火星上建立一个殖民地。

So creating a backup for humanity on another planet and exploring the possibilities of what artificial intelligence could be in this world, especially in the real world, AI with hundreds of millions, maybe billions of robots walking around.

因此，在另一个星球上为人类创建一个备份，并探索人工智能在这个世界上的可能性，特别是在现实世界中，人工智能有数亿甚至数十亿机器人在周围行走。

There will be billions of robots.

将会有数十亿个机器人。

1:27:45 That seems virtual certainty.

这似乎是板上钉钉的事情。

Well, thank you for building the future and thank you for inspiring so many of us to keep building and creating cool stuff, including kids.

好吧，感谢您建设未来，感谢您激励我们很多人继续建设和创造很酷的东西，包括孩子们。

You’re welcome.

不客气。

1:28:00 Go forth and multiply.

继续前进并繁衍。

Go forth, multiply.

前进，繁衍。

1:28:04 Thank you Elon.

谢谢埃隆。

Thanks for talking about it.

谢谢你谈论它。

Thanks for listening to this conversation with Elon Musk.

感谢您收听与埃隆·马斯克的对话。

And now, dear friends, here’s DJ Seo, the Co-Founder, President and COO of Neuralink.

现在，亲爱的朋友们，这是 Neuralink 联合创始人、总裁兼首席运营官 DJ Seo。

When did you first become fascinated by the human brain?

您什么时候第一次对人脑着迷？

For me, I was always interested in understanding the purpose of things and how it was engineered to serve that purpose, whether it’s organic or inorganic, like we were talking earlier about your curtain holders.

对我来说，我总是对理解事物的目的以及它是如何设计来服务于该目的感兴趣，无论是有机的还是无机的，就像我们之前讨论的窗帘架一样。

1:28:23 They serve a clear purpose and they were engineered with that purpose in mind.

它们有明确的目的，并且在设计时就考虑到了这一目的。

And growing up I had a lot of interest in seeing things, touching things, feeling things, and trying to really understand the root of how it was designed to serve that purpose.

在成长过程中，我对观察事物、触摸事物、感受事物非常感兴趣，并试图真正理解它是如何设计来实现这一目的的根源。

And obviously brain is just a fascinating organ that we all carry.

显然，大脑是我们所有人都拥有的一个令人着迷的器官。

It’s an infinitely powerful machine that has intelligence and cognition that arise from it.

它是一台无限强大的机器，具有由此产生的智能和认知。

And we haven’t even scratched the surface in terms of how all of that occurs.

对于这一切是如何发生的，我们甚至还没有触及表面。

1:29:17 But also at the same time, I think it took me a while to make that connection to really studying and building tech to understand the brain.

但与此同时，我认为我花了一段时间才将这种联系与真正研究和构建理解大脑的技术联系起来。

Not until graduate school.

直到研究生毕业。

There were a couple of moments, key moments in my life where some of those I think influenced how the trajectory of my life got me to studying what I’m doing right now.

在我生命中的一些时刻，一些关键时刻，我认为其中一些影响了我的生活轨迹，让我开始研究我现在正在做的事情。

One was growing up, both sides of my family, my grandparents had a very severe form of Alzheimer and it’s incredibly debilitating conditions.

一个是在我的成长过程中，我的家人、我的祖父母都患有一种非常严重的阿尔茨海默病，这是一种令人难以置信的衰弱状况。

I mean, literally you’re seeing someone’s whole identity and their mind just losing over time.

我的意思是，从字面上看，你会看到某人的整个身份和他们的思想随着时间的推移而消失。

And I just remember thinking how both the power of the mind, but also how something like that could really lose your sense of identity.

我只记得我在想，心灵的力量，以及类似的东西如何真的会失去你的认同感。

It’s fascinating that that is one of the ways to reveal the power of a thing by watching it lose the power.

令人着迷的是，这是通过观察事物失去力量来揭示其力量的方法之一。

Yeah, a lot of what we know about the brain actually comes from these cases where there are trauma to the brain or some parts of the brain that led someone to lose certain abilities.

是的，我们对大脑的了解实际上来自这些案例，即大脑或大脑某些部分受到创伤，导致某人失去某些能力。

1:30:17 And as a result there’s some correlation and understanding of that part of the tissue being critical for that function.

因此，对于该功能至关重要的组织部分存在一些相关性和理解。

And it’s an incredibly fragile organ, if you think about it that way.

如果你这样想的话，它是一个极其脆弱的器官。

But also it’s incredibly plastic and incredibly resilient in many different ways.

而且它在许多方面都具有令人难以置信的塑料性和令人难以置信的弹性。

And by the way, the term plastic as we’ll use a bunch, means that it’s adaptable.

顺便说一下，我们会经常使用“塑料”这个术语，意思是它具有适应性。

1:30:46 So neuroplasticity refers to the adaptability of the human brain?

那么神经可塑性是指人脑的适应性吗？

Correct.

正确的。

1:30:56 Another key moment that sort of influenced how the trajectory of my life have shaped towards the current focus of my life has been during my teenage year when I came to the US.

另一个影响我的人生轨迹如何塑造我当前生活焦点的关键时刻是我十几岁的时候，当时我来到了美国。

I didn’t speak a word of English.

我一句英语也没说。

There was a huge language barrier and there was a lot of struggle to connect with my peers around me because I didn’t understand the artificial construct that we have created called language, specifically English in this case.

存在巨大的语言障碍，与周围的同龄人联系有很多困难，因为我不理解我们创建的称为语言的人工结构，特别是在这种情况下的英语。

And I remember feeling pretty isolated, not being able to connect with peers around me.

我记得我感到非常孤立，无法与周围的同龄人联系。

So spent a lot of time just on my own reading books, watching movies, and I naturally sort of gravitated towards sci-fi books.

所以我花了很多时间看书、看电影，我自然而然地被科幻小说所吸引。

I just found them really, really interesting.

我发现它们真的非常非常有趣。

And also it was a great way for me to learn English.

这对我来说也是学习英语的好方法。

1:31:46 Some of the first set of books that I picked up are Enders Game, the whole saga by Orson Scott Card and Neuromancer from William Gibson and Snow Crash from Neal Stephenson.

我买的第一套书是《Enders Game》，奥森·斯科特·卡德的整个传奇故事，威廉·吉布森的《神经漫游者》和尼尔·斯蒂芬森的《雪崩》。

And movies like Matrix, what’s coming out around that time point that really influenced how I think about the potential impact that technology can have for our lives in general.

像《黑客帝国》这样的电影，在那个时间点上映的电影真正影响了我对技术对我们生活的潜在影响的看法。

1:32:11 So fast track to my college years, I was always fascinated by just physical stuff, building physical stuff and especially physical things that had some sort of intelligence.

进入大学时代后，我总是着迷于物理事物、建造物理事物，尤其是具有某种智能的物理事物。

And I studied electrical engineering during undergrad and I started out my research in MEMS, so micro electromechanical systems and really building these tiny nano structures for temperature sensing.

我在本科期间学习了电气工程，并开始了 MEMS、微机电系统的研究，并真正构建了这些用于温度传感的微小纳米结构。

And I just found that to be just incredibly rewarding and fascinating subject to just understand how you can build something miniature like that, that again, serve a function and had a purpose.

我发现这是一个令人难以置信的回报和令人着迷的主题，只是了解如何构建这样的微型东西，再次，服务于功能并有一个目的。

Then I spent large majority of my college years basically building millimeter wave circuits for next gen telecommunication systems for imaging.

然后，我大学的大部分时间都花在为下一代成像电信系统构建毫米波电路上。

And it was just something that I found very, very intellectually interesting.

我发现这在智力上非常非常有趣。

Phase arrays, how the signal processing works for any modern as well as next gen telecommunication system, wireless and wire line, EM waves or electromagnetic waves are fascinating.

相控阵、信号处理如何适用于任何现代以及下一代电信系统、无线和有线、EM 波或电磁波都是令人着迷的。

1:33:17 How do you design antennas that are most efficient in a small footprint that you have?

如何在较小的占地面积内设计最高效的天线？

How do you make these things energy efficient?

如何使这些东西变得节能？

That was something that just consumed my intellectual curiosity and that journey led me to actually apply to and find myself at PhD program at UC Berkeley, at this consortium called the Berkeley Wireless Research Center that was precisely looking at building…

这消耗了我的求知欲，这段旅程让我真正申请并找到了自己在加州大学伯克利分校的博士项目，在这个名为伯克利无线研究中心的联盟中，该联盟正是致力于建立……

At the time, we called it XG, similar to 3G, 4G, 5G, but the next, next generation G system and how you would design circuits around that to ultimately go on phones and basically any other devices that are wirelessly connected these days.

当时，我们称之为 XG，类似于 3G、4G、5G，但它是下一代 G 系统，以及如何围绕该系统设计电路，最终应用在手机和基本上任何其他当今无线连接的设备上。

So I was just absolutely just fascinated by how that entire system works and that infrastructure works.

所以我对整个系统的工作原理和基础设施的工作原理非常着迷。

1:34:07 And then also during grad school, I had sort of the fortune of having a couple of research fellowships that led me to pursue whatever project that I want.

然后，在读研究生期间，我有幸获得了一些研究奖学金，这使我能够追求任何我想要的项目。

And that’s one of the things that I really enjoyed about my graduate school career, where you got to kind of pursue your intellectual curiosity in the domain that may not matter at the end of the day, but is something that really allows you the opportunity to go as deeply as you want, as well as widely as you want.

这是我在研究生生涯中真正喜欢的事情之一，在那里你必须在某个领域追求你的求知欲，虽然最终可能并不重要，但它确实让你有机会按照你想要的方式深入、广泛地探索。

And at the time I was actually working on this project called the Smart Bandaid, and the idea was that when you get a wound, there’s a lot of other proliferation of signaling pathway that cells follow to close that wound.

当时我实际上正在研究一个名为“智能创可贴”的项目，其想法是，当你受伤时，细胞会遵循许多其他信号通路的增殖来闭合伤口。

And there were hypotheses that when you apply external electric field, you can actually accelerate the closing of that field by having basically electro taxing of the cells around that wound site.

有人假设，当你施加外部电场时，你实际上可以通过对伤口部位周围的细胞进行基本的电刺激来加速该电场的关闭。

1:35:06 And specifically not just for a normal wound, there are chronic wounds that don’t heal.

特别是不仅对于正常伤口，还有无法愈合的慢性伤口。

So we were interested in building some sort of a wearable patch that you could apply to facilitate that healing process.

因此，我们有兴趣构建某种可穿戴贴片，您可以使用它来促进愈合过程。

And that was in collaboration with Professor Michel Maharbiz, which was a great addition to my thesis committee and it really shaped the rest of my PhD career.

那是与米歇尔·马哈比兹教授合作的，他对我的论文委员会来说是一个很好的补充，它确实塑造了我余下的博士生涯。

So this would be the first time you interacted with biology, I suppose?

所以我想这将是你第一次与生物学互动？

Correct.

正确的。

1:35:37 I mean there were some peripheral end application of the wireless imaging and telecommunication system that I was using for security and bio imaging.

我的意思是，我用于安全和生物成像的无线成像和电信系统有一些外围终端应用。

But this was a very clear direct application to biology and biological system and understanding the constraints around that and really designing and engineering electrical solutions around that.

