TECH009：太空数据中心、AI教育、触觉机器人技术等，与Seb Bunney一起探讨

它们通过利用创新，将复杂性转化为增长。

It's how they leverage innovation to turn complexity into growth.

亚马逊广告正是这样做的，其背后是 AWS 人工智能的支持。

That's exactly what Amazon Ads is doing, powered by AWS AI.

每天，亚马逊广告处理数十亿次实时决策，优化着一个价值 310 亿美元的广告生态系统中的广告表现。

Every day, Amazon Ads processes billions of real time decisions, optimizing ad performance across a $31,000,000,000 advertising ecosystem.

其结果是，广告活动运行速度提升 30%，并能大规模带来可衡量的业务影响。

The result is campaigns that run 30% faster and deliver measurable business impact at scale.

而这也是亚马逊自身实现增长的方式。

And this is how Amazon itself drives growth.

它们的代理式人工智能将营销从一个资源密集型流程转变为智能自主系统，最大化投资回报率，并赋能营销人员专注于创意与战略。

Their agentic AI transforms marketing from a resource heavy process into an intelligent autonomous system that maximizes ROI and empowers marketers to focus on creativity and strategy.

亚马逊广告正在证明，人工智能驱动的广告不仅是未来，更是新的竞争优势。

Amazon Ads is proving that AI driven advertising isn't just the future, it's the new competitive advantage.

更棒的是，每一家企业都可以应用亚马逊内部精心打磨的同一套创新方法论。

And better yet, every enterprise can apply the same innovation playbook that Amazon perfected in house.

在 aws.com/ai/rstory 查看 Amazon Ads 的故事。

See the Amazon Ads story at aws.comai/rstory.

网址是 aws.com/ai/rstory。

That's aws.com/ai/rstory.

初创公司行动迅速。

Startups move fast.

借助 AI，它们交付速度更快，并更早吸引企业客户。

And with AI, they're shipping even faster and attracting enterprise buyers sooner.

但大单会带来更大的安全和合规要求。

But big deals bring even bigger security and compliance requirements.

SOC 2 并不总是足够。

A SOC two isn't always enough.

适当的安全措施可以促成或破坏一笔交易。

The right kind of security can make a deal or break it.

但哪位创始人或工程师能抽时间离开公司建设呢？

But what founder or engineer can afford to take time away from building their company?

Vanta 的人工智能和自动化功能可在数天内轻松为大单做好准备。

Vanta's AI and automation make it easy to get big deals ready in days.

Vanta 持续监控您的合规状态，确保未来的交易不会受阻。

And Vanta continuously monitors your compliance so future deals are never blocked.

此外，Vanta 随您一同成长，并在每一步都提供及时的支持。

Plus Vanta scales with you, backed by support that's there when you need it every step of the way.

随着人工智能改变法规和买家的期望，Vanta 知道何时需要什么，并已打造了最快、最便捷的路径，助您达成目标。

With AI changing regulations and buyers' expectations, Vanta knows what's needed and when, and they've built the fastest, easiest path to help you get there.

因此，认真的初创公司都会尽早通过 Vanta 实现安全合规。

That's why serious startups get secure early with Vanta.

我们的听众可在 vanta.com/billionaires 获得 1000 美元优惠。

Our listeners get $1,000 off at vanta.com/billionaires.

访问 vanta.com/billionaires，立减 1000 美元。

That's vanta.com/billionaires for $1,000 off.

新的一年到了，这是最终实现您梦想创业的最好时机。

It's the new year, which means that it's the best time to finally start the business you've been dreaming about.

就在几年前，我创办了自己的电子商务业务，而Shopify正是我起步时需要的工具。

Just a couple years ago, I launched my own e commerce business and Shopify was exactly the tool I needed to get started.

尽管许多人不断推迟自己的梦想，等到明年再行动，但我在这里要告诉你，现在就是抓住眼前机遇的时候。

While many people continually push off their dreams until the next year, I am here to tell you that now is the time to capitalize on the opportunities right in front of you.

Shopify为你提供了在线和线下销售所需的一切。

Shopify gives you everything you need to sell online and in person.

数百万创业者，包括我自己，都已经从普通人跃升为刚刚起步的创业者。

Millions of entrepreneurs, including myself, have already made this leap from household names to first time business owners just getting started.

你可以从数百个精美的模板中选择，并自定义它们，同时使用其内置的AI工具撰写产品描述或编辑产品图片。

Choose from hundreds of beautiful templates that you can customize and use their built in AI tools to write product descriptions or edit product photos.

随着你的成长，Shopify也会在每一步与你共同成长。

And as you grow, Shopify grows with you every step of the way.

在2026年，别再等待，立即用Shopify开始销售吧。

In 2026, stop waiting and start selling with Shopify.

立即注册每月1美元的试用版，今天就开始在shopify.com/wsb上销售。

Sign up for your $1 per month trial and start selling today at shopify.com/wsb.

前往 shopify.com/wsb。

Go to shopify.com/wsb.

那就是 shopify.com/wsb。

That's shopify.com/wsb.

在新的一年里，让 Shopify 伴你开启新篇章。

Hear your first this new year with Shopify by your side.

好的。

All right.

回到节目。

Back to the show.

我认为这一切的重大突破在于人工智能能够感知孩子是如何学习的。

I think the big breakthrough on all of this is going to be just the AI's ability to sense how the child learns.

看看你的孩子，任何有孩子的人都知道，每个孩子之间的学习方式都非常不同。

Just look at your kids, anybody that's got kids, you know, between one and the other, they learn very differently.

有些孩子需要通过例子来学习。

Some have to go through examples.

有些人必须，你知道，每个人的学习方式都不同。

Some have to, you know everybody has a different way of learning.

我给你举个例子。

I'll give you an example.

我喜欢有声书。

I love audiobooks.

我更喜欢有声书而不是实体书，实际上我更喜欢阅读，当我翻页的时候。

I prefer an audiobook over the physical I actually I prefer reading to me as I'm flipping the pages ultimately.

但如果必须在两者之间选一个，我其实更喜欢音频版本，因为我就是这种学习方式，学得更好一点，每个人都不一样。

But if I had to choose one over the other, would actually prefer the audio version because I just kinda learn that way a little bit better, and it's different for everybody.

所以举个例子，AI将会学习这些不同的人所使用的技术，或者最适合他们的方法，这是最好的表达方式。

And so just as an example, the AI is gonna learn these different techniques that people have or that's most optimal for them, the best way to frame it.

举个有趣的例子，当我还在西点军校上学时，我们总是参加一些考试，题目都是以这样的方式呈现的：比如在数学课上，题目会说，你有一枚炮弹，你要把它射向某个目标。

As a funny example, when I was a student at the military academy at West Point, we would always have these tests that were framed from like if you're in a math class, it was like, you have an artillery round and you're going to shoot it at whatever.

而且这些题目总是用某种军事场景来表述。

And it was like always framed for some type of military example.

我们总是翻白眼，心想：天哪，你能不能给个正常的题目，别总是搞这些军事相关的？

We would always roll our eyes and be like, oh my God, can you just give us a normal question and not some military type question?

但我用这个例子来说明如何设定问题的背景。

But I use that as an example to frame the framing of the thing.

假设你的孩子喜欢足球，或者你的女儿喜欢跳舞，或者其他任何事情。

Let's say your kid is he loves football or your daughter loves dance or whatever.

这些例子完全可以根据他们真正感兴趣的内容和思维方式来设计。

Like, the examples could always be framed in a way that is exactly what they wanna hear and how they want to think about it.

对吧？

Right?

所以我觉得这一点将会非常重要。

So I think that that's gonna be huge.

现在我认为还欠缺的一点是，在新冠疫情期间，孩子们不得不通过Zoom在线上课。

Now the part that I think is still lacking, when we went through COVID, the kids had to do online Zoom call like classes.

说实话，塞布，那简直是一团糟。

And to be quite honest with you, Seb, it was disastrous.

就像，那简直是一场灾难。

Like, it was just it was a train.

任何经历过面对面教育和通过电脑屏幕学习的人，都会知道电脑屏幕有一些优势，但面对面学习也有很多优势。

Anybody who's gone to in person education versus learning through a computer screen, like there's some advantages for the computer screen, but there's a lot of advantages to in person.

