Unlocking Reality: Donald Hoffman on Consciousness, Simulations, and the Limits of Space-Time | Impact Theory

它们的规模可以任你设定。

They can be as big as you want.

你可以有三个。

You can have three.

你可以有一百亿、一万亿个。

You can have a billion, a trillion.

你的马尔可夫矩阵有多大，就可以设多大。

Go as big as you Markov matrix as you want.

直接延伸到无穷大。

Go off to infinity.

真的是无穷多个状态。

Literally, infinity of states.

然后看看这一整套东西。

And then look at all this whole set.

这些就是所谓的单子。

These are all the quote unquote Monet.

所以一共有无穷多个。

So there's an infinite number of them.

他想要一种预先设定的和谐。

And he wanted a preestablished harmony.

那这到底是什么意思呢？

And Which is what?

这可能得花几分钟，但我可以讲讲我对预先设定和谐的理解。

So I I can this will take a couple minutes, but I can give it what I think the pre established harmony is.

其实这个想法非常简单。

So, if I have a and it's a very simple idea, actually.

假设我盯着红灯、绿灯和黄灯，我可以算出各种状态转换的概率——比如这是个有问题的红绿灯，红灯有时候会直接变黄，而不是变绿，等等。

So, if I if I have a if I'm looking at the red, green, and yellow lights, and I could write down the probability, say it's a bad traffic light, so sometimes red turns yellow instead of green and so forth.

这红绿灯有问题。

So it's a bad traffic light.

你把红绿灯所有状态变化的规律写成一个矩阵，就知道它到底是怎么运作的了。

So you write down the matrix of whatever this traffic light is doing, you know, how it changes from red, green, and yellow.

现在，假设我戴上一副眼镜，看不到黄色了。

Now suppose I put on a pair of glasses where I can't see yellow.

我只能看到红灯和绿灯之间的切换，但看的还是同一个交通灯。

All I can see is the red and green transitions, but I'm still looking at the same traffic light.

那我就会写出一个不同的矩阵。

Then I will write down a a different matrix.

这个矩阵只有两个状态：红和绿。

It only have two elements, red and green.

对吧？

Right?

我只能看到红和绿。

That's all I can see is red and green.

所以，如果现在是红灯，接下来可能是红灯或者绿灯；如果现在是绿灯，接下来也可能是绿灯或者红灯。

And so I'll have a probability if I'm seeing red now, I'll see red next or green next, and if I see green now, green next or red next.

这样我就得到了一个二乘二的矩阵。

So I'll get a two by two matrix.

但我看的是红绿黄三色灯，实际上只看得到红和绿，而红绿灯的切换本来就是由完整的三色系统控制的，所以当我只能看到红和绿的时候，它们的变化其实是由黄色那部分决定的。

But since it's I'm looking at the three by the the the red, green, and yellow light, but I'm only seeing the red, green, the red, green, and yellow transitions are, of course, controlling what I see in the red and green when I can only see red and green.

对吧？

Right?

这很明白。

It's it's clear.

我看到的所有这些切换，都是因为背后有个完整的三色系统在运作，只是我看不见黄色而已。

The all the transitions that I'm seeing there are because I'm seeing this bigger one, but I'm missing the yellow.

黄色这部分对我而言是隐藏的。

It's it's it's hidden from me.

这才是关键所在。

That's the key idea.

所以在数学上，如果我们有了红、绿、黄三色的完整转移矩阵，那我就能精确算出只包含红和绿的简化矩阵是什么样子。

So so in mathematics, we say, if you give me the matrix for the red, green, and yellow, then I will tell you exactly what the matrix has to be for the red and green.

而且只会有一个确定的结果。

There's gonna be only one.

对吧？

Right?

这叫做迹。

And it's called the trace.

这才是关键。

And that's the key.

这个发现早在1905年到现在的几十年间就已经有了。

This was discovered decades ago, between nineteen o five and now.

人们发现，如果我有一个大矩阵，只关注其中一部分状态，那么通常会存在一个唯一的矩阵。

It was discovered that if I have a big matrix, and I say, I'm only gonna look at a subset of the states, then there will be a unique in general, a unique matrix.

这里面有一些专业术语。

There are technical terms.

你需要满足遍历性之类的条件，但那些我们这里就不深入了。

You have to be ergodic and so forth, but that's beyond what we need to do here.

所以我简单说一下，一般来说——当然，对懂数学的人来说，前提是遍历性成立。

So but I'll just say in general, but for those who know the math, if it's ergodic.

但不管怎样。

But but anyway.

所以我有一个很大的矩阵，只看其中一部分状态，就会得到一个轨迹矩阵，也就是所谓的零意外矩阵。

So so I have a big matrix, and I look at a subset of states, then I will get a a trace matrix, which is exactly the zero surprise matrix.