但这是对生物学和生物系统的非常明确的直接应用，了解其周围的限制，并真正围绕其设计和工程电气解决方案。

So that was my first introduction and that’s also kind of how I got introduced to Michel.

这是我的第一次介绍，也是我认识米歇尔的方式。

He’s sort of known for remote control of beetles in the early two thousands.

他因远程控制两千出头的甲虫而闻名。

1:36:16 And then around 2013, obviously the holy grail when it comes to implantable system is to understand how small of a thing you can make, and a lot of that is driven by how much energy or how much power you can supply to it and how you extract data from it.

然后在 2013 年左右，显然，植入式系统的圣杯是了解你可以制造出多小的东西，而这很大程度上取决于你可以向它提供多少能量或多少功率以及如何从中提取数据。

At the time at Berkeley, there was this desire to understand in the neural space what sort of system you can build to really miniaturize these implantable systems.

当时在伯克利，人们希望了解在神经空间中可以构建什么样的系统来真正小型化这些可植入系统。

And I distinctively remember this one particular meeting where Michel came in and he’s like, “Guys, I think I have a solution.

我清楚地记得在一次特殊的会议上，米歇尔进来了，他说：“伙计们，我想我有一个解决方案。

The solution is ultrasound.” And then he proceeded to walk through why that is the case.

解决方案是超声波。”然后他继续解释为什么会出现这种情况。

And that really formed the basis for my thesis work called Neural dust system, that was looking at ways to use ultrasound as opposed to electromagnetic waves for powering as well as communication.

这确实构成了我名为“神经尘埃系统”的论文工作的基础，该论文正在研究使用超声波而不是电磁波来供电和通信的方法。

I guess I should step back and say the initial goal of the project was to build these tiny, about a size of a neuron, implantable system that can be parked next to a neuron, being able to record its state and being able to ping that back to the outside world for doing something useful.

我想我应该退后一步说，该项目的最初目标是构建这些微小的、大约神经元大小的可植入系统，可以停在神经元旁边，能够记录其状态，并能够将其返回给外界以做一些有用的事情。

And as I mentioned, the size of the implantable system is limited by how you power the thing and get the data off of it.

正如我所提到的，植入式系统的大小受到其供电方式和从中获取数据的方式的限制。

And at the end of the day, fundamentally, if you look at a human body, we’re essentially bag of salt water with some interesting proteins and chemicals, but its mostly salt water that’s very, very well temperature regulated at 37 degrees Celsius.

归根结底，如果你观察人体，我们本质上就是一袋盐水，里面含有一些有趣的蛋白质和化学物质，但其中大部分是盐水，温度非常非常好地控制在 37 摄氏度。

1:38:05 And we’ll get into how, and later why that’s an extremely harsh environment for any electronics to survive.

我们将探讨如何以及为什么对于任何电子产品来说，这是一个极其恶劣的环境。

As I’m sure you’ve experienced or maybe not experienced, dropping cell phone in a salt water in an ocean, it will instantly kill the device.

我确信您经历过或可能没有经历过，将手机掉入海洋的盐水中，设备会立即损坏。

But anyways, just in general, electromagnetic waves don’t penetrate through this environment well and just the speed of light, it is what it is, we can’t change it.

但无论如何，一般来说，电磁波不能很好地穿透这个环境，光速就是这样，我们无法改变它。

And based on the wavelength at which you are interfacing with the device, the device just needs to be big.

根据与设备连接的波长，设备只需要很大即可。

These inductors needs to be quite big.

这些电感器需要相当大。

And the general good rule of thumb is that you want the wavefront to be roughly on the order of the size of the thing that you’re interfacing with.

一般好的经验法则是，您希望波前大致与您所连接的物体的尺寸相同。

So an implantable system that is around 10 to a hundred micron in dimension in a volume, which is about the size of a neuron that you see in a human body, you would have to operate at hundreds of gigahertz.

因此，一个体积约为 10 至 100 微米的可植入系统（大约相当于人体神经元的大小）必须在数百千兆赫的频率下运行。

Which number one, not only is it difficult to build electronics operating at those frequencies, but also the body just attenuates to that very, very significantly.

第一，不仅很难构建在这些频率下运行的电子设备，而且身体也会非常非常显着地衰减。

1:39:23 So the interesting kind of insight of this ultrasound was the fact that ultrasound just travels a lot more effectively in the human body tissue compared to electromagnetic waves.

因此，这种超声波的有趣之处在于，与电磁波相比，超声波在人体组织中传播的效率要高得多。

And this is something that you encounter, and I’m sure most people have encountered in their lives when you go to hospitals that are medical ultrasound sonograph.

这是你会遇到的事情，我相信大多数人一生中都遇到过，当你去有医用超声检查的医院时。

And they go into very, very deep depth without attenuating too much, too much of the signal.

它们进入非常非常深的深度，而不会衰减太多太多的信号。

So all in all, ultrasound, the fact that it travels through the body extremely well and the mechanism to which it travels to the body really well is that just the wavefront is very different.

总而言之，超声波在人体中传播得非常好，而且它传播到身体的机制非常好，只是波前非常不同。

Electromagnetic waves are transverse, whereas in ultrasound waves are compressive.

电磁波是横向的，而超声波是压缩的。

It’s just a completely different mode of wavefront propagation.

这只是一种完全不同的波前传播模式。

And as well as, speed of sound is orders and orders of magnitude less than speed of light, which means that even at 10 megahertz ultrasound wave, your wavefront ultimately is a very, very small wavelength.

而且，声速比光速低几个数量级，这意味着即使在 10 兆赫兹的超声波下，波前最终也是一个非常非常小的波长。

1:40:37 So if you’re talking about interfacing with the 10 micron or a hundred micron type structure, you would have 150 micron wavefront at 10 megahertz.

因此，如果您谈论的是与 10 微米或 100 微米类型结构的接口，那么您将在 10 兆赫兹下拥有 150 微米波前。

And building electronics at those frequencies are much, much easier and they’re a lot more efficient.

在这些频率下构建电子设备要容易得多，而且效率也高得多。

So the basic idea was born out of using ultrasound as a mechanism for powering the device and then also getting data back.

因此，基本想法诞生于使用超声波作为设备供电并获取数据的机制。

So now the question is how do you get the data back?

那么现在的问题是如何取回数据？

The mechanism to which we landed on is what’s called backscattering.

我们发现的机制就是所谓的反向散射。

This is actually something that is very common and that we interface on a day-to-day basis with our RFID cards, radio frequency ID tags.

这实际上是很常见的事情，我们每天都会与 RFID 卡、射频 ID 标签进行交互。

Where there’s actually rarely in your ID a battery inside, there’s an antenna and there’s some sort of coil that has your serial identification ID, and then there’s an external device called the reader that then sends a wavefront and then you reflect back that wavefront with some sort of modulation that’s unique to your ID.

实际上，您的身份证件中很少有电池，但有一个天线和某种带有您的序列识别 ID 的线圈，然后有一个称为读取器的外部设备，它会发送一个波前，然后您使用您的身份证件特有的某种调制反射回该波前。

That’s what’s called backscattering fundamentally.

从根本上来说，这就是所谓的后向散射。

1:41:50 So the tag itself actually doesn’t have to consume that much energy.

所以标签本身实际上不需要消耗那么多能量。

That was the mechanism through which we were thinking about sending the data back.

这就是我们考虑发回数据的机制。

When you have an external ultrasonic transducer that’s sending ultrasonic wave to your implant, the neural dust implant, and it records some information about its environment, whether it’s a neuron firing or some other state of the tissue that it’s interfacing with.

当您有一个外部超声波换能器向您的植入物（神经尘埃植入物）发送超声波时，它会记录有关其环境的一些信息，无论是神经元放电还是与之接触的组织的其他状态。

And then it just amplitude modulates the wavefront that comes back to the source.

然后它只是对返回源的波前进行幅度调制。

And the recording step would be the only one that requires any energy.

记录步骤将是唯一需要能量的步骤。

1:42:27 So what would require energy in that low step?

那么在这么低的步骤中需要什么能量呢？

Correct.

正确的。

1:42:33 So it is that initial startup circuitry to get that recording, amplifying it, and then just modulating.

因此，最初的启动电路用于获取录音、放大录音，然后进行调制。

And the mechanism to which that you can enable that is there is this specialized crystal called piezoelectric crystals that are able to convert sound energy into electrical energy and vice versa.

实现这一点的机制是一种称为压电晶体的特殊晶体，它能够将声能转换为电能，反之亦然。

So you can kind of have this interplay between the ultrasonic domain and electrical domain that is the biological tissue.

因此，超声波域和生物组织的电域之间可以产生这种相互作用。

So on the theme of parking very small computational devices next to neurons, that’s the dream, the vision of brain computer interfaces.

因此，关于将非常小的计算设备放置在神经元旁边的主题，这就是脑机接口的梦想和愿景。

1:43:04 Maybe before we talk about Neuralink, can you give a sense of the history of the field of BCI?

也许在我们谈论 Neuralink 之前，您能先介绍一下 BCI 领域的历史吗？

What has been maybe the continued dream and also some of the milestones along the way of the different approaches and the amazing work done at the various labs?

也许持续的梦想是什么，以及不同方法和各个实验室所做的令人惊叹的工作过程中的一些里程碑？

I think a good starting point is going back to 1790s.

我认为一个好的起点是 1790 年代。

I did not expect that.

我没想到会这样。

Where the concept of animal electricity or the fact that body’s electric was first discovered by Luigi Galvani, where he had this famous experiment where he connected set of electrodes to a frog leg and ran current through it, and then it started twitching and he said, “Oh my goodness, body’s electric.” So fast forward many, many years to 1920s where Hans Berger, who’s a German psychiatrist, discovered EEG or electroencephalography, which is still around.

动物电的概念或身体电的事实是由路易吉·加尔瓦尼（Luigi Galvani）首先发现的，他做了一个著名的实验，他将一组电极连接到青蛙腿上，并通过它流过电流，然后青蛙开始抽搐，他说：“哦，天哪，身体带电了。”时间快进很多很多年，到了 1920 年代，德国精神病学家汉斯·伯杰 (Hans Berger) 发现了脑电图 (EEG) 或脑电图描记术，这种技术至今仍然存在。

1:43:41 There are these electrode arrays that you wear outside the skull that gives you some sort of neural recording.

您将这些电极阵列戴在颅骨外，为您提供某种神经记录。

That was a very, very big milestone that you can record some sort of activities about the human mind.

这是一个非常非常大的里程碑，你可以记录一些关于人类思维的活动。

And then in the 1940s there were these group of scientists, Renshaw, Forbes and Morison that inserted these glass micro electrodes into the cortex and recorded single neurons.

然后在 20 世纪 40 年代，Renshaw、Forbes 和 Morison 等科学家小组将这些玻璃微电极插入皮层并记录单个神经元。

The fact that there’s signal that are a bit more high resolution and high fidelity as you get closer to the source, let’s say.

事实上，当您靠近信号源时，信号的分辨率和保真度会更高一些。

And in the 1950s, these two scientists, Hodgkin and Huxley showed up-

在 20 世纪 50 年代，霍奇金和赫胥黎这两位科学家出现了——

1:45:00 These two scientists, Hodgkin and Huxley showed up and they built this beautiful, beautiful models of the cell membrane and the ionic mechanism, and had these circuit diagram.