我想知道，一旦你开始深入探讨——我不确定每个人都会同意这一点，但也许类人机器人AI会有所不同，因为你是在进行面对面的互动。

And I wonder if once you start getting into I don't know that everybody would agree with this, but maybe the humanoid robot AI is going to maybe have a difference because you're having a in person interaction.

所以，一旦到了

And so Once where

我们十年、十五年或二十年后，会在AI的这些想法上走到哪一步呢？

are we in ten or fifteen or twenty years with respect to some of these ideas with AI

你可以把东西带到真实环境中，去黑板前，或者去户外活动，亲眼看看各种东西。

you put something into the physical environment and you could go over to a chalkboard or you could go on a field day and you could see whatever.

我不知道。

I don't know.

我认为，当学习开始融入物理空间，而不仅仅是总盯着电脑屏幕时，学习就会进入一个全新的层次。

I think that the learning kinda takes on a whole new level when you start incorporating it into physical space versus just always looking at something on a computer screen.

也许你可以在VR环境中做到这一点。

And maybe you can do that with a VR environment.

我个人不喜欢这些戴在脸上的东西。

I personally don't like these things on my face.

我觉得它们非常烦人，但有些人可能喜欢，或者也能通过这种方式学习。

I find them to be highly annoying, but some people might like it or learn that way as well.

我觉得你说得非常对，这是一个非常重要的观点。因为当我们讨论这些科技播客中的任何一点时，我们往往在谈论最新的技术以及它如何影响我们，但任何与我们和我们的世界互动、影响我们行为的事物，都必然有其利弊。

I think you are spot on and I think that's a really important point to mention because I think that when we're talking about any of these points in any of these tech podcasts we do, we're kind of talking about what is the newest technology and how it's impacting us, but there's always going to be pros and cons to everything that kind of interacts with us and our world and how we show up.

当我想到教育时，我认为知识只是其中一方面，但当我们上学时，我们与同龄人在一起，与老师互动，这其中还包含了人际连接。

And what I think about when I think about education is that the knowledge is one aspect of it, but when we're in school, we're around our peers, when we're interacting with our teacher, there is also the human connection.

我认为，在我们成长的求学阶段，我们也在学习如何调节自己的神经系统。

And I think part of during our developmental years while we're in school, we're also learning how to regulate our nervous system.

因此，我们与老师共同调节情绪，当老师平静而稳定时，这有助于我们的学习。那么，当我们用数字实体取代这些真实的人时，这会带来什么影响呢？

And so we're co regulating with the teacher, the teacher when they're calm and grounded that helps us learn and so what does that do when we're replacing these physical beings with this digital entity?

这个数字实体能够具备情感吗？

Is that digital entity going to be able to be emotional?

它们会具备那种同理心和支持他们的能力吗？

Are they going to have that empathy and that capacity to support them?

还是说，你根本无法用数字实体来替代这种联系？

Or can you just completely not replace that with a digital entity?

从共鸣的角度来看，我们所互动的东西要多得多。

Like, there's far more from a resonance perspective that we're interacting with.

而这正是我觉得非常有趣的地方。

And and that's that's what I find really interesting.

我完全同意。

I wholeheartedly agree.

我告诉你，我妻子听到我说‘哦，是的’时，

I'd tell you my wife, hearing me say, oh, yeah.

你竟然会让一个AI人形机器人来教她，她一定会对此感到厌恶。

You're gonna have an AI humanoid robot, like, teaching the she would just be disgusted by such a comment.

我认为，有很多人可能会说：天哪。

And I think that there's a lot of people out there that probably would be like, oh my god.

这听起来像是一个绝对可怕的功能。

That sounds like a absolute nightmare of a feature.

而且，嘿，也许它对某些人来说确实是个噩梦，我不确定，但我能理解为什么有些人会需要这些功能，因为我觉得从中学到的个性化教育，远比一个精通历史、凡事都带着历史视角的老师要出色得多，而那个学生可能根本讨厌历史。

And, hey, maybe it is a nightmare of a few I don't I don't know, but I can see the demand for some of these things taking place because I think that the the customized education that you're gonna get out of that is so far superior than some teacher that is an expert in history and everything comes with a history lens and maybe the student that they're teaching hates history.

根本受不了历史，结果一整年你坐在那里，和另外30个学生一起，所有内容都被打上了历史的烙印。

Can't stand history and then everything is flavored with history for an entire year as you're sitting there with 30 other students.

所以我认为，当我们回望你我成长过程中所接受的教育时，会发现它将显得如此过时，我认为无论人们是否喜欢，教育的未来都将朝这个方向发展。

And so I think that when we look back at the education that you and I, you know, grew up with, it's gonna be so archaic to where I think a lot of this is going, whether people like that.

完全同意，而且有句著名的佛教格言是说，你不该用爬树的能力来评判一条鱼，我觉得学校系统对我来说，我百分百就是那条鱼，更像是一块躺在地上的石头，根本不可能爬得上树。

Totally and there's that famous, I think it's the Buddhist quote which is like, you shouldn't judge a fish by its ability to climb a tree and it's just like I think the school system, I was 100% the fish, was craving less so a fish as a rock just laying on the floor, there was no way that I could climb a freaking tree.

所以对我来说，上学时我从没觉得自己属于那里，但一旦离开学校，我找到了有声读物，才发现了自己真正适合的学习方式。

And so I think that school to me, I never felt like I fit in, once I left school and I was able to find audio books, I was able to find my ability to find how I learn.

天啊，那真是意义深远。

Oh my God, it was it was profound.

阿门。

Amen.

所以我认为，我们该如何在个性化学习和人际连接之间找到平衡呢？我认为最大的问题就是这个。

So I think that how do we find this balance between I think the biggest question is it's like, how do we find this balance between, like, personalized learning while also having human connection?

我认为，我们至今还没有找到这种平衡。

And I think I think that's something that we really haven't found that balance yet.

我经常和人们聊天，他们问我：‘你上学时是个什么样的学生？’

I have conversations with people all the time and they ask me, oh, what kind of student were you present?

说实话，塞布，我讨厌学校。

And to be honest with you, Seb, I hated school.

我讨厌它，因为我从没觉得学过任何我真正感兴趣的东西。

I hated it because I didn't feel like I was ever learning something that I actually had an interest in.

我总是被迫接受一些我毫无兴趣的内容，而且这些内容从不以让我感兴趣的方式呈现。

I was always being force fed this stuff that I had no interest in, and it was never framed in a way that interested me.

是啊。

Yeah.

我大学毕业后的经历也是一样。

And same experience after I got out of college.

我只是开始阅读一些我真正想深入了解的东西，然后我就爱上了，因为我专注于自己想学的内容。

I just started reading things that, like, I wanted to know more about, and then I just loved it because I was focusing on things that I wanted to learn about.

而在整个高中时期，我从未有过这样的体验。

And just truly never had that experience the whole time in high school.

我在大学里确实喜欢工程课，但除此之外，那些我不得不修的课程——比如，我大学时还被迫选了一门诗歌课，班上都是些，你知道的，我们当时多大？

And, I mean, I enjoyed the engineering classes I had in college, but beyond that, like all the other stuff that I I literally had to take a poetry class in college with a bunch of, you know, how old were we?

18、19岁的男生们，站在教室里互相朗读莎士比亚。

18, 19 year old dudes all standing in, you know, a classroom reading Shakespeare to each other.

天啊。

Like, good god.

杀了我吧。

Shoot me.

那真是最糟糕的经历，因为我根本不喜欢那种东西。

Like, it was the worst it was the worst experience ever because I I don't like that stuff.

对不起。

I'm sorry.

如果你喜欢这些东西，我很抱歉。

If you love if you love that stuff, I'm sorry.

我不喜欢这些东西。

I don't like that stuff.

而这正是学习这件事的核心——人们终于可以投身于自己真正感兴趣的事物。

And that's that's to the whole point of this learning thing is, like, people can lean into finally lean into things that really interest them.

我简直无法想象，如果从小就开始这样做，并持续二十年，一个人始终被引导去探索他真正感兴趣的主题，而且由世界上最好的老师来教授他所热爱的X和Z领域，会是什么样子。

And, like, I can't even imagine what that's gonna do from if you start that out early and you do it for twenty years where the person is being led down a thing that actually interests them, and you're being taught by the world's greatest teacher on subject x, and z that that person's interested in.

我无法想象，当他们到了二十岁左右时，会取得怎样的成果。

I cannot even imagine what those results would look like by the time they're, you know, 20 years old.

但问题就在这里。

That's the thing though.