你所说的零意外是什么意思？

What do you mean zero surprise?

零意外就是说，我预测的概率转移，和实际发生的一模一样。

Zero surprise means that the probability transitions that I predict are exactly the probability transitions that I will have.

所以从概率上讲，这正是我会看到的结果。

So this is probabilistically exactly what I will see.

所以轨迹运算——抱歉，轨迹运算会给出一个零意外的结果。

So the trace logic so I'm sorry, the trace operation gives you a zero surprise.

它精确地呈现了你将看到的一切。

It gives you exactly what you will see.

是因为你把可能性限制得太过彻底，只剩下非常有限的几种情况吗？

Because you have limited the options so profoundly that there's just a very finite number?

因为这本质上就是一个数学定理，

Well, because it's just a mathematical theorem, basically, that

但之前那个定理不是跟红绿蓝有关吗？

But wasn't it a mathematical theorem with red, green, and blue?

我得说得严谨一点。

Or I should be careful.

有了红、绿、蓝三种颜色，你可以任选一个矩阵。

So with the red, green, and blue, you can pick any matrix you want.

随便选三个，就会有一个

Pick any three then there's a

新的。

new one.

当我减少一种颜色时，突然就变成了零意外的情况。

When I reduce by one color, all of a sudden I get to a no surprise one.

没错。

That's right.

所以从这三个3x3矩阵中选一个，如果你固定了这个矩阵，那红和绿就被限制住了，我会给你一个对应的2x2矩阵。

So pick one of those 3x3 matrices, and if you fix that matrix, then you restrict red and green, I will give you a 2x2 matrix that is matched to that particular 3x3.

如果你换一个3x3矩阵，对应的2x2也会不一样。

If you give me a different 3x3, it'll be a different 2x2.

对。

Yep.

明白吗？

Okay?

所以不管你给我哪个3x3矩阵，我只要取它的迹，就能得到完全匹配的2x2矩阵，这个矩阵会准确呈现你在看不到黄色时所看到的红绿变化。

So whatever 3x3 you give me, when I take the trace, I will get exactly the right 2x2, which gives you exactly the transitions on the red and green that you will see if you can't see the

黄色。

yellow.

对。

Mhmm.

这就叫迹，其实只是一个数学定理，它到底在告诉我们什么？

So that's called the trace, and it's just a mathematical theorem that What's it telling us?

它告诉我们，当你面对一个更大的系统，却无法看到它的所有状态时，它能精确告诉你能观察到什么，以及在概率上你具体会看到什么。

It's telling us exactly when you have a bigger system, and you can't see all of its states, it will tell you exactly what you can see, what exactly you will see probabilistically, exactly.

这个在头戴设备的类比中有什么意义呢？

Why does that matter in the headset analogy?

结果是，

It turns

这一点其实已经为人所知很多年了。

out well, so this has been known for for many, many decades.

而我在大约两年前发现的是，这为我们提供了一种适用于所有马尔可夫链、所有观察者的逻辑。

And what I what I discovered about two years ago was that this gives us a logic on the set of all Markov chains, on the set of all observers.

它把它们全都联系在了一起。

It ties them together.

它建立了一种预先协调的和谐，统一了所有这些观察者。

It gives a pre established harmony that unites all of these observers.

所以，我先来谈谈一种布尔逻辑，一种简单的逻辑，就像计算机科学里那种布尔逻辑。

So a logic I'll talk first about a boolean logic, a simple logic, a boolean logic that we have in, like, in in, you know, computer science.

你可以把各种东西组合起来，用与、或、非这些运算符来处理。

You can take things and take their ands and their ors and and the nots and and so forth.

所以这是一种逻辑体系。

So that's a kind of logic.

而所谓的轨迹逻辑，并不是布尔逻辑。

And the logic of the so the trace logic is not Boolean.

它比布尔逻辑更通用。

It's more general than Boolean.

这是一种非布尔逻辑。

It's a non Boolean logic.

但它在局部范围内是布尔的。

It's but it's locally boolean.

比如，如果我固定一个矩阵，假设它有一千个状态，然后观察它所有的可能轨迹——这些轨迹数量庞大，但如果你只看其中一部分，

So if I take any particular if you fix a matrix, say it has a thousand states in it, and you look at all of its possible traces, and there's, you know, lots and lots of traces, if you look at just those You mean

轨迹就是子集吗？

traces are just subsets?

子集加上这个矩阵。

Subsets plus the matrix.

对吧？

Right?

子集和它对应的矩阵，矩阵会告诉你在这个状态子集内会发生哪些转移。

The the the subset and the matrix on that subset, which is telling you the transitions that you're gonna see on that subset of states.

明白了。

Okay.

对吧？

Right?