霍奇金和赫胥黎这两位科学家出现了，他们建立了这个非常漂亮的细胞膜和离子机制模型，并有这些电路图。

And as someone who’s an electrical engineer, it’s a beautiful model that’s built out of these partial differential equations, talking about flow of ions and how that really leads to how neurons communicate.

作为一名电气工程师，这是一个由这些偏微分方程构建的美丽模型，讨论离子流以及它如何真正导致神经元如何通信。

And they won the Nobel Prize for that 10 years later in the 1960s.

10 年后的 20 世纪 60 年代，他们因此获得了诺贝尔奖。

1:45:29 So in 1969, Eb Fetz from University of Washington published this beautiful paper called Operant Conditioning of Cortical Unit Activity, where he was able to record a single unit neuron from a monkey and was able to have the monkey modulated based on its activity and reward system.

因此，1969 年，华盛顿大学的 Eb Fetz 发表了这篇名为《皮质单元活动的操作调节》的精彩论文，其中他能够记录猴子的单个单元神经元，并能够根据猴子的活动和奖励系统来调节猴子。

So I would say this is the very, very first example, as far as I’m aware, of close loop brain computer interface or BCI.

所以我想说，据我所知，这是闭环脑机接口（BCI）的第一个例子。

The abstract reads, “The activity of single neurons in precentral cortex of unanesthetized monkeys was conditioned by reinforcing high rates of neuronal discharge with delivery of a food pellet.

摘要中写道：“未麻醉猴子中央前皮层中单个神经元的活动是通过食物颗粒的输送来增强神经元放电率来调节的。

1:46:01 Auditory or visual feedback of unit firing rates was usually provided in addition to food reinforcement.” Cool.

除了食物强化之外，通常还会提供单位射击速度的听觉或视觉反馈。”凉爽的。

So they actually got it done.

所以他们实际上做到了。

They got it done.

他们完成了。

1:46:24 This is back in 1969.

这要追溯到1969年。

” After several training sessions, monkeys could increase the activity of newly isolated cells by 50 to 500% above rates before reinforcement.” Fascinating.

“经过几次训练后，猴子可以将新分离的细胞的活性比强化前提高 50% 到 500%。”迷人。

Brain is very [inaudible 01:46:45].

大脑非常[听不清 01:46:45]。

And so from here, the number of experiments grew.

从这里开始，实验的数量开始增加。

Yeah.

是的。

1:46:49 Number of experiments, as well as set of tools to interface with the brain have just exploded.

实验的数量以及与大脑接口的工具集都呈爆炸式增长。

And also, just understanding the neural code and how some of the cortical layers and the functions are organized.

而且，只需了解神经代码以及一些皮质层和功能的组织方式即可。

So the other paper that is pretty seminal, especially in the motor decoding, was this paper in the 1980s from Georgopoulos that discovered that there’s this thing called motor tuning curve.

因此，另一篇非常具有开创性的论文，尤其是在电机解码方面，是 Georgopoulos 在 20 世纪 80 年代发表的论文，它发现了一种叫做电机调谐曲线的东西。

So what are motor tuning curves?

那么什么是电机调谐曲线呢？

It’s the fact that there are neurons in the motor cortex of mammals, including humans, that have a preferential direction that causes them to fire.

事实上，哺乳动物（包括人类）的运动皮层中的神经元具有导致它们放电的优先方向。

So what that means is, there are a set of neurons that would increase their spiking activities when you’re thinking about moving to the left, right, up, down, and any of those vectors.

所以这意味着，当你考虑向左、向右、向上、向下以及任何这些向量移动时，有一组神经元会增加它们的尖峰活动。

And based on that, you could start to think, well, if you can’t identify those essential eigenvectors, you can do a lot.

基于此，你可以开始思考，好吧，如果你不能识别那些基本的特征向量，你可以做很多事情。

And you can actually use that information for actually decoding someone’s intended movement from the cortex.

你实际上可以使用这些信息来从皮层解码某人的预期运动。

So that was a very, very seminal paper that showed that there is some sort of code that you can extract, especially in the motor cortex.

所以这是一篇非常非常具有开创性的论文，它表明你可以提取某种代码，特别是在运动皮层中。

So there’s signal there.

所以那里有信号。

1:48:11 And if you measure the electrical signal from the brain that you could actually figure out what the intention was.

如果你测量大脑的电信号，你实际上可以弄清楚意图是什么。

Correct.

正确的。

1:48:20 Yeah, not only electrical signals, but electrical signals from the right set of neurons that give you these preferential direction.

是的，不仅是电信号，还有来自正确神经元组的电信号，它们可以为您提供这些优先方向。

Okay.

好的。

1:48:29 So going slowly towards Neuralink, one interesting question is, what do we understand on the BCI front, on invasive versus non-invasive, from this line of work?

因此，慢慢走向 Neuralink，一个有趣的问题是，我们对 BCI 前沿、侵入性与非侵入性这一工作的理解是什么？

How important is it to park next to the neuron?

停在神经元旁边有多重要？

What does that get you?

这会给你带来什么？

That answer fundamentally depends on what you want to do with it.

这个答案从根本上取决于你想用它做什么。

1:48:49 There’s actually incredible amount of stuff that you can do with EEG and electrocortical graph, ECOG, which actually doesn’t penetrate the cortical layer or parenchyma, but you place a set of electrodes on the surface of the brain.

实际上，你可以使用脑电图和皮层电图（ECOG）做大量的事情，它实际上不会穿透皮质层或实质，但你可以在大脑表面放置一组电极。

So the thing that I’m personally very interested in is just actually understanding and being able to just really tap into the high resolution, high fidelity, understanding of the activities that are happening at the local level.

因此，我个人非常感兴趣的是真正理解并能够真正利用高分辨率、高保真度来理解当地正在发生的活动。

And we can get into biophysics, but just to step back to use analogy, because analogy here can be useful, and sometimes it’s a little bit difficult to think about electricity.

我们可以进入生物物理学，但只是退一步使用类比，因为类比在这里可能有用，有时思考电有点困难。

At the end of the day, we’re doing electrical recording that’s mediated by ionic currents, movements of these charged particles, which is really, really hard for most people to think about.

归根结底，我们正在进行的是由离子电流、这些带电粒子的运动介导的电记录，这对大多数人来说真的很难想象。

1:49:45 But turns out, a lot of the activities that are happening in the brain and the frequency bandwidth with which that’s happening, is actually very, very similar to sound waves and our normal conversation audible range.

但事实证明，大脑中发生的许多活动以及发生这些活动的频率带宽实际上与声波和我们正常对话的可听范围非常非常相似。

So the analogy that typically is used in the field is, if you have a football stadium, there’s a game going on.

因此，在该领域通常使用的类比是，如果你有一个足球场，那就有一场比赛正在进行。

If you stand outside the stadium, you maybe get a sense of how the game is going based on the cheers and the boos of the home crowd, whether the team is winning or not.

如果你站在体育场外，你也许可以根据主场球迷的欢呼声和嘘声来了解比赛的进展情况，无论球队是否获胜。

But you have absolutely no idea what the score is, you have absolutely no idea what individual audience or the players are talking or saying to each other, what the next play is, what the next goal is.

但你完全不知道比分是多少，你完全不知道个别观众或球员在互相交谈或说些什么，下一场比赛是什么，下一个目标是什么。

So what you have to do is you have to drop the microphone into the stadium and then get near the source into the individual chatter.

所以你要做的就是把麦克风扔进体育场，然后靠近声源进入个人的谈话声。

In this specific example, you would want to have it right next to where the huddle is happening.

在这个特定的示例中，您可能希望将其放置在拥挤发生的地方旁边。

1:50:47 So I think that’s kind of a good illustration of what we’re trying to do when we say invasive or minimally invasive or implanted brain computer interfaces versus non-invasive or non-implanted brain interfaces.

所以我认为，当我们说侵入性或微创或植入式脑机接口与非侵入性或非植入式脑机接口时，这很好地说明了我们正在努力做的事情。

It’s basically talking about where do you put that microphone and what can you do with that information.

它基本上是在讨论你把麦克风放在哪里以及你可以用这些信息做什么。

So what is the biophysics of the read and write communication that we’re talking about here as we now step into the efforts at Neuralink?

那么，当我们现在开始 Neuralink 的工作时，我们在这里讨论的读写通信的生物物理学是什么？

Yeah.

是的。

1:51:18 So brain is made up of these specialized cells called neurons.

因此，大脑是由这些称为神经元的特殊细胞组成的。

There’s billions of them, tens of billions, sometimes people call it a hundred billion, that are connected in this complex yet dynamic network that are constantly remodeling.

有数十亿、数百亿，有时人们称之为千亿，它们连接在这个不断重塑的复杂而动态的网络中。

They’re changing their synaptic weights, and that’s what we typically call neuroplasticity.

它们正在改变突触权重，这就是我们通常所说的神经可塑性。

And the neurons are also bathed in this charged environment that is latent with many charge molecules like potassium ions, sodium ions, chlorine ions.

神经元也沐浴在这种带电环境中，潜藏着许多带电分子，如钾离子、钠离子、氯离子。

And those actually facilitate these, through ionic current, communication between these different networks.

这些实际上通过离子电流促进了这些不同网络之间的通信。

1:52:08 And when you look at a neuron as well, they have these membrane with a beautiful, beautiful protein structure called the voltage selective ion channels, which in my opinion, is one of nature’s best inventions.

当你观察神经元时，它们的膜具有美丽的蛋白质结构，称为电压选择性离子通道，在我看来，这是大自然最好的发明之一。

In many ways, if you think about what they are, they’re doing the job of a modern day transistors.

从很多方面来说，如果你考虑一下它们是什么，就会发现它们正在做现代晶体管的工作。

Transistors are nothing more, at the end of the day, than a voltage-gated conduction channel.

归根结底，晶体管只不过是一个电压门控传导通道。

And nature found a way to have that very, very early on in its evolution.

大自然在其进化的非常非常早期就找到了一种方法来实现这一点。

And as we all know, with the transistor, you can have many, many computation and a lot of amazing things that we have access to today.

众所周知，通过晶体管，你可以进行很多很多的计算，以及我们今天可以获得的很多令人惊奇的东西。

So I think it’s one of those, just as a tangent, just a beautiful, beautiful invention that the nature came up with, these voltage-gated ion channels.

所以我认为这是其中之一，就像切线一样，只是大自然想出的一个美丽的发明，这些电压门控离子通道。

I suppose there’s, on the biological of it, every level of the complexity, of the hierarchy, of the organism, there’s going to be some mechanisms for storing information and for doing computation.

我想，从生物学的角度来看，有机体的各个层次的复杂性、层次结构，都会有一些存储信息和进行计算的机制。

1:53:02 And this is just one such way.

这只是其中一种方式。

But to do that with biological and chemical components is interesting.

但用生物和化学成分来做到这一点很有趣。

Plus, when neurons, it’s not just electricity, it’s chemical communication, it’s also mechanical.

另外，对于神经元来说，它不仅仅是电，它是化学通讯，它也是机械通讯。

These are actual objects that vibrate, they move.

这些是振动、移动的实际物体。

It’s all of that.

这就是全部。

Yeah, actually there’s a lot of really, really interesting physics that are involved in kind of going back to my work on ultrasound during grad school, there were groups and there are still groups looking at ways to cause neurons to actually fire an action potential using ultrasound wave.