只是在努力寻找那个平衡点。

It's just trying to find that balance.

我完全同意，但问题是，我们该如何实现这一点？

I I wholeheartedly agree, and it's just like, how do we achieve that?

我们如何找到这种平衡？

How do we find this balance?

人性的温度与技术。

The human touch and technology.

这是一个具有挑战性的问题。

This is a challenging one.

总会有取舍。

There's always a give and a take.

是的。

Yeah.

我有几个关于人工智能的话题想提一下。

I have a couple AI topics that I wanted to bring up.

第一个，我会在这里的屏幕上展示出来。

The first one, and I'm gonna put up here on the screen.

这个来自著名的AI人士安德烈·卡帕西。

This one's from Andre Kaparthi, the very famous AI.

他曾短暂担任特斯拉的AI负责人，后来成为OpenAI的创始人之一。

He was head of AI at Tesla for a little bit, and then he was one of the founders at OpenAI.

说实话，在我看过的所有教AI的YouTube视频中，他可能是我最欣赏的网络人士。

To be quite honest with you, of the YouTube videos I've watched of people teaching AI, he is probably my favorite of anybody on the Internet.

他是一位非常出色的老师，能把复杂的内容讲得通俗易懂。

He is such a great teacher, and he makes things so accessible.

他发过一条推文，我觉得这条推文非常有趣，值得在这里特别提一下。

And he had this tweet, and I just think that it's kind of really interesting tweet, and I think it's worthy of highlighting here.

他说，不要把大语言模型当作有意识的实体，而应将其视为模拟器。

He says, don't think of LLMs large language models as entities, but as simulators.

例如，在探索一个话题时，不要问：‘你对X、Y和Z有什么看法？’

For example, when exploring a topic, don't ask, what do you think about x, y, and z?

这里没有‘你’。

There's no you.

下次试试问：‘什么样的人群最适合探讨X、Y和Z？’

Next time, try, what would be a good group of people to explore x, y, and z?

他们会怎么说？

What would they say?

大型语言模型可以模拟多种视角，但它并没有像我们习惯的那样，长时间思考过x、y和z并形成自己的观点。

The LLM can channel simulate many perspectives, but it hasn't thought about x, y, and z for a while and over time and formed its own opinions in the way we're used to.

如果你通过使用‘你’来强迫它，它会根据其微调数据的统计特性，采用一个隐含的人格嵌入向量，然后模拟这种人格。

If you force it via the use of you, it will give you something by adopting a personality embedding vector implied by the statistics of its fine tuning data and then simulate that.

这样做没问题，但其神秘感远没有人们天真地向AI提问时所赋予的那么大。

It's fine to do, but there's a lot less mystique to it than I find people naively attribute to asking an AI.

所以他的主要观点是：不要说‘你怎么看’。

So his big point here is don't say what do you think.

他建议人们说：‘谁是最适合对此发表意见的群体？’

He's suggesting people say, who would be the best group of people to have an opinion on this?

然后问AI：‘这个群体的人会怎么想？’

And then ask the AI, what would this group of people think?

我觉得这种方法很有用也很重要，它触及了那些真正编程实现这些系统的人的核心。

I find that to be useful and important, and it gets to the heart of a person who's literally, like, programmed these things.

他正在揭示某种偏见，这种偏见能帮你获得更清晰、更准确的答案，而这正是你所追求的。

He's getting at something that is showing you a bias that will give you a cleaner and more accurate answer that you're going after.

所以我认为这一点对人们来说很重要。

So I think that that's important for people to understand.

对此有什么评论吗？

Any comments there?

如果我们有时间，我们可以讨论一下这个话题，这涉及到我们俩都听过的CEO日记播客，他在Deo中提到的一个观点让我觉得非常有趣，这个观点与我们当前讨论的内容一致：如果我们使用社交媒体，我们的算法会根据我们的兴趣推送信息。

There's so we may talk about this if we have time, and it's around kind of this idea of it was a diary of a CEO podcast that both of us have listened to and something that he mentioned in Deo that I thought was really interesting that kind of is in alignment with this is this idea that if we go on social media, our algorithm is giving us information based upon our interests.

它并不一定提供客观的信息。

It's not necessarily giving us the objective information.

他还讨论了这样一个观点：如果你去问AI，它对某个有争议的话题有什么看法？

And what he also discussed is this idea that if you go and ask AI what does it think about maybe this controversial subject?

根据你的地理位置，如果某个地区的人持有截然不同的观点，AI就会给出那个地区的答案。

Depending on your location, if in one location they have very different views, it's going to give you that answer.

如果你在另一个地区，它就会给出不同的答案。

If you're in another location, it's going to give you a different answer.

所以我觉得有趣的是，你得弄清楚如何提示AI，以获得最客观的答案。

So I think the thing that is interesting is that you've got to figure out how to prompt AI to give you the most objective answer.

这是我个人发现的一点，我很想听听你的看法，因为我一直在使用AI，并且能够给AI设定一个角色。

So this is something that I've personally found and I'm curious to hear your thoughts as I've been using AI, being able to give the AI a persona.

比如说，‘我希望你以一位精通X、Y和Z的专家开发者的身份提供你的观点’，我想听到这样的视角。

So saying like, hey, I want your perspective if you were say an expert developer that has a knowledge in X, Y and Z, I want that perspective.

因此，能够通过特定的视角来看待问题，我认为人们并没有要求AI从特定视角出发分析问题，而AI只会给出它认为你希望听到的答案，从而让我们更依赖AI。

And so being able to look at it through a specific lens and I don't think people ask AI to look at something through a specific lens And with that in mind, it's just gonna give us what it thinks we want to hear so that we're just more attentive to AI.

是的。

Yeah.

是的。

Yeah.

我们稍后在节目中会播放这段采访的几个片段。

And we are gonna play a couple clips from that interview here later on in the show.

但没错，非常精彩的评论。

But, yeah, outstanding comment.

下一个话题有点政治性，我提出来是因为我觉得这个话题挺有意思的。

This next one's a little political, and I'm just I'm bringing it up because I I find it kind of an interesting topic.

总统最近出面，试图剥夺各州的权利。

The president has recently come out and he's trying to take away the state's rights.

我们美国有50个不同的州，每个州似乎都在制定自己的人工智能法律和规定。

So we have 50 different states here in The US and they all seem to be coming up with their own AI laws and rules.

而总统现在打算本周发布一项行政命令，强制所有州服从联邦层面的人工智能法规，以统一我们国家在这方面的做法。

And the president is now trying to say that there's going to be an executive order on AI this week that forces all of the states to fall under the federal level AI mandate of how we're going to go about this as a country.

作为一个坚定支持州权、主张尽可能将责任从联邦政府下放至州层级的人，这一点让我很痛心。

And as a person that firmly believes in states' rights and pushing responsibility as much out of the federal government and down to the state level as possible, this one pains me.

但与此同时，我也在思考，我理解这种逻辑：我们正处在一场全球竞赛中，面对的是其他超级大国的快速推进，不可能让一个州占优势而另一个州处于劣势。

But at the same time, I'm also looking at it and I understand the logic that we're in a global race and you're up against some other superpowers that are moving out and not in a situation that one state is advantageous, the other one is not.

于是大型科技公司会选择在更有利的州设立据点，但它们也担心当地政局的变动和政策的不稳定性。

So then the big tech companies set up shop and the one that's advantageous, but they also are concerned about the whims of that local government shifting and changing.

因此，它们不得不将整个基础设施迁移到另一个现在看起来对商业更有利的州。

And then they have to move all of this infrastructure to another state that now looks like it's more advantageous for doing business there.

我认为，我猜测正在发生的是，总统正受到大型科技公司的压力，要求他们出台一项统一的政策。

And I think what I suspect what's happening is the president is getting pressured by the big tech companies to have something that's unilateral across the board.

他们再也不用担心，在某个州XYZ被‘抽走地毯’，从而不得不将巨额资本支出转移到另一个州，还要承担成本、开支，最重要的是，转换到另一个州所需的时间。

They never have to think about, my words, getting rug pulled in state XYZ, so that they have to then move all of this huge amount of CapEx into another state and the cost and expense and most importantly, the time to make that conversion over to a different state.

我认为，他们正在游说总统这么做。

I think what they're doing is they're lobbying him to do this.

我想知道你对这些有什么看法。

I'm curious what you think about some of this.