比如我从三色的红、绿、黄简化到红和绿时，就是在告诉你从红到绿、绿到红这些转移的概率。

For like when I went from the three by three red green and yellow to the red green, I was telling you what's the probability of transitioning from red to green or green to red and so forth.

所以我当时说的，就是这些概率的具体数值。

So I was telling you exactly what those probabilities are.

这正是我想表达的意思。

So that's what I'm saying.

你随便选一个大矩阵，然后看看它的所有迹。

You you fix any big matrix and then look at all of its traces.

对。

Mhmm.

这些迹合起来就构成了一套布尔逻辑。

And then the all of those traces form a Boolean logic.

或者说是迹逻辑中的一个布尔子系统。

Or Boolean sub logic of the trace logic.

但整个逻辑本身并不是布尔的。

But the whole logic itself is not Boolean.

这才是关键。

So that is the key.

现在我们有了一个观察者的理论。

This is we now have a theory of an observer.

非常简单。

Very, very simple.

这是一个马尔可夫矩阵。

It's the Markov matrix.

我们有预先设定的和谐，也就是这种迹逻辑。

We have the pre established harmony, which is this trace logic.

然后我们就可以用这些矩阵来构建时空。

And then we can start to use these matrices to build spacetime.

我给你一个具体的提示，就一个具体的提示，告诉你这会怎么展开。

So I'll give you just a concrete clue, just a concrete clue about how this is going to go.

在马尔可夫链理论中，除了矩阵之外，通常还会有个计数器。

It's standard in Markov chain theory to have, in addition to the matrix, a little counter.

每次你的体验发生变化，比如从红色变成绿色，就把计数器加一。

Every time you change an experience, like from red to green, you just increment the counter.

其实没我讲的这么复杂。

It's really as simple as I'm making it.

所以不妨先从零开始计数。

So maybe start the counter off at zero.

现在我看到红色了。

Now I'm seeing red.

哦，这是第一次。

Oh, that's one.

现在，嗯，变成绿色了。

Now oh, now it's green.

不是第二次。

Not two.

现在是黄色了。

Now it's yellow.

明白了。

Okay.

第三次。

Three.

以此类推。

And so forth.

你只需要一直累加计数器就行。

You just keep adding the counter.

注意现在计数器会发生什么变化。

Notice what happens to counters now.

如果我有红、绿、黄三种颜色的矩阵，但只能看到红色和绿色的话。

If I have the red, green, and yellow matrix, and then I can only see red and green.

对。

Mhmm.

注意，红绿组合有自己的计数器。

Notice the red green has its own counter.

对吧？

Right?

整个矩阵有自己的计数器，但它不会算上黄色。

The the matrix has its own counter, but it doesn't count the yellows.

因为它看不到黄色。

It because it doesn't see yellow.

所以，那个涵盖红、绿、黄三种颜色的更大矩阵，它的计数器会比只算红和绿的那个计数器跑得更快。

So the counter that's bigger, on the bigger matrix, so the red, green, yellow, is going to count faster than the counter on the red and green.

对。

Mhmm.

这正是相对论里时间膨胀的原理。

That is going to give us time dilation in relativity theory.

这就是我们从马尔可夫链过渡到爱因斯坦狭义和广义相对论中时间膨胀现象的精确方式。

That's the precise way that we will be able to go from Markov chains to the time dilation that we see in Einstein's special and general relativity.

我们先暂停一下，但后面还有更多精彩内容，别走开。

We're hitting pause for a moment, but there's plenty more ahead, so don't go anywhere.

每个企业都存在这个问题——大量时间浪费在对账表格、追报表、反复在彼此不互通的系统间重新输入数据上。

Every business has it, a ton of it, hours buried in reconciling spreadsheets, chasing down reports, reentering the same data across systems that don't talk to one another.

这些成本从不会出现在损益表里，但它们实实在在地消耗着你的时间、效率，还有那些总是来得太晚的决策。

Nobody puts it on the P and L, but it is costing you in time, in speed, and decisions that just come too late.

解决方案就是甲骨文的NetSuite。

The solution is Netsuite by Oracle.

这是超过四万三千家企业信赖的首选AI云ERP系统。

It's the number one AI cloud ERP trusted by over 43,000 businesses.

它把财务、库存、电商、人力资源和客户关系管理全都整合到一个统一的数据源中。

It brings your financials, inventory, commerce, HR, and CRM into one source of truth.

再也不用从五个不同的地方拼凑数据了。

No more stitching together data from five different places.

现在有了NetSuite AI连接器，你可以用任何你喜欢的AI工具，直接访问你真实的业务数据。

Now with Netsuite AI Connector, use the AI of your choice to tap into your actual business data.

任何你想问的问题，都能得到解答。

Ask every question you've ever had.