是的，实际上有很多非常非常有趣的物理学涉及到我在研究生院期间关于超声波的工作，有一些小组并且仍然有一些小组正在研究如何使用超声波使神经元真正激发动作电位。

1:53:36 And the mechanism to which that’s happening is still unclear, as I understand.

据我了解，发生这种情况的机制仍不清楚。

It may just be that you’re imparting some sort of thermal energy and that causes cells to depolarize in some interesting ways.

可能只是你传递了某种热能，导致细胞以一些有趣的方式去极化。

But there are also these ion channels, or even membranes, that actually just open up as pore as they’re being mechanically shook, vibrated.

但也有这些离子通道，甚至是膜，它们实际上只是在受到机械摇动、振动时像毛孔一样打开。

There’s just a lot of elements of these, move particles, which again, that’s governed by diffusion physics, movements of particles.

其中有很多元素，移动粒子，这又是由扩散物理学和粒子运动控制的。

And there’s also a lot of interesting physics there.

那里还有很多有趣的物理学。

Also, not to mention, as Roger Penrose talks about, there might be some beautiful weirdness in the quantum mechanical effects of all of this.

另外，更不用说，正如罗杰·彭罗斯所说，所有这一切的量子力学效应可能存在一些美丽的怪异。

Oh, yeah.

哦，是的。

And he actually believes that consciousness might emerge from the quantum mechanical effects there.

他实际上相信意识可能是从那里的量子力学效应中产生的。

1:54:44 So there’s physics, there’s chemistry, there’s biology, all of that is going on there.

所以有物理学、有化学、有生物学，所有这些都在那里发生。

Oh, yeah.

哦，是的。

1:54:54 Yes, there’s a lot of levels of physics that you can dive into.

是的，您可以深入研究很多级别的物理知识。

But yeah, in the end, you have these membranes with these voltage-gated ion channels that selectively let these charged molecules that are in the extracellular matrix, in and out.

但是，是的，最终，这些膜带有这些电压门控离子通道，可以选择性地让细胞外基质中的这些带电分子进出。

And these neurons generally have these resting potential where there’s a voltage difference between inside the cell and outside the cell.

这些神经元通常具有静息电位，其中细胞内部和细胞外部之间存在电压差。

And when there’s some sort of stimuli that changes the state such that they need to send information to the downstream network, you start to see these orchestration of these different molecules going in and out of these channels.

当某种刺激改变状态从而需要将信息发送到下游网络时，您开始看到这些不同分子进出这些通道的编排。

They also open up.

他们也开放。

More of them open up once it reaches some threshold, to a point where you have a depolarizing cell that sends an action potential.

一旦达到某个阈值，就会有更多的细胞打开，达到一个可以发送动作电位的去极化细胞的程度。

So it’s just a very beautiful kind of orchestration of these molecules.

所以这只是这些分子的一种非常美丽的编排。

And what we’re trying to do when we place an electrode or parking it next to a neuron is that you’re trying to measure these local changes in the potential.

当我们放置电极或将其停放在神经元旁边时，我们试图做的是测量这些局部电位变化。

Again, mediated by the movements of the ions.

再次，由离子的运动介导。

1:56:17 And what’s interesting, as I mentioned earlier, there’s a lot of physics involved.

有趣的是，正如我之前提到的，其中涉及很多物理知识。

And the two dominant physics for this electrical recording domain is diffusion physics and electromagnetism.

该电记录领域的两个主导物理学是扩散物理学和电磁学。

And where one dominates, where Maxwell’s equation dominates versus Fick’s law dominates depends on where your electrode is.

麦克斯韦方程占主导地位与菲克定律占主导地位取决于你的电极在哪里。

If it’s close to the source, mostly electromagnetic-based.

如果它靠近源头，则主要是基于电磁的。

When you’re further away from it, it’s more diffusion-based.

当你离它更远时，它就更加基于扩散。

So essentially, when you’re able to park it next to it, you can listen in on those individual chatter and those local changes in the potential.

所以本质上，当你能够把它停在它旁边时，你可以倾听那些个人的闲聊和潜力的局部变化。

And the type of signal that you get are these canonical textbook neural spiking waveform.

您获得的信号类型是这些规范教科书的神经尖峰波形。

The moment you’re further away, and based on some of the studies that people have done, Christof Koch’s lab, and others, once you’re away from that source by roughly around a hundred micron, which is about a width of a human hair, you no longer hear from that neuron.

当你距离更远时，根据人们所做的一些研究，克里斯托夫·科赫的实验室和其他人，一旦你远离那个源大约一百微米，大约是人类头发的宽度，你就不再听到那个神经元的声音。

You’re no longer able to have the system sensitive enough to be able to record that particular local membrane potential change in that neuron.

您不再能够让系统足够灵敏来记录该神经元中特定的局部膜电位变化。

1:57:36 And just to give you a sense of scale also, when you look at a hundred micron voxel, so a hundred micron by a hundred micron by a hundred micron box in a brain tissue, there’s roughly around 40 neurons, and whatever number of connections that they have.

为了给你一种规模感，当你观察 100 微米的体素时，即脑组织中 100 微米×100 微米×100 微米的盒子，大约有 40 个神经元，以及它们具有的连接数量。

So there’s a lot in that volume of tissue.

所以这个组织体积里有很多东西。

So the moment you’re outside of that, there’s just no hope that you’ll be able to detect that change from that one specific neuron that you may care about.

因此，一旦你脱离了这种状态，你就不可能从你可能关心的某个特定神经元中检测到这种变化。

But as you’re moving about this space, you’ll be hearing other ones.

但当您在这个空间中移动时，您会听到其他声音。

1:58:03 So if you move another a hundred micron, you’ll be hearing chatter from another community.

因此，如果你再移动一百微米，你就会听到来自另一个社区的喋喋不休。

Correct.

正确的。

And so the whole sense is, you want to place as many as possible electrodes, and then you’re listening to the chatter.

所以整体的感觉是，你想要放置尽可能多的电极，然后你就可以听到喋喋不休的声音。

Yeah, you want to listen to the chatter.

是的，你想听闲聊。

1:58:20 And at the end of the day, you also want to basically let the software do the job of decoding.

归根结底，您还想让软件完成解码工作。

And just to kind of go to why ECOG and EEG work at all.

只是想了解一下 ECOG 和 EEG 为何有效。

When you have these local changes, obviously it’s not just this one neuron that’s activating, there’s many, many other networks that are activating all the time.

当你发生这些局部变化时，显然不仅仅是这个神经元在激活，还有很多很多其他网络一直在激活。

And you do see sort of a general change in the potential of this electrode, this charged medium, and that’s what you’re recording when you’re farther away.

你确实看到了这个电极、这个带电介质的电势的总体变化，这就是你在更远的地方记录到的。

I mean, you still have some reference electrode that’s stable in the brain, that’s just electro- active organ, and you’re seeing some combination, aggregate action, potential changes, and then you can pick it up.

我的意思是，你的大脑中仍然有一些稳定的参考电极，那只是电活性器官，你会看到一些组合、聚合作用、潜在的变化，然后你就可以拿起它。

It’s a much slower changing signals.

这是一个变化缓慢得多的信号。

But there are these canonical oscillations and waves like gamma waves, beta waves, when you sleep, that can be detected because there’s sort of a synchronized global effect of the brain that you can detect.

但是，当你睡觉时，可以检测到这些典型的振荡和波，例如伽马波、β波，因为你可以检测到大脑的同步全局效应。

And the physics of this go, if we really want to go down that rabbit hole, there’s a lot that goes on in terms of why diffusion physics at some point dominates when you’re further away from the source.

而这个过程的物理学，如果我们真的想深入那个兔子洞，那么当你远离源头时，为什么扩散物理学在某些时候占主导地位，有很多事情会发生。

It is just a charged medium.

它只是一种带电介质。

So similar to how when you have electromagnetic waves propagating in atmosphere or in a charged medium like a plasma, there’s this weird shielding that happens that actually further attenuates the signal as you move away from it.

与电磁波在大气或等离子体等带电介质中传播时类似，会发生这种奇怪的屏蔽，当您远离信号时，信号实际上会进一步衰减。

So yeah, you see, if you do a really, really deep dive on the signal attenuation over distance, you start to see one over R square in the beginning and then exponential drop off, and that’s the knee at which you go from electromagnetism dominating to diffusion physic dominating.

所以，是的，你看，如果你对信号随距离的衰减进行非常非常深入的研究，你会开始看到一开始超过 R 平方，然后呈指数下降，这就是你从电磁主导转向扩散物理主导的拐点。

But once again, with the electrodes, the biophysics that you need to understand is not as deep because no matter where you’re placing it, you’re listening to a small crowd of local neurons.

但再一次，对于电极，你需要了解的生物物理学并不那么深，因为无论你把它放在哪里，你都会听到一小群局部神经元的声音。

Correct, yeah.

正确，是的。

2:00:32 So once you penetrate the brain, you’re in the arena, so to speak.

所以，可以这么说，一旦你穿透了大脑，你就进入了竞技场。

And there’s a lot of neurons.

而且有很多神经元。

There are many, many of them.

有很多很多。

But then again, there’s a whole field of neuroscience that’s studying how the different groupings, the different sections of the seating in the arena, what they usually are responsible for, which is where the metaphor probably falls apart because the seating is not that organized in an arena.

但话又说回来，整个神经科学领域都在研究竞技场中座位的不同分组、不同部分，以及它们通常负责什么，这就是这个比喻可能不成立的地方，因为竞技场中的座位并不是那么有组织。

Also, most of them are silent.

而且，大多数人都是沉默的。

2:00:56 They don’t really do much.

他们实际上并没有做太多事情。

Or their activities are…

或者他们的活动是……

You have to hit it with just the right set of stimulus.

你必须用正确的刺激来达到目的。

So they’re usually quiet.

所以他们通常都很安静。

They’re usually very quiet.

他们通常很安静。

2:01:09 Similar to dark energy and dark matter, there’s dark neurons.

与暗能量和暗物质类似，也有暗神经元。

What are they all doing?

他们都在做什么？

When you place these electrodes, again, within this hundred micron volume, you have 40 or so neurons.

当你再次将这些电极放置在这百微米的体积内时，你就会有 40 个左右的神经元。

Why do you not see 40 neurons?

为什么你看不到 40 个神经元？

Why do you see only a handful?

为什么你只看到少数？

What is happening there?

那里发生了什么？

Well, they’re mostly quiet, but when they speak, they say profound shit.

嗯，他们大多都很安静，但当他们说话时，他们会说一些深刻的废话。

2:01:25 That’s the way I’d like to think about it.

这就是我想考虑的方式。

Anyway, before we zoom in even more, let’s zoom out.

不管怎样，在我们进一步放大之前，让我们先缩小一下。

So how does Neuralink work from the surgery to the implant, to the signal and the decoding process, and the human being able to use the implant to actually affect the world outside?

那么 Neuralink 从手术到植入，再到信号和解码过程是如何工作的，以及人类如何能够使用植入物来实际影响外面的世界呢？

And all of this, I’m asking in the context of, there’s a gigantic historic milestone that Neuralink just accomplished in January of this year.

所有这一切，我问的是，Neuralink 在今年 1 月刚刚实现了一个巨大的历史性里程碑。

Putting a Neuralink implant in the first human being, Noland.

将 Neuralink 植入物植入第一个人诺兰德体内。

And there’s been a lot to talk about there about his experience because he’s able to describe all the nuance and the beauty and the fascinating complexity of that experience of everything involved.