我知道这是个关于美国和美国政治的话题，但我不确定。

I know this is a US topic, except US politics, but I don't know.

我能理解科技公司为什么这么做。

I can understand why the tech companies are doing this.

他们只是想尽量降低对自身的风险。

They just were trying to minimize risk to them.

我怀疑正是他们在推动这件事，但如果你有什么评论的话。

I suspect they're the ones that are pushing on this, but if you have any comments.\

这项监管措施让我非常纠结，我觉得他们正在陷入比特币的迷宫。

This regulation is something that I'm so torn on and I think they're going down the Bitcoin rabbit hole.

我认为很多比特币支持者都处于这种立场：算了，干脆去监管化吧。

I think a lot of Bitcoiners find themselves in this position where they're like, you know what, let's deregulate.

我们相信自由市场最清楚什么是最好的，如果你希望迈向一个有增长的社会，就不该阻碍信息进入自由市场。

We believe that the free market knows what is best and ultimately if you want to move towards a society with growth, you don't want to impede that information making it to the free market.

而我真正纠结的地方在于，我不认为这件事非黑即白。

And I think the thing that I really struggle with is I don't necessarily think that it is black and white.

不是完全无监管，就是全面监管，而是介于两者之间。

It's either no regulation or regulation, it's somewhere in between.

对我来说，一个完美的例子是：有些公司可以利用环境破坏来获利，却不必承担相应的成本。

Like a perfect kind of example to me is something around the lines of you can have a corporation that can be capitalizing off the destruction of the environment and they don't have to front that cost.

那么，监管机构是否应该介入，要求那些从自然环境中获利的企业承担相应的经济负担？毕竟他们通过破坏森林、开采矿物和提取其他自然资源来盈利。

And so then does a regulator have to put something in place where there is going to be a financial burden for then capitalising off the natural environment because they're making profit from destroying forest, extracting minerals, extracting XYZ from our natural environment.

因此，我认为这些公司的根本目标显然是利润，如果没有监管，他们就会继续肆无忌惮地从我们的自然环境中攫取资源。

And so I think that a lot of these corporations, their bottom line is obviously profit and there's no regulation, they're going to complete and they're going to continue to extract from our natural environment.

所以我认为，对于许多这些AI实体来说，情况也是如此。

So I think the same thing is true for a lot of these AI entities.

如果它们的目标是打造最快学习、最强大的AI引擎，即使存在毁灭人类的风险，我们也会继续这样做。

If their goal is how do we build the fastest learning, most powerful AI engine, even if there is a risk of destroying the human race, we're just going to keep doing it.

所以问题是，是否需要建立监管机制，因为自由市场没有能力对这样的事情进行抵制？

So it's just like, does there need to be regulation in place because the free market doesn't have the capacity to push back on something like that?

我不确定，我没有完全成型的观点，因为我能从两个方面来看待这个问题。

I'm not sure, I don't have a fully formed opinion on it because I can see it from both sides.

我不知道你的想法是什么。

I don't know what your thoughts are

那里。

there.

是的，从州权的角度来看，普雷斯顿，其中一个优势是它在各州之间创造了竞争。我们先不谈AI，就拿石油来说。

Yeah, from a state's rights standpoint, Preston one of the advantages is it creates competition between the states in that if And let's not use AI, let's just use oil.

假设你是一家炼油厂。

Let's say you're an oil refinery.

假设某个州的公民非常反感，因为这对环境造成了影响，或者从审美角度来说也不行，总之无所谓。

Let's say that citizens of one state just really don't like it because of what it does to the environment, what it does just from a aesthetic standpoint, whatever, doesn't matter.

明白吗？

Okay?

而另一个州却说：哦，我们不在乎。

And then another state is like, Oh no, we don't care.

我们希望所有这些产业都到我们这里来。

We want all that business to come in here.

我们希望所有的商业活动都来。

We want all that commerce.

所以我们将会对这个行业实行支持监管。

So we're going to be pro regulation around that particular industry.

所以，让我们来举个非常假设性的例子，只是为了说明一下利弊。

So let's and this is very hypothetical, just kind of illustrate the pros and cons.

那个允许这种情况发生并广泛蔓延的公司或州。

The company or the state that allows this to come in and just proliferate everywhere.

我们就说这有点过头了。

Let's just say that it's over the top.

而且这是世界上最丑陋、或者在美国最丑陋的居住地之一，而另一个限制更严格的州则要美丽得多。

And it's just one of the ugliest states in the world or in The United States to live, where the other one that was more restrictive is a much more beautiful state.

如果你是那种不想看到这些东西的人，那你就会搬走，用脚投票，搬到更符合你需求的另一个州。

If you're the type of person that doesn't want to be looking at that kind of stuff, well, then you'd move, you'd vote with your feet and move to the other state in the one that's more desirable for you.

我们假设那个支持石油的州有很多税收优惠。

Let's say there's a bunch of tax advantages in the one that that is pro oil.

对吧？

Right?

你不想多花钱，所以你会搬到那个州。

You don't wanna be paying more money than you have to, so you move into that state.

在这种情况下，你实际上是在创造一种竞争，促使人们迁移到最符合他们生活生态系统需求的州。

So in that scenario, you're creating competition for people to migrate to the state that aligns with what it is they want out of their ecosystem that they're living in the most.

当谈到人工智能时，这里有一点不同，那就是正在开发的产品。

When it comes to AI, where this is a little bit different is the product that's being built here.

我不知道这对一个州相比另一个州是利大于弊还是弊大于利，这正是为什么说这不同——因为你实际上是在构建智能。

I don't know that it has a benefit or a negative for that state over another one would be kind of the argument of why, hey, this is different because you're literally building intelligence.

我认为可以从数据中心的角度，即数据的规模、数据中心的占地面积及其能耗来论证，这些因素在各州之间会引发担忧。

I think you could make the argument from a data center standpoint in the footprint that the data and that that would be the argument, is the data center and the footprint and the size of that footprint and the energy consumption of that footprint would be concerning from state to state.

有些人可能出于各种原因不希望自己的州有这么多数据中心。

Some people might not want all of those data centers in their state for whatever reason.

但也有些人会欢迎，因为为了服务这些数据中心，将建设大量的能源基础设施。

Some might love it because of the energy infrastructure that's going to be built out to service it all.

所以这是一个很难的问题。

So those are it's a hard one.

确实是个难题。

It's hard one.

这个观点很有趣，因为我认为物理基础设施不应该由联邦层面来监管。

That's interesting point because I think the physical infrastructure, I don't believe that should be regulated on a federal level.

就像如果一个州想创造更多就业机会，想吸引更多数据中心，那当然应该是他们的自由选择。

It's just like if a state wants to open up jobs, they want to have more data centers, for sure by all means that's their choice.

然而，当一项技术的益处或负面影响不仅波及该州，还可能影响整个人类或国家时，突然间，一个州所做的决定可能影响的范围远超其本地范围。

However, when the benefit or the negative effects of a technology not only impact that state but they potentially impact humanity or the nation, all of a sudden it's just like a state could be making a decision that impacts far more than its locality.

这正是我觉得非常、非常有趣的地方。

That's the thing that I find really, really interesting.

那么，我们在什么时候需要某些监管，因为人们所做的决定在更大范围内造成了更严重的损害？

So at what point do we need certain regulation because people are making decisions that are far more detrimental on a bigger scale.

然后问题来了：谁在制定监管？我认为这触及了监管的核心——即使我们能实现完美的监管，也必须问清楚：究竟是谁在制定监管？谁来监管监管者，以确保监管的公正性？资金又从何而来以支持这些监管？

And then the question is, who is and this is I think what gets to the heart of regulation it's just like cool, okay this is great if we can have perfect regulation but you've got to ask who is actually creating the regulation and who is regulating the regulators to ensure the regulators are being fair and where's the money coming from to help fund this regulation?

很多时候，比如你看制药行业，负责监管制药的机构的资金其实来自大型制药公司。

Many times like if you look at, I don't know, the pharmaceutical industry, the pharmaceutical regulating agencies are funded by big pharma.

所以这就出现了利益冲突，因此情况变得极其复杂。

So it's just like there's these conflicts of interest, so that's where it gets so convoluted.

而且我们稍后在节目中还会讨论特里斯坦·哈里斯的对话，如果你对监管政策这个方向感兴趣，或者喜欢这个话题，我强烈建议你听完整个访谈，因为其中对是否应该监管这个问题的探讨非常深入。

And again, we're going to talk about this Tristan Harris discussion later in the show, and if you want to go down the regulatory policy path and if you like this particular topic, then I would highly encourage you to listen to that entire interview because it gets very heavy in this in this domain as to whether you should or shouldn't.