关于他的经历有很多值得谈论的地方，因为他能够描述所涉及的一切经历的所有细微差别、美丽和令人着迷的复杂性。

But on the technical level, how does Neuralink work?

但从技术层面来说，Neuralink 是如何工作的呢？

So there are three major components to the technology that we’re building.

因此，我们正在构建的技术由三个主要组成部分组成。

2:02:26 One is the device, the thing that’s actually recording these neural chatters.

一是设备，它实际上记录这些神经的喋喋不休。

We call it N1 Implant or The Link.

我们称之为 N1 植入物或 The Link。

And we have a surgical robot that’s actually doing an implantation of these tiny, tiny wires that we call threads that are smaller than human hair.

我们有一个外科手术机器人，它实际上正在植入这些非常非常小的电线，我们称之为比人类头发还小的线。

And once everything is surgerized, you have these neural signals, these spiking neurons, that are coming out of the brain, and you need to have some sort of software to decode what the users intend to do with that.

一旦一切都经过手术，你就会得到这些从大脑发出的神经信号，这些尖峰神经元，你需要某种软件来解码用户打算用它做什么。

So there’s what’s called the Neuralink Application or B1 App that’s doing that translation.

因此，有一种所谓的 Neuralink 应用程序或 B1 应用程序正在执行这种翻译。

It’s running the very, very simple machine learning model that decodes these inputs that are neural signals and then convert it to a set of outputs that allows our first participant, Noland, to be able to control a cursor on the screen.

它运行非常非常简单的机器学习模型，对这些神经信号输入进行解码，然后将其转换为一组输出，使我们的第一个参与者诺兰能够控制屏幕上的光标。

And this is done wirelessly?

这是无线完成的吗？

And this is done wirelessly.

这是通过无线方式完成的。

2:03:33 So our implant is actually a two-part.

所以我们的植入物实际上是由两部分组成。

The link has these flexible tiny wires called threads that have multiple electrodes along its length.

该链接有这些称为线的柔性细线，沿其长度有多个电极。

And they’re only inserted into the cortical layer, which is about three to five millimeters in a human brain, in the motor cortex region.

它们只插入到运动皮层区域的皮层中，皮层在人脑中大约三到五毫米。

That’s where the intention for movement lies in.

这就是运动的意图所在。

And we have 64 of these threads, each thread having 16 electrodes along the span of three to four millimeters, separated by 200 microns.

我们有 64 根这样的线，每根线有 16 个电极，跨度为 3 到 4 毫米，间距为 200 微米。

So you can actually record along the depth of the insertion.

所以你实际上可以沿着插入的深度进行记录。

And based on that signal, there’s custom integrated circuit or ASIC that we built that amplifies the neural signals that you’re recording and then digitizing it and then has some mechanism for detecting whether there was an interesting event that is a spiking event, and decide to send that or not send that through Bluetooth to an external device, whether it’s a phone or a computer that’s running this Neuralink application.

基于该信号，我们构建了定制集成电路或 ASIC，放大您正在记录的神经信号，然后将其数字化，然后具有某种机制来检测是否存在有趣的事件，即尖峰事件，并决定是否通过蓝牙将其发送到外部设备，无论是运行此 Neuralink 应用程序的手机还是计算机。

So there’s onboard signal processing already just to decide whether this is an interesting event or not.

因此，机载信号处理已经可以决定这是否是一个有趣的事件。

2:04:50 So there is some computational power on board in addition to the human brain?

那么除了人脑之外，船上还有一些计算能力吗？

Yeah.

是的。

2:05:00 So it does the signal processing to really compress the amount of signal that you’re recording.

因此，它会进行信号处理，以真正压缩您正在记录的信号量。

So we have a total of thousand electrodes sampling at just under 20 kilohertz with 10 bit each.

因此，我们总共有 1000 个电极，采样频率略低于 20 kHz，每个电极 10 位。

So that’s 200 megabits that’s coming through to the chip from thousand channel simultaneous neural recording.

因此，这是从千通道同步神经记录传输到芯片的 200 兆比特。

And that’s quite a bit of data, and there are technology available to send that off wirelessly.

这是相当多的数据，并且有技术可以无线发送这些数据。

But being able to do that in a very, very thermally-constrained environment that is a brain.

但能够在一个非常非常热受限的大脑环境中做到这一点。

So there has to be some amount of compression that happens to send off only the interesting data that you need, which in this particular case for motor decoding is, occurrence of a spike or not.

因此，必须进行一定程度的压缩，以便仅发送您需要的有趣数据，在电机解码的特殊情况下，即是否出现尖峰。

And then being able to use that to decode the intended cursor movement.

然后能够使用它来解码预期的光标移动。

So the implant itself processes it, figures out whether a spike happened or not with our spike detection algorithm, and then sends it off, packages it, sends it off through Bluetooth to an external device that then has the model to decode, okay, based on these spiking inputs, did Noland wish to go up, down, left, right, or click or right click or whatever.

因此，植入物本身会对其进行处理，通过我们的尖峰检测算法确定尖峰是否发生，然后将其发送、打包，通过蓝牙将其发送到外部设备，然后由模型进行解码，好吧，根据这些尖峰输入，诺兰是否希望向上、向下、向左、向右，或单击或右键单击或其他任何操作。

All of this is really fascinating, but let’s stick on the N1 Implant itself.

所有这一切都非常令人着迷，但让我们继续关注 N1 植入物本身。

2:06:23 So the thing that’s in the brain.

所以大脑里的东西。

So I’m looking at a picture of it, there’s an enclosure, there’s a charging coil, so we didn’t talk about the charging, which is fascinating.

所以我正在看它的图片，有一个外壳，有一个充电线圈，所以我们没有谈论充电，这很有趣。

The battery, the power electronics, the antenna.

电池、电力电子设备、天线。

Then there’s the signal processing electronics.

然后是信号处理电子设备。

I wonder if there’s more kinds of signal processing you can do?

我想知道是否可以进行更多种类的信号处理？

That’s another question.

那是另一个问题了。

And then there’s the threads themselves with the enclosure on the bottom.

然后是螺纹本身，外壳位于底部。

So maybe to ask about the charging.

所以也许要问一下充电的问题。

So there’s an external charging device?

那么有外接充电设备吗？

Yeah, there’s an external charging device.

是的，有一个外部充电设备。

2:07:03 So yeah, the second part of the implant, the threads are the ones, again, just the last three to five millimeters are the ones that are actually penetrating the cortex.

所以，是的，植入物的第二部分，即螺纹，同样，最后三到五毫米是真正穿透皮质的螺纹。

Rest of it is, actually most of the volume, is occupied by the battery, rechargeable battery, and it’s about a size of a quarter.

剩下的，实际上大部分的体积，都被电池，充电电池占据了，大约有四分之一的大小。

I actually have a device here if you want to take a look at it.

我这里实际上有一个设备，如果你想看一下的话。

This is the flexible thread component of it, and then this is the implant.

这是它的柔性螺纹组件，然后这是植入物。

So it’s about a size of a US quarter.

所以它大约相当于美国25美分硬币的大小。

It’s about nine millimeters thick.

它大约有九毫米厚。

So basically this implant, once you have the craniectomy and the directomy, threads are inserted, and the hole that you created, this craniectomy, gets replaced with that.

所以基本上这个植入物，一旦你进行了颅骨切除术和直切术，就会插入线，并且你创建的孔，这个颅骨切除术，会被替换。

So basically that thing plugs that hole, and you can screw in these self-drilling cranial screws to hold it in place.

所以基本上这个东西会堵塞那个洞，你可以拧入这些自钻颅骨螺钉将其固定到位。

And at the end of the day, once you have the skin flap over, there’s only about two to three millimeters that’s obviously transitioning off of the top of the implant to where the screws are.

一天结束时，一旦你将皮瓣翻过来，就会发现只有大约两到三毫米的部分明显从植入物顶部过渡到螺钉所在的位置。

And that’s the minor bump that you have.

这就是你所遇到的小问题。

Those threads look tiny.

这些线看起来很小。

2:08:22 That’s incredible.

那真是难以置信。

That is really incredible.

这实在是太不可思议了。

That is really incredible.

这实在是太不可思议了。

And also, you’re right, most of the actual volume is the battery.

而且，你是对的，大部分实际体积是电池。

This is way smaller than I realized.

这比我意识到的要小得多。

Also, the threads themselves are quite strong.

而且，线程本身非常坚固。

They look strong.

他们看起来很坚强。

And the thread themselves also has a very interesting feature at the end of it called the loop.

线程本身还有一个非常有趣的功能，称为循环。

2:08:42 And that’s the mechanism to which the robot is able to interface and manipulate this tiny hair-like structure.

这就是机器人能够连接和操纵这种微小的毛发状结构的机制。

And they’re tiny.

而且它们很小。

2:08:55 So what’s the width of a thread?

那么螺纹的宽度是多少？

So the width of a thread starts from 16 micron and then tapers out to about 84 micron.

因此，螺纹的宽度从 16 微米开始，然后逐渐减小到约 84 微米。

2:08:58 So average human hair is about 80 to 100 micron in width.

因此，人类头发的平均宽度约为 80 至 100 微米。

This thing is amazing.

这件事太神奇了。

2:09:13 This thing is amazing.

这件事太神奇了。

Yes, most of the volume is occupied by the battery, rechargeable lithium ion cell.

是的，大部分体积被电池、可充电锂离子电池占据。

2:09:16 And the charging is done through inductive charging, which is actually very commonly used.

而且充电是通过感应充电来完成的，这其实是很常用的。

Your cell phone, most cell phones, have that.

你的手机，大多数手机，都有这个功能。

The biggest difference is that for us, usually when you have a phone and you want to charge it on the charging pad, you don’t really care how hot it gets.

最大的区别是，对我们来说，通常当你有手机并且想在充电板上充电时，你并不真正关心它有多热。

Whereas, in for us, it matters.

然而，对我们来说，这很重要。

There is a very strict regulation and good reasons to not actually increase the surrounding tissue temperature by two degrees Celsius.

有一个非常严格的规定和充分的理由实际上不会将周围组织的温度增加两摄氏度。

So there’s actually a lot of innovation that is packed into this to allow charging of this implant without causing that temperature threshold to reach.

因此，实际上包含了很多创新，可以在不达到温度阈值的情况下为该植入物充电。

2:10:03 And even small things like, you see this charging coil and what’s called a ferrite shield.

甚至像这样的小东西，你会看到这个充电线圈和所谓的铁氧体屏蔽。

So without that ferrite shield, what you end up having when you have resonant inductive charging is that the battery itself is a metallic can, and you form these eddy currents from external charger and that causes heating, and that actually contributes to inefficiency in charging.

因此，如果没有铁氧体屏蔽，当进行谐振感应充电时，电池本身就是一个金属罐，并且从外部充电器形成这些涡流，从而导致发热，这实际上会导致充电效率低下。

So this ferrite shield, what it does, is that it actually concentrate that field line away from the battery and then around the coil that’s actually wrapped around it.

因此，这个铁氧体屏蔽的作用是，它实际上将磁场线集中在远离电池的地方，然后集中在实际缠绕在其周围的线圈周围。

There’s a lot of really fascinating design here to make it, I mean, you’re integrating a computer into a biological, a complex biological system.

这里有很多非常有趣的设计，我的意思是，你正在将计算机集成到生物、复杂的生物系统中。

Yeah, there’s a lot of innovation here.

是的，这里有很多创新。

2:10:52 I would say that part of what enabled this was just the innovations in the wearable.

我想说，这一切的部分原因在于可穿戴设备的创新。

There’s a lot of really, really powerful tiny, low-power microcontrollers, temperature sensors, or various different sensors and power electronics.

有很多非常非常强大的微型、低功耗微控制器、温度传感器或各种不同的传感器和电力电子设备。

A lot of innovation really came in the charging coil design, how this is packaged, and how do you enable charging such that you don’t really exceed that temperature limit, which is not a constraint for other devices out there.

许多创新确实来自于充电线圈的设计、如何封装，以及如何实现充电，这样就不会真正超过温度限制，这对其他设备来说不是一个限制。

So let’s talk about the threads themselves.

那么让我们来谈谈线程本身。

2:11:28 Those tiny, tiny, tiny things.

那些微小的、微小的、微小的东西。

So how many of them are there?

那么他们有多少人呢？

You mentioned a thousand electrodes.

你提到了一千个电极。

How many threads are there and what do the electrodes have to do with the threads?

有多少根螺纹，电极与螺纹有什么关系？

So the current instantiation of the device has 64 threads, and each thread has 16 electrodes for a total of 1,024 electrodes that are capable of both recording and stimulating.

因此，该设备的当前实例有 64 个线程，每个线程有 16 个电极，总共 1,024 个电极能够同时记录和刺激。

2:11:42 And the thread is basically this polymer-insulated wire.

线基本上就是这种聚合物绝缘线。

The metal conductor is the kind of a tiramisu cake of ti, plat, gold, plat, ti and they’re very, very tiny wires.

金属导体是一种提拉米苏蛋糕，由钛、铂、金、铂、钛组成，它们是非常非常细的电线。

Two micron in width.

宽度为两微米。

So two one-millionth of meter.

所以百万分之二米。

It’s crazy that that thing I’m looking at has the polymer-insulation, has the conducting material and has 16 electrodes at the end of it.

太疯狂了，我看到的那个东西有聚合物绝缘体、导电材料，并且末端有 16 个电极。

On each of those thread.

在每个线程上。

Yeah, on each of those threads.

是的，在每一个线程上。

Correct.

正确的。

16, each one of those 64.

16 个，这 64 个中的每一个。

Yes, you’re not going to be able to see it with naked eyes.

是的，你无法用肉眼看到它。

And to state the obvious, or maybe for people who are just listening, they’re flexible?

说实话，或者对于那些只是在听的人来说，他们很灵活？

Yes, that’s also one element that was incredibly important for us.

是的，这也是对我们来说非常重要的一个因素。

2:12:48 So each of these threads are now, as I mentioned, 16 micron in width, and then they taper to 84 micron, but in thickness they’re less than five micron.

因此，正如我提到的，现在每根螺纹的宽度均为 16 微米，然后逐渐变细至 84 微米，但厚度小于 5 微米。

And in thickness it’s mostly a polyimide at the bottom and this metal track and then another polyimide.

从厚度上看，它的底部主要是聚酰亚胺，金属轨道然后是另一个聚酰亚胺。

So two micron of polyimide, 400 nanometer of this metal stack and two micron of polyimide sandwiched together to protect it from the environment that is 37 degrees C bag of salt water.

因此，将 2 微米的聚酰亚胺、400 纳米的金属堆叠和 2 微米的聚酰亚胺夹在一起，以保护其免受 37 摄氏度的盐水袋环境的影响。

Maybe can you speak to some interesting aspects of the material design here?

也许您能在这里谈谈材料设计的一些有趣的方面吗？

2:13:26 What does it take to design a thing like this and to be able to manufacture a thing like this?

设计这样的东西并能够制造这样的东西需要什么？

For people who don’t know anything about this kind of thing.

对于对这种事情一无所知的人来说。

So the material selection that we have is not, I don’t think it was particularly unique.

所以我们的材料选择不是，我不认为它是特别独特的。

2:13:40 There were other labs and there are other labs that are kind of looking at similar material stack.

还有其他实验室也在研究类似的材料堆栈。

There’s kind of a fundamental question, and still needs to be answered, around the longevity and reliability of these microelectrodes that we call, compared to some of the other more conventional neural interfaces devices that are intracranial, so penetrating the cortex, that are more rigid, like the Utah Array.

与其他一些更传统的颅内神经接口设备相比，这些微电极的寿命和可靠性仍然需要回答，而且仍然需要回答，这些微电极位于颅内，因此穿透皮层，更加刚性，如犹他阵列。

That are these four by four millimeter kind of silicon shank that have exposed recording site at the end of it.

这就是这种四乘四毫米的硅柄，其末端暴露了记录位置。

And that’s been kind of the innovation from Richard Normann back in 1997.

这就是 Richard Normann 在 1997 年的创新。

It’s called the Utah Array because he was at University of Utah.

它被称为犹他阵列，因为他在犹他大学。

And what does the Utah Array look like?

犹他阵列是什么样子的？

2:14:36 So it’s a rigid type of [inaudible 02:14:41]?

所以这是一种僵化的类型 [听不清 02:14:41]？

Yeah, so we can actually look it up.

是的，所以我们实际上可以查一下。

2:14:40 Yeah, so it’s a bed of needle.

是的，所以这是一个针床。

There’s-

有——

Okay, go ahead.

好的，继续吧。

2:14:52 I’m sorry.

对不起。

Those are rigid shanks.

这些是刚性的柄。

Rigid, yeah, you weren’t kidding.

刚性，是的，你不是在开玩笑。

And the size and the number of shanks vary anywhere from 64 to 128.

柄的尺寸和数量从 64 到 128 不等。

2:14:57 At the very tip of it, is an exposed electrode that actually records neural signal.

在它的最顶端，是一个裸露的电极，实际上记录了神经信号。

The other thing that’s interesting to note is that unlike neural link threads that have recording electrodes that are actually exposed iridium oxide recording sites along the depth, this is only at a single depth.

另一件有趣的事情是，与具有记录电极的神经连接线程不同，记录电极实际上沿着深度暴露于氧化铱记录位点，这仅在单一深度。

So these Utah Array spokes can be anywhere between 0.5 millimeters to 1.5 millimeter, and they also have designs that are slanted.

因此，这些 Utah Array 辐条的长度可以在 0.5 毫米到 1.5 毫米之间，而且它们也有倾斜的设计。

So you can have it inserted at different depths, but that’s one of the other big differences.

因此，您可以将其插入不同的深度，但这是其他重大差异之一。

And then, the main key difference is the fact that there’s no active electronics.

然后，主要的区别是没有有源电子设备。

These are just electrodes, and then there’s a bundle of a wire that you’re seeing, and then that actually then exits the craniotomy that then has this port that you can connect to for any external electronic devices.

这些只是电极，然后是一束你看到的电线，然后实际上从开颅手术中出来，然后有一个端口，你可以连接到任何外部电子设备。

They are working on, or have, the wireless telemetry device but it still requires a through-the-skin port, that actually is one of the biggest failure modes for infection for the system.

他们正在开发或拥有无线遥测设备，但它仍然需要一个穿透皮肤的端口，这实际上是系统感染的最大故障模式之一。

What are some of the challenges associated with flexible threads?

与柔性螺纹相关的一些挑战有哪些？

2:16:06 Like for example, on the robotic side, R1, implanting those threads.

例如，在机器人方面，R1，植入这些线程。

How difficult is that task?

这项任务有多困难？

Yeah, so as you mentioned, they’re very, very difficult to maneuver by hand.

是的，正如您提到的，它们非常非常难以手动操作。

2:16:19 These Utah Arrays that you saw earlier, they’re actually inserted by a neurosurgeon actually positioning it near the site that they want.

您之前看到的这些犹他阵列实际上是由神经外科医生插入的，实际上将其放置在他们想要的位置附近。

And then there’s a pneumatic hammer that actually pushes them in.

然后有一个气动锤将它们实际推入。

So it’s a pretty simple process and they’re easy to maneuver.

所以这是一个非常简单的过程，而且很容易操作。

But for these thin-film arrays, they’re very, very tiny and flexible.

但对于这些薄膜阵列来说，它们非常非常小且灵活。

So they’re very difficult to maneuver.

所以他们很难操纵。

So that’s why we built an entire robot to do that.

这就是为什么我们建造了一个完整的机器人来做到这一点。

2:16:55 There are other reasons for why we built the robot, and that is ultimately we want this to help millions and millions of people that can benefit from this.

我们建造机器人还有其他原因，最终我们希望它能够帮助数以百万计的人从中受益。

And there just aren’t that many neurosurgeons out there.

而且那里的神经外科医生并不多。

And robots can be something that we hope can actually do large parts of the surgery.

我们希望机器人能够真正完成大部分手术。

But the robot is this entire other sort of category of product that we’re working on.

但机器人是我们正在开发的完全不同类型的产品。

And it’s essentially this multi- axis gantry system that has the specialized robot head that has all of the optics and this kind of a needle-retracting mechanism that maneuvers these threads via this loop structure that you have on the thread.

本质上，这个多轴龙门系统具有专门的机器人头，该机器人头具有所有光学器件和这种针缩回机构，可通过在线上的循环结构操纵这些线。

So the thread already has a loop structure by which you can grab it?

那么线程已经有一个循环结构，你可以通过它来抓取它？

Correct.

正确的。

So this is fascinating.

所以这很有趣。

2:17:56 So you mentioned optics.

所以你提到了光学。

So there’s a robot, R1, so for now, there’s a human that actually creates a hole in the skull.

所以有一个机器人，R1，所以现在，有一个人实际上在头骨上打了一个洞。

And then after that, there’s a computer vision component that’s finding a way to avoid the blood vessels.

之后，有一个计算机视觉组件正在寻找一种避开血管的方法。

And then you’re grabbing it by the loop, each individual thread, and placing it in a particular location to avoid the blood vessels and also choosing the depth of placement, all that.

然后你抓住它的环，每根单独的线，并将其放置在特定位置以避免血管，并选择放置深度，所有这些。

So controlling every, the 3D geometry, of the placement?

那么要控制每个 3D 几何形状吗？

Correct.

正确的。

2:18:31 So the aspect of this robot that is unique is that it’s not surgeon-assisted or human-assisted.

因此，该机器人的独特之处在于它不需要外科医生辅助或人工辅助。

It’s a semi-automatic or automatic robot.

它是一个半自动或自动机器人。

Obviously, there are human component to it, when you’re placing targets, you can always move it away from major vessels that you see.

显然，其中有人为因素，当你放置目标时，你总是可以将其移离你看到的主要船只。

But we want to get to a point where one click and it just does the surgery within minutes.

但我们希望达到这样的程度：只需点击一下，几分钟内就能完成手术。

So the computer vision component finds great targets, candidates, and the human approves them, and the robot does…

因此，计算机视觉组件找到了很好的目标、候选者，人类批准了它们，机器人也批准了……

2:18:57 Does it do one thread at a time?

它一次只执行一个线程吗？

Or does it do them [inaudible 02:19:08]?

或者它对他们有影响吗 [听不清 02:19:08]？

It does one thread at a time.

它一次执行一个线程。

2:19:07 And that’s actually also one thing that we are looking at ways to do multiple threads at a time.

这实际上也是我们正在寻找同时执行多个线程的方法的一件事。

There’s nothing stopping from it.