被采访者斯蒂芬·巴雷特的观点带有很强的偏见。

It's a very biased point of view by the person being interviewed, which is, Stephen Barlett.

对不起。

I'm sorry.

史蒂文·巴莱特正在采访特里斯坦·哈里斯。

Steven Barlette's doing the interview of Tristan Harris.

特里斯坦显然对此有非常强烈的观点。

Tristan obviously has a a very strong opinion in that.

但我认为，让人们听听这种相反的观点或论点，无论他们是否认同，都是有好处的。

But I think that it's good just for a person to kind of hear that counterargument or that argument and whether it you know, something that they agree with or not.

那么，我们来看一下我关于AI话题的最后一个内容。

So let's go to the last one that I have on just the AI topics here.

我要把这个放到屏幕上。

I'm gonna put this up on the screen.

这很有趣，因为我们正在讨论谷歌的泰坦。

And this is interesting because we're talking about Google's Titans.

这个人在这里发了这条帖子。

This person had this post here.

那么，这到底是什么？

And so, like, what is this?

Titans 是谷歌推出的一种新架构，它能让语言模型在运行时拥有类似长期记忆的能力。

Titans is a Google's new architecture type that gives a language model something like a real long term memory while the model is running.

他们在这里展示了一张图表，显示该模型能处理一千万个标记，并仍保持约70%的准确率，这简直不可思议。

And so they have a chart up here and they're showing how it's ingesting 10,000,000 tokens and it's still maintaining around 70% accuracy, which I guess is insane.

它还与其他一些模型在处理一千万个标记时的表现进行了对比，但这些模型的表现远不及谷歌在Titans上取得的成果。

It has some of the other models and what they do with 10,000,000 tokens and it's nowhere close to what Google has uncovered here with titans.

我觉得这个点很有趣，因为我经常想把很长的文档输入到一个超长上下文窗口中，希望它仍能正常执行，但结果总是变得很卡顿。

This one I found interesting because I find myself wanting to just put really large documents into a really long context window asking it to still perform, and it gets laggy.

显然，在长期记忆方面，目前还存在某种缺失。

And clearly, there's something missing when it comes to long term memory.

这篇论文发表于2025年。

So this paper came out in 2025.

它并不是上周发布的，但确实是在今年发布的。

It didn't come out this past week, but it did come out this year.

这篇论文的标题或副标题是‘在测试时学习记忆’。

And the title or the subtitle of the paper is learning to memorize at test time.

你在使用这个系统时输入的某些token大小，我曾试着去理解它是如何工作的。

And some of the token sizes that you're putting through this thing, I was playing around with it trying to understand how it works.

我这里有一段话，就直接念给大家听，让大家思考一下它是如何实现的。

I got this paragraph that I'm just going to read for folks to kind of think through how it's doing this.

想象一下，你正在浏览社交媒体动态，大多数帖子通常是表情包或朋友的午餐照片。

Imagine you're scrolling through your social media feed and most posts are usually stuff like memes or friends' lunch pics.

你的大脑，就像AI一样，预期这些无关内容，并大多会忽略或遗忘它们以节省空间。

Your brain, like the AI, expects that junk and mostly ignores or forgets it to to save space.

但突然间，出现了一条关于你最爱乐队的惊喜演唱会门票抽奖信息。

But suddenly, there's a post about a surprise concert ticket giveaway for your favorite band.

这个惊喜吸引了你的注意力，因此你会记住细节，比如报名截止时间和规则，同时让那些无聊的帖子逐渐淡出。

That surprise grabs your attention, so you remember the details like the entry deadline and rules while letting the boring post fade away.

在Titans中，AI利用基于梯度的计算所产生的‘惊喜信号’，从海量数据流中识别出意外或重要的信息，并决定将这些有用的信息存入长期记忆。

In titans, the AI uses a similar surprise signal from calculations based on gradients to spot unexpected or important info in the huge stream of data deciding to store that useful bit in its long term memory.

这样，它就能保留对后续任务（如回答问题）有价值的信息，而不会被无关的重复内容占据空间。

This way, it keeps what's valuable for later tasks like answering questions without cluttering up the irrelevant repeats.

所以我对这个问题的直接跟进是：好吧。

So my immediate follow on question for that was, okay.

那么，AI 是如何知道那个演唱会门票或那个特定乐队在信息流中是令人意外的呢？

So then how does the AI know that that concert ticket or that particular band is a surprise in the feed?

我得到的回答是，它只是在查看它所训练的海量数据，也就是整个互联网。

And what I got back was it's just looking at the sheer amount of training data that it was trained on, which is the whole internet.

当它处理包含数百万个标记的文档时，它会发现与它所训练的整个数据集相比独一无二的内容。

As it's going through said document that has these millions and millions of tokens in size, it is finding something that is unique in reference to the entire dataset that it was trained on.

而这个梯度正是让它能够说：‘哦，这和我预测或预期的不一样，所以我要记住它。’

And that gradient is what's allowing it to say, oh, that was different than what I would have predicted or expected, so then it remembers it.

我觉得特别有趣的是，多年前我采访过撰写克劳德·香农传记的作者。

And what I find so fascinating so, years ago, I interviewed the author that covered Claude Shannon.

当我们讨论信息论时，我清楚地记得他说：‘普雷斯顿，信息就是惊喜。’

And when we were covering information theory, I distinctly remember him saying, Preston is just surprise.

他的算法，他的数学算法，只是在寻找惊喜。

Like, his algorithm, his mathematical algorithm is just looking for surprise.

当它处理这个‘泰坦’内容时，我心想，哇，这简直就是信息论在实现长期记忆方面的应用。

And when it was going through this titans thing, I was just like, wow, this is literally just like information theory in order to do long term memory.

这太令人震撼了，对吧？

It's just mind blowing, right?

太迷人了，老兄。

It's fascinating, dude.

那是什么？

What was it?

信息论里有个词，比如我给你寄一封信，信里会有很多噪音，而你最终要找的是惊喜，是新的信息。这有点像是信息的反面——信息就是噪音，我们寻找的是噪音中的信号。就像你有一块大理石，要剔除所有多余的部分，才能找到里面隐藏的雕像。我一时想不起那个词了，但当你说到这个时，我想起我最喜爱的一场TED演讲之一，我想你也读过那本书，是唐纳德·霍夫曼做的，他讨论的是：我们看到的是现实本身吗？

There's that word in information theory which is if I send a letter to you there's going to be a lot of noise in that letter ultimately you're looking for the surprise the new piece of information and it was kind of like the inverse of information information is noise what we're looking for is the signal in that noise and so it's kind of like if you've got a block of marble you're cutting away all that excess to find the statue inside and I've got blanked on that but what comes to mind as you're talking about this is there's one of my favorite TED talks we've ever listened to, and I think you've read the book, is by a guy called Donald Hoffman and he talks about do we see reality as it is?

对。

Yeah.

他打了个比方，我强烈推荐大家去听一听，只要搜索TED演讲‘唐纳德·霍夫曼’，‘我们看到的是现实吗？’

And he gives the analogy and I highly recommend anyone going out and listening to this, just type in TED talk Donald Hoffman, do we see reality So as it good.

他提到澳大利亚有一种叫末日甲虫的生物，这种甲虫已经存在了大约三亿年。

And he mentions there's this it was like a doom beetle in Australia and this doom beetle has been around for something like three hundred million years.

所以你可能会认为，既然它已经存在了三亿年，那它肯定能看到现实的本来面目。

So you would assume if it's been around three hundred million years, of course it sees reality as it is.

但后来澳大利亚人开始把棕色短颈啤酒瓶扔进沙漠，突然间，这种末日甲虫几乎濒临灭绝，因为它在大脑中形成了一条启发式规则：越大、越棕，就越好。

But then the Australians started throwing these beer bottles, these brown stubby beer bottles into the desert, and all of a sudden this doom beetle nearly went extinct because it had effectively created a hack, a rule of thumb in its brain that is, hey, the bigger, the browner, the better.

它会去试图和这些棕色啤酒瓶交配，差点彻底灭绝。

And it just goes and tries to mate with these brown beer bottles, nearly went completely extinct.

因此，澳大利亚政府不得不介入，禁止使用啤酒瓶，之后这种甲虫才开始逐渐恢复。

So the Australian government had to step in, ban beer bottles, and all of a sudden that beetle started to kind of recover.

但我觉得这个演讲真正有趣的地方，以及他想表达的观点是：我们并不是为了看清现实本身而进化的，而是为了适者生存。

But what I find really interesting about this talk and what he's trying to get at is this idea that we are not optimised to see reality as it is, we're actually optimised for survival of the fittest.

因此，我们接收所有这些信息，剔除所有噪音，试图建立这些经验法则，以最大限度地提高我们的生存几率。

And so what we do is we take in all this information, we discard all the noise and we try and create these rules of thumb to basically maximize our chances of survival.

所以，当一列火车朝我们驶来时，我们并不会去处理所有信息，想：这到底是什么东西？

So if there's a train coming towards us, we're not processing all this information being like, what is this thing?

我想知道它有多快？

I wonder how fast it is going?

它以多快的速度撞上我的身体？

At what velocity does it going to hit my body?

我们忽略了所有这些数据，只是简单地想：火车朝我冲来了，我得赶紧躲开。

And we ignore all that data and we're just like, trains coming towards me, I need to get out of the way.

我们建立了一条经验法则，来识别这是危险的。

We've created a rule of thumb to recognize that this is dangerous.

我之所以说这些，是因为AI，什么会让AI感到惊讶？

Now the reason why I say all of this is because AI, what is surprise to the AI?

当它处理一千万个标记时，它如何知道哪些信息是有价值的？

What information when it's looking at 10,000,000 tokens, how does it know what is valuable?

它究竟在优化什么？

What is it actually optimizing for?

因为我们优化的是适者生存，那它又在优化什么？

Because we're optimizing for survival of the fittest, what is it optimizing for?

因此，我很好奇能更深入地研究这些模型，弄清楚它们究竟在试图提取什么信息——要么是开发者必须手动编码进去，告诉我们哪些类型的信息才是我们真正关注的；要么是模型自己发现的，而这些发现真的相关吗？

And so that's where I think I'm curious to dig deeper into these models and try and understand like what are they trying to pull out because either a developer has to code that in, this is the type of information that we're actually looking for or it's figuring that out itself and is that actually relevant?

我很想听听你对此的看法或想法。

I'm curious to hear your take or your thoughts on that.

我不确定自己是否充分了解这些梯度是如何确定的。

I don't know that I have enough information on how these gradients are determined.

我知道我见过埃隆·马斯克和其他人发的帖子，他们强烈指出，使用整个互联网、你能获取到的每一条数据作为训练集，并不一定能为任务X、Y、Z训练出最好的模型。

I do know that I've seen posts by Elon Musk and others that really speak to the idea that the training dataset, like using the entire internet, every single thing that you can get, isn't necessarily leading to the best model for task X, Y, and Z.

它能训练好吗？

Is it going to train?

它在理解英语或语言方面会表现得非常好吗？

Is it going to do really well for understanding English or the language?

会的。

Yes.

但如果你的目标是理解医学讨论，它会是最优的吗？

Is it going to be most optimal if you're trying to understand a medical discussion?

不。

No.

所以你需要深入思考，好吧，我们有一个完全或近乎完全理解语言的AI。

So then you got to get into, okay, so we have an AI that understands the language perfectly or near perfectly.

现在我们要将它应用到另一个模型上，这个模型更专注于我们输入的训练数据。

And now we're going to apply that to a different model that then is more focused on the training set as to what we're putting in it.

但所有这些都远远超出了我的理解范围，因为我对此所做的研究太少。

But all of that is way outside of my depth of understanding as far as how much research I've done on it.

我看到过很多该领域重量级人物的讨论，他们谈论数据集的筛选，确保你不会把所有东西都塞进去，因为这可能会带来灾难性后果，或者在特定任务中竞争会变得激烈。

Have seen a lot of conversations by heavy hitters in the space talking about curation of the dataset and making sure that you don't just put everything in there because it's somewhat disastrous or how competitive it'll be in particular tasks.

我认为这是一个非常重要的观点，即最终仍需由个人来判断什么是有价值的，什么是没有价值的。

I think that this is really an important point, which is that I think it still falls on the individual to determine what is valuable and what is not.

我认为你可能会看到一项庞大而庞大的研究或论文，它输出大量信息，声称这些是有效的、有效的、有效的。

I think that you can have this huge, huge research study or research paper that you're looking at and it's spitting out a whole bunch of information saying this is valid, this is valid, this is valid.

但对于普通人来说，也许会觉得：酷，这太棒了。

And to the average individual maybe like cool, this is rad.

但对于一位了解该领域的科研人员来说，他仍然能够再次筛选，区分出有用的信息与干扰信息。

But to a scientific researcher who understands the subject he's able to again filter through that again and separate out the signal from the noise.

我认为，正是在这一点上，人类在短期内不会被取代，因为各领域的专家能够辨别什么是有价值的，什么是没有价值的。

I think that's where ultimately humans aren't going to be replaced anytime soon because specialists in their field are able to see what is valuable and what is not.

就像我们比特币爱好者，当看到报纸文章或《纽约时报》的帖子说‘比特币的耗电量超过了阿根廷’时，我们立刻就能明白。

It's just how, as Bitcoiners, we can see when there's a newspaper article or a New York Times post or whatever that says, oh man, Bitcoin is consuming more energy than Argentina.

这明显具有误导性，并不准确。

It's just like, well, that's misleading and that's not necessarily accurate.

因此，我认为你仍然需要成为某一领域的专家，并深入理解该主题，才能判断这些模型输出结果的有效性。

So I think that you still need to be a specialist and deeply understand the topic to understand the validity of the output from a lot of these models.

是的。

Yeah.

塞布，我们进入本节的最后一个话题吧，时间有点超了。

Seb, let's go to our last topic for this show, and we're running a little long.

所以我想我们得安排另一期节目，把其他话题放进去，因为我甚至觉得我们可能都来不及讨论特里斯坦·哈里斯这个话题了。

So I think we're gonna have to schedule another one to put some other topics in there because I don't even think we're gonna get to this Tristan Harris discussion.

但你本来想聊聊远距离触觉反馈。

But you wanted to talk about long distance haptic touch.

那你来主导这个话题吧。

Go ahead and take this topic away.

当然。

Totally.

所以这件事啊，天哪。

So this is something that oh, man.

它让我大吃一惊。

It blew me away.

我甚至都不知道这东西存在。

I didn't even know this existed.

基本上，我偶然看到了马里奥·诺尔费尔特的这篇帖子，我不知道能不能打开帖子给你们看看。

Essentially, I stumbled upon this post by Mario Norfelt, I wonder if I can even open up the post and I'll show you guys.

前几天我刷推特的时候，偶然看到了马里奥的这篇帖子，科学家们实际上已经研发出了这种远距离触觉反馈技术。

So I was kind of scrolling Twitter the other day and I stumbled upon this post by Mario and basically what scientists have kind of created are these long distance haptic touch.

对于那些不熟悉‘触觉’这个词的人，简单来说，就是如何创建柔性贴片，将触觉引入虚拟现实、增强现实等领域。

So for those that aren familiar with this word haptic, s basically it s like how do we create flexible patches that bring touch into virtual reality, augmented reality and such.

这篇帖子中提供了一些信息，但我进一步深入挖掘，找到了这项研究的原始科学论文，它深入探讨了这一主题。

And so it gives a little bit of information in this post but I started to dig a little deeper and I found the original scientific study which kind of dove into this.

你可以在这里找到这项真正的科学研究，它被称为‘用于多功能增强触觉交互的皮肤附着式触觉贴片’。

The actual scientific study you can find here and it was basically called skin attached haptic patch for versatile and augmented tactile interaction.

因此，摘要中提到的第一部分是：可穿戴的触觉界面可以通过在皮肤上添加触觉模拟，与传递给用户的视觉和听觉信息相结合，从而增强沉浸式体验和虚拟增强现实系统。

And so this is basically the abstract says the first little bit, it's like wearable tactile interfaces that can enhance immersive experiences and virtual augmented reality systems by adding tactile simulation to the skin along with visual and auditory information delivered to the user.

对我来说，最让我感到非常非常着迷的是这些小小的触觉贴片。

And so to me what I found really, really fascinating about this is these little haptic touch patches.