没有什么可以阻止它。

You can have multiple kind of engagement mechanisms.

您可以有多种参与机制。

But right now, it’s one-by-one.

但现在，却是一一的。

And we also still do quite a bit of just kind of verification to make sure that it got inserted.

我们仍然做了相当多的验证以确保它被插入。

If so, how deep?

如果有，有多深？

Did it actually match what was programmed in?

它真的与编程的内容相符吗？

And so on and so forth.

等等。

And the actual electrodes are placed at differing depths in the…

实际的电极放置在......的不同深度。

2:19:36 I mean, it’s very small differences, but differences.

我的意思是，差异很小，但是差异。

Yeah.

是的。

And so there’s some reasoning behind that, as you mentioned, it gets more varied signal.

正如您所提到的，这背后有一些原因，它会获得更加多样化的信号。

Yeah, we try to place them all around three or four millimeter from the surface.

是的，我们尝试将它们放置在距表面约三到四毫米的位置。

2:20:00 …

……

it’s three or four millimeter from the surface just because the span of the electrode, those 16 electrodes that we currently have in this version, spans roughly around three millimeters.

它距表面三到四毫米，只是因为电极的跨度（我们目前在这个版本中拥有 16 个电极）的跨度大约为 3 毫米。

So we want to get all of those in the brain.

所以我们希望将所有这些都纳入大脑中。

This is fascinating.

这很有趣。

2:20:16 Okay, so there’s a million questions here.

好的，这里有一百万个问题。

If we could zoom in specifically on the electrodes.

如果我们可以专门放大电极。

What is your sense, how many neurons is each individual electrode listening to?

您的感觉是什么？每个单独的电极监听了多少个神经元？

Yeah, each electrode can record from anywhere between zero to 40, as I mentioned earlier.

是的，正如我之前提到的，每个电极可以记录 0 到 40 之间的任何位置。

2:20:27 But practically speaking, we only see about at most two to three, and you can actually distinguish which neuron it’s coming from by the shape of the spikes.

但实际上，我们最多只能看到两到三个，而且你实际上可以通过尖峰的形状来区分它来自哪个神经元。

Oh, cool.

哦，酷。

I mentioned the spike detection algorithm that we have, it’s called BOSS algorithm, Buffer Online Spike Sorter.

我提到了我们拥有的尖峰检测算法，它被称为 BOSS 算法，Buffer Online Spike Sorter。

Nice.

好的。

It actually outputs at the end of the day six unique values, which are the amplitude of these negative going hump, middle hump, positive going hump, and then also the time at which these happen.

它实际上在一天结束时输出六个唯一值，即这些负向驼峰、中间驼峰、正向驼峰的幅度，以及这些发生的时间。

2:20:59 And from that, you can have a statistical probability estimation of, “Is that a spike?

由此，您可以进行统计概率估计，“这是一个峰值吗？

Is it not a spike?” And then based on that, you could also determine, “Oh, that spike looks different than that spike, it must come from a different neuron.”

不是尖刺吗？”然后基于此，你还可以确定，“哦，那个尖峰看起来与那个尖峰不同，它一定来自不同的神经元。”

Okay.

好的。

2:21:27 So that’s a nice signal processing step from which you can then make much better predictions about if there’s a spike, especially in this kind of context, where there could be multiple neurons screaming.

因此，这是一个很好的信号处理步骤，您可以通过它对是否存在尖峰做出更好的预测，特别是在这种可能有多个神经元尖叫的情况下。

And that that also results in you being able to compress the data better at the of the day.

这也使您能够在当天更好地压缩数据。

Yeah.

是的。

Okay, that’s-

好吧，那就是——

And just to be clear, I mean, the labs do this what’s called spike sorting.

需要明确的是，我的意思是，实验室进行了所谓的尖峰排序。

2:21:46 Usually once you have the fully digitized signals and then you run a bunch of different set of algorithms to tease apart, it’s just all of this for us is done on the device.

通常，一旦您拥有完全数字化的信号，然后运行一组不同的算法来梳理，我们的所有这些都是在设备上完成的。

On the device.

在设备上。

In a very low power, custom-built ASIC digital processing unit.

采用功耗极低的定制 ASIC 数字处理单元。

Highly heat constrained.

高度受热约束。

Highly heat constrained.

高度受热约束。

2:22:15 And the processing time from signal going in and giving you the output is less than a microsecond, which is a very, very short amount of time.

从信号输入到输出的处理时间不到一微秒，这是一个非常非常短的时间。

Oh, yeah.

哦，是的。

2:22:25 So the latency has to be super short.

所以延迟必须超短。

Correct.

正确的。

Oh, wow.

哦，哇。

2:22:28 Oh, that’s a pain in the ass.

哦，那真是令人痛苦。

That’s really tough.

这真的很难。

Yeah, latency is this huge, huge thing that you have to deal with.

是的，延迟是你必须处理的一个非常非常大的问题。

2:22:32 Right now the biggest source of latency comes from the Bluetooth, the way in which their packetized and we bin them in a 15 millisecond time window.

目前，延迟的最大来源来自蓝牙，即它们的分组方式，我们将它们分组在 15 毫秒的时间窗口中。

Oh, interesting, so it’s communication constrained.

哦，有趣，所以沟通受到限制。

2:22:44 Is there some potential innovation there on the protocol used?

所使用的协议是否有一些潜在的创新？

Absolutely.

绝对地。

Okay.

好的。

Yeah.

是的。

2:22:49 Bluetooth is definitely not our final wireless communication protocol that we want to get to.

蓝牙绝对不是我们最终想要的无线通信协议。

It’s highly-

这是高度-

Hence, the N1 and the R1.

因此，N1 和 R1。

2:22:59 I imagine that increases [inaudible 02:23:03].

我想这会增加 [听不清 02:23:03]。

Nx, Rx.

NX、Rx。

Yeah, that’s the communication protocol because Bluetooth allows you to communicate, gets farther distances than you need to, so you can go much shorter.

是的，这就是通信协议，因为蓝牙允许您进行通信，获得比您需要的更远的距离，因此您可以走得更短。

Yeah.

是的。

2:23:16 The only, well, the primary motivation for choosing Bluetooth is that, I mean, everything has Bluetooth,

选择蓝牙的唯一主要动机是，我的意思是，一切都有蓝牙，

All right, so you can talk to any device.

好的，这样您就可以与任何设备通话了。

Interoperability is just absolutely essential, especially in this early phase.

互操作性是绝对必要的，尤其是在这个早期阶段。

2:23:23 And in many ways, if you can access a phone or a computer, you can do anything.

在很多方面，如果你可以使用手机或电脑，你就可以做任何事情。

It’ll be interesting to step back and actually look at, again, the same pipeline that you mentioned for Noland.

退后一步，再次实际看看您为诺兰提到的同一条管道，这会很有趣。

2:23:35 What does this whole process look like from finding and selecting a human being, to the surgery, to the first time he’s able to use this thing?

从寻找和选择一个人，到进行手术，到他第一次能够使用这个东西，这整个过程是什么样的？

We have what’s called a patient registry that people can sign up to hear more about the updates.

我们有所谓的患者登记处，人们可以注册以了解有关更新的更多信息。

2:23:56 And that was a route to which Noland applied.

这就是诺兰申请的一条路线。

And the process is that once the application comes in, it contains some medical records, and we …

这个过程是，一旦申请进来，它包含一些医疗记录，我们......

Based on their medical eligibility, there’s a lot of different inclusion/exclusion criteria for them to meet.

根据他们的医疗资格，他们需要满足许多不同的纳入/排除标准。

2:24:22 And we go through a prescreening interview process with someone from Neuralink, and at some point we also go out to their homes to do a BCI home audit.

我们与 Neuralink 的人员进行了预筛选面试过程，有时我们还会去他们的家中进行 BCI 家庭审核。

Because one of the most revolutionary part about having this in one system that is completely wireless, is that you can use it at home.

因为将其集成到一个完全无线的系统中最具革命性的部分之一是您可以在家中使用它。

You don’t actually have to go to the lab and go to the clinic to get connectedorized to these specialized equipment that you can’t take home with you.

实际上，您不必去实验室或去诊所即可连接到这些无法带回家的专用设备。

2:24:51 So that’s one of the key elements of when we’re designing the system that we wanted to keep in mind, people hopefully would want to be able to use this every day in the comfort of their homes.

因此，这是我们在设计系统时要牢记的关键要素之一，希望人们能够每天在舒适的家中使用它。

And so part of our engagement and what we’re looking for during BCI home audit is to just understand their situation, what other assisted technology that they use.

因此，我们参与的一部分以及我们在 BCI 家庭审计期间所寻求的就是了解他们的情况以及他们使用的其他辅助技术。

And we should also step back and say that the estimate is 180,000 people live with quadriplegia in the United States, and each year an additional 18,000 suffer a paralyzing spinal cord injury.

我们还应该退一步说，据估计，美国有 180,000 人患有四肢瘫痪，而且每年还有 18,000 人遭受瘫痪脊髓损伤。

2:25:14 So these are folks who have a lot of challenges living a life in terms of accessibility, in terms of doing the things that many of us just take for granted day to day.

因此，这些人在生活中面临着很多挑战，在可及性方面，在做我们许多人每天认为理所当然的事情方面。

2:25:42 And one of the things, one of the goals of this initial study is to enable them to have digital autonomy where they by themselves can interact with a digital device using just their mind, something that you’re calling telepathy, so digital telepathy.

其中之一，这项初步研究的目标之一是使他们能够拥有数字自主权，他们自己可以仅用他们的思想与数字设备进行交互，这就是你所说的心灵感应，数字心灵感应。

Where a quadriplegic can communicate with a digital device in all the ways that we’ve been talking about.

四肢瘫痪者可以通过我们一直在谈论的所有方式与数字设备进行通信。

Control the mouse cursor enough to be able to do all kinds of stuff, including play games and tweet and all that kind of stuff.

足够控制鼠标光标，以便能够执行各种操作，包括玩游戏和发推文以及所有此类操作。

And there’s a lot of people for whom life, the basics of life, are difficult because of the things that have happened to them.

对很多人来说，生活、生活的基本要素都因为发生在他们身上的事情而变得困难。

Yeah.

是的。

2:26:24 I mean, movement is so fundamental to our existence.

我的意思是，运动对于我们的存在至关重要。

I mean, even speaking involves movement of mouth, lip, larynx.

我的意思是，即使说话也涉及嘴、嘴唇、喉部的运动。

And without that, it’s extremely debilitating.

如果没有这一点，它会非常令人衰弱。

And there are many, many people that we can help.

我们可以帮助很多很多人。

I mean, especially if you start to look at other forms of movement disorders that are not just from spinal cord injury, but from a ALS, MS, or even stroke, or just aging, that leads you to lose some of that mobility, that independence, it’s extremely debilitating.

我的意思是，特别是如果你开始关注其他形式的运动障碍，这些运动障碍不仅来自脊髓损伤，还来自肌萎缩性脊髓侧索硬化症、多发性硬化症，甚至中风，或者只是衰老，这会导致你失去一些活动能力、独立性，这是极其虚弱的。

And all of these are opportunities to help people, to help alleviate suffering, to help improve the quality of life.

所有这些都是帮助人们、帮助减轻痛苦、帮助改善生活质量的机会。

2:27:09 But each of the things you mentioned is its own little puzzle that needs to have increasing levels of capability from a device like a Neuralink device.