如果我没记错的话，它们的尺寸约为1.1毫米，但体积虽小却功能强大，能够产生压力和高频振动。

Essentially they are, if I remember correctly, they're 1.1 millimeters in size and they're extremely powerful for their size and able to create both pressure and high frequency vibrations.

所以，如果你戴上一副装有这些触觉贴片的手套，而另一个人也戴着同样的手套，并触摸了纹理、形状、边缘、字母或三维表面，你就能通过这些触觉传感器感受到他们所感受到的一切。

So if you were to wear a glove with these haptic touches on your fingers, would be able to if someone else was say wearing the same glove and they went and touched textures, shapes, edges, letters, three d surfaces, you could feel what they are feeling through these haptic touch sensors.

那么，这有什么作用呢？

And so what does that do?

我觉得我们可以在很多不同的方面使用这项技术。

I think we could use this in so many different ways.

你可以通过Zoom与某人互动，比如现在我们正在交谈，如果我们想要拥抱或以涉及触觉的方式交流，我就能感受到你的感受，从而获得一种更三维、更细致的互动体验；同时，我们也可以从虚拟现实和增强现实的角度来看，你可以穿戴某些设备，比如手套或全身服，并戴上VR眼镜。

Either you could basically have an interaction with someone, let's say through Zoom, you and I are talking right now and if we went to hug or we went to communicate in a way that we wanted to involve touch, I could feel what you were feeling and have a much more like three-dimensional detailed interaction, but we could also look at it from the perspective of like virtual reality and augmented reality where you'll be able to wear certain things, whether it's gloves or a suit, and you'd be able to wear VR goggles.

你将置身于一个模拟环境中，不仅能通过视觉，还能通过物理触觉与这个模拟世界互动。

You'd be inside a simulation and you could be interacting with the simulation not just from the visual sense but also from the physical touch sense.

对我来说，这简直令人难以置信。

And that to me, I think, is is mind blowing.

所以我很想听听你对这种触觉传感器的看法。

So I'm curious to hear your thoughts on kind of this haptic touch sensor.

是的。

Yeah.

当塞布在说话的时候，对于只听音频的观众，我放了一段名为Fluid Reality的公司的视频。

So while Seb was talking there, for people that are just listening to the audio, I put up a a video of a company that it's called Fluid Reality.

这家公司正在研发这些触觉传感器。

And the company is putting these haptic sensors.

这家公司在使用这项技术时，旨在让人形机器人更好地感知它们所接触的物体，从而判断在与物体互动时，是该用力捏苹果还是轻柔地拿取。

The way that they're using that particular company is doing is they're trying to give the humanoid robots better sensing of how they're feeling different objects and whether they should be squeezing an apple very hard or very lightly as they're interacting with it.

现在你们在屏幕上可以看到，如果正在观看这段视频，会看到一个人正戴着一副手套，训练具有塞布所描述感知能力的人形手部。

Then you can see up on the screen right now, if you're seeing the video of a person that's wearing a glove that's training one of these humanoid hands with the sensing capability that Seb was describing.

当我们仅关注机器人执行某些任务的能力时，比如洗衣或其他工作，你很快就会意识到，机器人在与物体互动时施加的压力至关重要。

And when we just look at the robot's ability to do certain tasks, like let's say it's doing laundry or it's doing whatever, you can very quickly realize that the amount of pressure that it's applying to the objects that it's interacting with become really important.

这不仅关系到功耗管理，也关系到避免捏碎或损坏所持物体。

It becomes important from a power management standpoint, from just not breaking the object that it's holding or damaging it.

我认为，触觉技术是人形机器人发展中至关重要的一部分，也是未来发展的重点方向。

And I think that haptics are a huge part of humanoid robots and where a lot of that's gonna be going.

这绝对是值得我们关注的技术。

And this is something that is definitely worth paying attention to.

我这里还有几个视频，现在就快速放出来给大家看看。

And I have a couple more videos here I'm just gonna quickly put up on the screen for people to see.

这是特斯拉机器人，对于只听音频的观众，我来展示一下它的手部动作和机械结构有多么精准。

And this is the Tesla robot for people that are just on the audio showing you how crazy accurate the hand gestures and the mechanics in the hands are.

我不知道这个特定的人形机器人在触觉压力反馈方面进展到什么程度了。

And I don't know where they're at from a haptic pressure feedback standpoint on this particular humanoid robot.

但从手势来看，它似乎内置了很高的灵敏度。

But by the looks of the hand gestures, it looks like it has a lot of sensitivity built into it.

但我

But I

我认为，据我了解，触觉感知是机器人领域目前正努力克服的最大挑战之一。

think this is one of from my understanding, haptic touch is one of the biggest hurdles I think robotics is trying to overcome right now.

没错。

Yeah.

因为让一个像叉车那样四处移动、拾取集装箱、放下集装箱的机器人工作是一回事，

Because it's one thing to have a robot that is a forklift truck that is just going around doing its own thing, picking up containers, moving those containers, dropping those containers.

但让一个在厨房里操作的机器人能够拿起鸡蛋而不压碎它，或者能够进行更精细的压力敏感操作，比如使用螺丝刀并感知何时开始拧坏螺丝，或者进行手术、缝纫、烹饪等，这完全是另一回事——我认为，这些任务所需的精细运动技能远超我们人类通常所意识到的。

But it's another thing to have a robot that say operates in the kitchen and can pick up an egg without crushing it or to be able to do more pressure sensitive things like use a screwdriver and understand when it's starting to strip the screw or starting to do things like surgery or sewing and cooking and things where ultimately I think there's a lot more fine motor skills than we really take into account as humans.

我们的双手通过触觉感知提供了大量信息，我们基于这些信息自主做出决策，而将这种能力引入机器人领域，我认为我们还有很长的路要走。

We are being given so much information through our hands, through our sensory touch, and making decisions on that information that we're doing autonomously and being able to bring that into the robotic world, I think we're still a little ways away.

据我了解，这些触觉传感器一套手部设备的价格可能在1万到5万美元之间。

Because from my understanding a lot of these haptic touch sensors can cost anywhere from like 10,000 to $50,000 for a set of hands that are able to do things like this.

因此，我们离让每个家庭都能用上这种技术还很遥远。

So we're a long way off having this available to everybody in their household.

但正如你之前提到的，如果我们能实现十倍的性能提升，成本降低90%，突然之间只需几千美元就能做到，那就会容易得多。

But to your point earlier, if we get a 10x improvement, 90% reduction in price and all of a sudden you can be doing this for a few thousand dollars, I think it's going to be far, far easier.

但话说回来，我觉得无论是从机器人技术的角度，还是从个人体验的角度，这都极其迷人。

But again, I just think it's so fascinating both on the robotic side of things and on the individual side of things.

所以我想提一件事，想听听你的看法，我们是什么时候来着？

And so one thing that I just wanted to kind of bring up and I'm curious to hear your thoughts, we had a when was it?

三月份我们一起去杰克逊霍尔滑雪时，曾聊过一个话题，不知道你还记不记得，关于我们是否活在模拟世界中。

In March we were skiing in Jackson Hole together and we ended up having a conversation, don't know if you remember, on are we in a simulation?

当时我们坐在桌边，讨论了我们是否活在模拟世界中，以及AI和当今一些机器人技术是被创造出来的，还是被重新发现的——即它们其实早已存在？

We were kind of sat at a table and we had this conversation of are we in a simulation and this idea of is AI and say some of the robotics we're using today, is it being created or is it being rediscovered as in it already exists?

因此，当我想到触觉技术时，我会联想到电影《阿凡达》，想到作为人类，我们渴望融入社群。

And so when I think about haptic touch what comes to mind is the movie Avatar and I think about as humans we want to feel a part of community.

我们想要创造价值。

We want to create value.

我们希望感受到自己有目标。

We want to feel like we've got purpose.

那么我们该怎么做呢？

And so what do we do?

我们开始创造产品。

We start to create products.

我们开始创造技术。

We start to create technology.

我们开始建立公司。

We start to kind of create companies.

我们走进这个世界。

And we go out into this world.

作为回应，我们创造了生产力和效率的提升，并开始开发能够取代我们的技术。

In response, we create advancements in productivity and efficiency and such and we start creating technology that starts to replace us.

所以，当我们开始取代自己时，突然间就失去了那种意义感。

So all of a sudden when we start to replace ourselves, we're losing that sense of purpose.

我们失去了那种能感受到自己创造价值的社区归属感。

We're losing that sense of community where we feel like we're creating value.

因此，当社会正在衰退，突然间抑郁、自杀、药物滥用等问题激增时，作为社区，我们该怎么做？

And so at that point, if society is degrading and all of a sudden there's rising rates of depression and suicide and substance abuse and all of these things, what do we do as a community?

如果技术足够先进，你可以创建一个像《阿凡达》那样带有触觉反馈的模拟世界，让我们能够进入一个更简单、更原始的时代，重新开始创造价值，然后一遍又一遍地重复这个过程——我们究竟多少次已经踏入过这种潜在的模拟世界？

Well if technology is advanced enough you could create a simulation with haptic touch like Avatar where we're able to step into a world and move back to a much simpler time and back to that simpler time where we can start to create value again and then we go and do it again and again and it's like how many times have we stepped into a potential simulation.

现在，我并不是说这就是我相信的观点，但我认为这是一个有趣的想法，因为我们正处在一个节点上：在未来三到四十年内，如果我们能创造出令人难以置信的真实模拟世界，以至于无法区分物理世界和这些模拟世界，我一点都不会感到惊讶。

Now I'm not necessarily saying this is what I believe but I think it's an interesting thought because we are at a point where it wouldn't surprise me if in the next thirty-forty years we're able to create simulations that feel unbelievably realistic and we can't differentiate between the physical world and more of these simulations.

当提到模拟理论时，埃隆·马斯克最常提到的观点之一，正是你所描述的，塞布。

It's one of Elon Musk's biggest talking points when this topic of simulation theory comes up is exactly what you described, Seb.

但我们先把这个思想实验留给你。

But we'll leave you with that thought experiment.

我不知道。

I don't know.

对吧？

Right?

我没有明确的观点，但我觉得这是一个非常有趣的思想实验，很有趣去慢慢推敲。

I I don't have an opinion, but I do find it to be a fascinating thought experiment and fun to kind of just tease out.

因为速度——我们怎么知道呢？

Because the pace how would we know?

速度？另外一件我觉得很有趣的事是，这些AI环境都是临时拼凑出来的。

The pace Well, and the other thing that I that I find fascinating are all these AI environments that are just kind of ad hoc making up.

比如他们展示了这些环境的视频。

Like they're showing videos of these things.

我不确定我们之前有没有在播客里放过，但也许早前放过。

I don't know if we've ever put it on the screen of one of the podcasts, but maybe we did earlier.

但它们只是在构建这样一个环境，你看着一段视频，里面有人走在街上，看起来完全真实。

But it's just making up this environment and you're watching this video of somebody walking down a street and it looks completely real.

这完全是AI实时生成的。

It's just being made up on the fly by AI.

然后还有一个记忆组件，如果你转过身，几分钟前你看到一棵树，当你转身朝相反方向走时，那棵树在一段时间内仍然会在那里。

And then there's a bit of a memory component that if you turn around and that you saw a tree just a couple minutes ago, if you turn around and start walking back the other direction, the tree will still be there for a certain amount of time.

是的，这个环境就是被实时生成的。

And yeah, it's just that environment's being made up.

所以我无法想象二十年后，这种技术会发展到什么地步，以及在虚拟现实中，人们在虚拟世界中体验过的事物，其记忆回溯会是什么样子。

So I can't imagine in twenty years where this is and then the memory recall of what's been experienced in this made up world that somebody's going around and sensing in virtual reality.

我想我说这些，是因为你在提出的这个观点——你如何知道你所经历的是真实的世界？

And I guess I say all this because it's teasing out this idea that you bring up, which is like, how do you know what is real and what you're experiencing is reality?

因为当你进入任何这些理论物理学的讨论时，他们会告诉你，你以为自己知道的一切，根本不是那么回事。

Because you go into any of these theoretical physics conversations and they'll tell you that what you think you know is not that at all.

从量子力学的角度来看，情况截然不同。

It's very, very, very different from a quantum mechanics standpoint.

但不管怎样。

But anyway.

当然。

Absolutely.

我们在之前的节目中讨论过，英伟达正在创建一个名为Cosmos的模拟系统，让人们能够在进入现实世界之前，在模拟环境中测试机器人，而这些模拟世界如今已经变得极其逼真。

And we've spoken about it on previous episodes where NVIDIA is creating, I think it's Cosmos, which is their simulation so that people are able to test robotics in a simulated world before they enter the physical world and these simulated worlds are getting so realistic now.

它们具备流体动力学、重力，能够施加各种各样的物理压力等。

They have fluid dynamics, they have gravity, they're able to apply all of these very various physics pressures and such.

所以我认为我们所处的世界真的非常非常迷人。

So I think that the world we're living in is really, really fascinating.

我们正处在一个将物理世界与数字世界融合的阶段，以至于我们还能否将它们区分开来？

We're at a point where we're kind of joining the physical and the digital world to the point where it's just like, we going to be able to separate them?

因为我认为，有一些青少年和孩子在这样的新世界中成长，由于他们不断在数字世界中进行社交、情感互动，有时会分不清什么是真实的，什么不是。

Because I do think that there's these teenagers, there's kids that are growing up in this new world and sometimes they get confused between what is real and what's not because of that constantly interacting socially, digitally, everything emotionally in the digital world.

是的。

Yeah.

好了，各位。

Alright, guys.

我们今天就到这里。

We're gonna wrap here.

希望你们喜欢这次的对话。

I hope you guys enjoyed the conversation.

塞布，我们可能会在两周后再次进行这次讨论，来聊聊那些我们还没来得及涉及的其他话题。

Seb, we'll probably do this again in, like, two weeks to kinda cover the other topics that we didn't even get to.

最重要的是，我们想聊聊特里斯坦·哈里斯在CEO日记——史蒂文·巴特利特播客中的讨论。

Most importantly, we wanna cover this Tristan Harris discussion on the diary of a CEO, Steven Bartlett's podcast.

那里提到了几个不同的话题，我们想在节目中进一步探讨，同时也会继续聊更多科技相关的内容。

A couple different topics that were brought up there we wanna discuss on the show in addition to more tech topics.

如果你们喜欢这个节目，请在评论区告诉我们。

If you guys are loving this, let us know in the comments.

我们玩得很开心。

We're having fun.

希望你们在听我们分享这些多样话题时也感到有趣。

Hopefully, you're having fun hearing some of the the different topics that we're bringing to you.

如果你有任何想听的话题建议，请在X上告诉我们，我们一定会努力在下一期节目中融入这些内容。

And if you have any recommendations of things you wanna hear, bring them to us on X, and we'll be sure to try to incorporate it into the next show that we do.

所以，Seb，给大家推荐一下你的书或者其他你想推广的内容，然后我们就结束吧。

So Seb, give people a hand off to your book or anything else that you wanna highlight that's out there and then we'll go ahead and wrap.

当然，我们非常感谢任何愿意花时间去实践这些步骤和教训的人。

Absolutely and again, we really appreciate anyone who just takes the time to give these steps those lessons.

所以，如果你任何时候想听听我们对这项技术或那项技术的看法，欢迎随时在评论区分享。我们真的很想聊聊当今世界正在发生的事情，分享我们的观点。

So if at any point you're just like, oh man, I'd love to hear your guys' perspective on this technology or that technology, feel free to just share it in the comments And yeah, we really want to go talk about what's happening in the world today and just kind of share our perspective.

你可以在Twitter上找到我，用户名是said bunny，也就是B U N N E Y。我的书是《金钱的头号成本》或《B代表比特币》。

Yeah, you can find me at said bunny and that's B U N N E Y on Twitter and my book, The Heading Cost of Money or B is for Bitcoin.

但再次感谢大家收听。

But again, appreciate everyone giving us a listen.

感谢收听TIP。

Thank you for listening to TIP.

请在您喜爱的播客应用中关注Infinite Tech，不错过任何一期节目。

Make sure to follow Infinite Tech on your favorite podcast app and never miss out on our episodes.

要获取我们的节目笔记和课程，请访问theinvestorspodcast.com。

To access our show notes and courses, go to the investorspodcast.com.

本节目仅用于娱乐目的。

This show is for entertainment purposes only.

在做任何决定之前，请咨询专业人士。

Before making any decisions, consult a professional.

本节目版权归投资者播客网络所有。

This show is copyrighted by The Investor's Podcast Network.

未经书面许可，不得进行 syndication 或重播。

Written permissions must be granted before syndication or rebroadcasting.