但你提到的每一件事都是它自己的小难题，需要像 Neuralink 设备这样的设备不断提高功能水平。

2:27:24 And so the first one you’re focusing on is, it’s just a beautiful word, telepathy.

所以你关注的第一个是，心灵感应，这是一个美丽的词。

So being able to communicate using your mind wirelessly with a digital device.

因此能够用你的思想与数字设备进行无线通信。

Can you just explain exactly what we’re talking about?

你能准确解释一下我们在说什么吗？

Yeah, I mean, it’s exactly that.

是的，我的意思是，正是如此。

2:27:40 I mean, I think if you are able to control a cursor and able to click and be able to get access to a computer or a phone, I mean, the whole world opens up to you.

我的意思是，我认为如果你能够控制光标并能够点击并能够访问计算机或手机，我的意思是，整个世界都会向你敞开。

And I mean, I guess the word “telepathy,” if you think about that as just definitionally being able to transfer information from my brain to your brain without using some of the physical faculties that we have, like voices.

我的意思是，我猜“心灵感应”这个词，如果你认为它的定义是能够将信息从我的大脑转移到你的大脑，而不需要使用我们拥有的一些身体能力，比如声音。

But the interesting thing here is I think the thing that’s not obviously clear is how exactly it works.

但有趣的是，我认为尚不清楚的是它到底是如何工作的。

2:28:13 In order to move a cursor, there’s at least a couple of ways of doing that.

为了移动光标，至少有几种方法可以做到这一点。

One is you imagine yourself maybe moving a mouse with your hand, or you can then, which no one talked about, imagine moving the cursor with your mind.

一种是你想象自己用手移动鼠标，或者你可以想象用你的意念移动光标，但没有人谈论这一点。

2:28:44 But it’s like there is a cognitive step here that’s fascinating, because you have to use the brain and you have to learn how to use the brain, and you have to figure it out dynamically because you reward yourself if it works.

但这就像这里有一个令人着迷的认知步骤，因为你必须使用大脑，你必须学习如何使用大脑，你必须动态地弄清楚它，因为如果它有效，你就会奖励自己。

I mean, there’s a step that …

我的意思是，有一个步骤......

This is just a fascinating step because you have to get the brain to start firing in the right way.

这只是一个令人着迷的步骤，因为你必须让大脑以正确的方式开始放电。

And you do that by imagining …

你通过想象来做到这一点……

Like fake it till you make it.

就像假装一样，直到你成功为止。

And all of a sudden it creates the right kind of signal that, if decoded correctly, can create the effect.

突然之间，它会产生正确的信号，如果正确解码，就可以产生效果。

And then there’s noise around that that you have to figure all of that out.

然后周围就有噪音，你必须弄清楚所有这些。

But on the human side, imagine the cursor moving is what you have to do.

但从人的角度来看，想象一下光标移动是你必须做的。

Yeah.

是的。

2:29:27 He says using the force.

他说用武力。

The force.

力量。

2:29:29 I mean, isn’t that just fascinating to you that it works?

我的意思是，它的功效难道不令你着迷吗？

To me, it’s like, holy shit, that actually works.

对我来说，天哪，这确实有效。

You could move a cursor with your mind.

你可以用意念移动光标。

As much as you’re learning to use that thing, that thing is also learning about you.

就像你在学习如何使用那个东西一样，那个东西也在了解你。

2:29:41 Our model’s constantly updating the way to say, “Oh, if someone is thinking about this sophisticated forms of spiking patterns, that actually means to do this.”

我们的模型不断更新说的方式：“哦，如果有人正在考虑这种复杂形式的尖峰模式，那实际上意味着要这样做。”

So the machine is learning about the human and the human is learning about the machine, so there is a adaptability to the signal process and the decoding step, and then there’s the adaptation of Nolan, the human being.

因此，机器正在学习人类，人类也在学习机器，因此存在对信号处理和解码步骤的适应性，然后是人类诺兰的适应。

2:30:02 The same way, if you give me a new mouse and I move it, I learn very quickly about its sensitivity, so I learn to move it slower.

同样，如果你给我一个新鼠标，我移动它，我很快就会了解它的灵敏度，因此我会学着更慢地移动它。

And then there’s other signal drift and all that kind of stuff they have to adapt to, so both are adapting to each other.

然后还有其他信号漂移以及它们必须适应的所有此类东西，因此两者都在相互适应。

Correct.

正确的。

That’s a fascinating software challenge, on both sides.

对于双方来说，这都是一个令人着迷的软件挑战。

2:30:34 The software on both, on the human software and the [inaudible 02:30:41] software.

软件既包括人类软件，也包括[听不清 02:30:41] 软件。

The organic and the inorganic.

有机和无机。

The organic and the inorganic.

有机和无机。

2:30:43 Anyway.

反正。

Sorry to rudely interrupt.

抱歉粗鲁地打扰了。

So there’s the selection that Noland has passed with flying colors.

诺兰出色地通过了这个选择。

Everything, including that it is a BCI-friendly home, all of that.

一切，包括它是一个 BCI 友好型住宅，所有这些。

So what is the process of the surgery, implantation, the first moment when he gets to use the system?

那么手术、植入、第一次使用该系统的过程是怎样的呢？

The end-to-end, we say patient end to patient out, is anywhere between two to four hours.

端到端（我们称患者端到患者出）的时间为两到四个小时。

2:31:06 In the particular case for Noland it was about three and a half hours, and there’s many steps leading to the actual robot insertion.

在诺兰的特殊情况下，大约需要三个半小时，并且有许多步骤导致实际的机器人插入。

So there’s anesthesia induction, and we do intra-op CT imaging to make sure that we’re drilling the hole in the right location.

因此，我们进行麻醉诱导，并进行术中 CT 成像，以确保我们在正确的位置钻孔。

And this is also pre-planned beforehand.

而这也是事先就计划好的。

2:31:34 Someone like Noland would go through fMRI and then they can think about wiggling their hand.

像诺兰这样的人会进行功能磁共振成像，然后他们可以考虑摆动自己的手。

Obviously due to their injury it’s not going to actually lead to any sort of intended output, but it’s the same part of the brain that actually lights up when you’re imagining moving your finger to actually moving your finger.

显然，由于他们的受伤，它实际上不会导致任何预期的输出，但当你想象移动手指和实际移动手指时，大脑的同一部分实际上会被点亮。

And that’s one of the ways in which we can actually know where to place our threads because we want to go into what’s called the hand knob area in the motor cortex.

这是我们实际上知道在哪里放置线程的方法之一，因为我们想要进入运动皮层中所谓的手柄区域。

And as much as possible, densely put our electrode threads.

并且尽可能密集地放置我们的电极线。

2:32:11 So we do intra-op CT imaging to make sure and double-check the location of the craniectomy.

因此，我们进行术中 CT 成像，以确保并仔细检查去骨瓣的位置。

And the surgeon comes in, does their thing in terms of skin incision, craniectomy, so drilling of the skull, and then there’s many different layers of the brain.

外科医生进来，做他们的工作，包括皮肤切口、颅骨切除、颅骨钻孔，然后是大脑的许多不同层。

There’s what’s called a dura, which is a very, very thick layer that surrounds the brain.

有一种叫做硬脑膜的东西，它是围绕大脑的一层非常非常厚的层。

That gets actually resected in a process called [inaudible 02:32:38].

这实际上是在一个名为 [听不清 02:32:38] 的过程中被切除的。

And that then expose the pia in the brain that you want to insert.

然后暴露你想要插入的大脑中的软脑膜。

2:32:43 And by the time it’s been around anywhere between one to one and a half hours, robot comes in, does his thing, placement of the targets, inserting of the thread.

大约一到一个半小时后，机器人就会进来，做他的事情，放置目标，插入线。

That takes anywhere between 20 to 40 minutes.

这需要 20 到 40 分钟。

In the particular case for Noland, it was just under or just over 30 minutes.

对于诺兰德来说，这个时间只有不到 30 分钟或略多于 30 分钟。

And then after that, the surgeon comes in, there’s a couple other steps of actually inserting the dural substitute layer to protect the thread as well as the brain.

之后，外科医生进来，还有其他几个步骤，实际插入硬脑膜替代层以保护线和大脑。

And then screw in the implant and then skin flap and then suture, and then you’re out.

然后拧入植入物，然后皮瓣，然后缝合，然后就可以了。

So when Noland woke up, what was that like?

那么当诺兰醒来时，那是什么样的呢？

2:33:18 What was the recovery like, and when was the first time he was able to use it?

恢复情况如何？他第一次能够使用它是什么时候？

He was actually immediately after the surgery, like an hour after the surgery, as he was waking up, we did turn on the device, make sure that we are recording neural signals.

他实际上是在手术后不久，大约手术后一个小时，当他醒来时，我们确实打开了设备，确保我们正在记录神经信号。

2:33:27 And we actually did have couple signals that we noticed that he can actually modulate.

事实上，我们确实有几个信号，我们注意到他实际上可以调节。

And what I mean by modulate is that he can think about clenching his fist and you could see the spike disappear and appear.

我所说的调节的意思是，他可以考虑握紧拳头，你可以看到尖峰消失并出现。

That’s awesome.

太棒了。

And that was immediate, immediate after in the recovery room.

那是立即的，就在康复室之后。

How cool is that?

那有多酷？

Yeah, absolutely.

是的，绝对是。

That’s a human being …

那是一个人类……

2:34:06 I mean, what did that feel like for you?

我的意思是，这对你来说是什么感觉？

This device and a human being, a first step of a gigantic journey?

这个设备和一个人，是一个巨大旅程的第一步吗？

I mean, it’s a historic moment, even just that spike, just to be able to modulate that.

我的意思是，这是一个历史性的时刻，即使只是那个峰值，只是为了能够调节它。

Obviously there have been other, as you mentioned, pioneers that have participated in these groundbreaking BCI investigational early feasibility studies.

显然，正如您提到的，还有其他先驱者参与了这些开创性的 BCI 早期可行性研究。

2:34:22 So we’re obviously standing on the shoulders of the giants here, we’re not the first ones to actually put electrodes in a human brain.

所以我们显然是站在巨人的肩膀上，我们并不是第一个真正将电极放入人脑的人。

2:34:44 But I mean, just leading up to the surgery, I definitely could not sleep.

但我的意思是，就在手术前，我绝对无法入睡。

It’s the first time that you’re working in a completely new environment.

这是您第一次在全新的环境中工作。

We had a lot of confidence based on our benchtop testing or preclinical R&D studies that the mechanism, the threads, the insertion, all that stuff is very safe and that it’s obviously ready for doing this in a human.

根据我们的台式测试或临床前研发研究，我们非常有信心，认为机制、螺纹、插入，所有这些东西都是非常安全的，而且它显然已经准备好在人类身上做到这一点。

But there’s still a lot of unknown unknown about can the needle actually insert?

但针到底能不能插入，还有很多未知的地方？

I mean, we brought something like 40 needles just in case they break, and we ended up using only one.

我的意思是，我们带了大约 40 根针，以防万一它们断裂，但最终我们只使用了一根。

But I mean, that was the level of just complete unknown because it’s a very, very different environment.

但我的意思是，那是完全未知的水平，因为这是一个非常非常不同的环境。

And I mean, that’s why we do clinical trial in the first place, to be able to test these things out.