如果你的年收入达到七位数或以上，免费领取NetSuite的《AI实战指南》，访问 netsuite.com/theory 获取。

If your revenues are in the 7 figures or higher, get Netsuite's free business guide, Demystifying AI at netsuite.com/theory.

这份指南免费提供，地址是 netsuite.com/theory。

The guide is free to you at netsuite.com/theory.

再次提醒，访问 netsuite.com/theory 就能领取。

Again, that's netsuite.com/theory.

我们来聊聊财富积累中最强大的力量——复利。

Let's talk about the most powerful force in wealth building, compounding.

为什么这么多年来，人们一直在谈论它，其实是有原因的。

There is a reason that people have talked about this forever.

你让资产发挥作用，它就会不断产生更多的自己。

You put an asset to work, it generates more of itself.

随着时间推移，曲线会越来越向你有利的方向倾斜。

Over time, the curve bends way in your favor.

每一个认真积累财富的人都懂复利的运作方式。

Every serious wealth builder knows how compounding works.

但大多数黄金持有者，却完全错过了这一点。

Most gold owners, though, have completely opted out of it.

Monetary Metals 改变了这个局面。

Monetary Metals changes that equation.

他们创造了一种方式，让你的黄金能获得收益，而且不是以美元支付，而是直接增加更多的实物黄金。

They've built a way for your gold to earn yield paid not in dollars, but in more physical gold.

每年最高可达4%的收益，每月都能增加更多黄金，复利增长在一种无法被印制、通胀或贬值的资产上。

Up to 4% per year, more ounces, stacking every single month, compounding in one of the only assets that can't be printed, inflated, or debased.

顺便说一下，保管和保险都包含在内，这样就不会有隐藏费用蚕食你的收益。

And storage and insurance are included, by the way, so hidden fees don't eat your yield.

普通人正在不知不觉地承受通货膨胀的损失。

The average person is passively absorbing inflation.

这才是你通过复利摆脱通胀的方法。

This is how you compound your way out of it.

点击节目说明中的链接，或访问 monetarymetals.com/impact 了解更多信息。

Click the link in the show notes or visit monetarymetals.com/impact to learn more.

本节目为付费广告。

This is a paid advertisement.

我们来聊聊你到底有多脆弱。

Let's talk about how exposed you really are.

假设你在路上别了别人一下。

Let's say you cut someone off in traffic.

他们气疯了。

They're furious.

他们拍下你的车牌号。

They snap a photo of your license plate.

几分钟内，他们就能通过人肉搜索网站查到你的家庭住址、电话号码，甚至你孩子的名字。

Within minutes, they can pull up your home address on a people search site, your phone number, the names of your kids.

你的个人信息正躺在数百个数据经纪网站上——地址、电话、邮箱，连你的社保号都不例外。

Your personal information is sitting on hundreds of data broker sites, your address, phone number, email, even your social security number.

Incogni 能帮你彻底清除这些信息。

Incogni shuts that down.

他们会追踪你在数百个网站上的数据，并全部删除。

They track your data down across hundreds of sites and remove it.

你只需要授权一次，剩下的他们全包了。

You authorize them once, they handle the rest.

而且只要信息一重现，他们就会继续帮你删掉。

And they keep removing it as it comes back.

此外，你可以自行提交任何显示你信息的链接，他们的团队会帮你彻底删除。

Plus, with custom removals, you can send them any link where your data shows up, and their team is gonna get it taken down.

Incogni 是唯一获得德勤独立认证的数据删除服务，效果有据可查。

Incogni is the first and only data removal service independently verified by Deloitte, so you know it actually works.

前往 incogni.com/impact，使用代码 impact，年度套餐立享六折优惠。

Go to incogni.com/impact and use code impact for 60% off an annual plan.

免费试用三十天，不满意随时取消。

Try it risk free for thirty days.

就是 incogni.com/impact，别忘了用代码 impact。

That's incogni.com/impact, and be sure to use code impact.

感谢你一直听下去。

Thanks for sticking around.

我们马上继续回到正题。

Let's get right back into the action.

到底少了什么，才导致这些数据不一致？

What's missing that causes the counters to be different?

你只是没看到全部情况。

You just aren't seeing everything.

重点在于，这其实是不一样的。

The idea is going be the different

为什么它会跟光速联系得这么紧密？

Why is it mapped so specifically to the speed of light?

所以这意思是……

So that's a

这是个好问题。

good question.

如果我要提出这样的说法，那光速到底是在哪里冒出来的呢？

So if I'm going to make this kind of claim, where does the speed of light come out in all of this?

对吧？

Right?

其实我更想问的是，能量等于质量这个计算过程是怎么回事。

Well, maybe more the computation of energy equals mass is probably what I'm asking.

光速会体现为矩阵中状态之间转换的最快速方式。

The speed of light is going to come out as the fastest way that you can transition among states in in the matrix.

还有一种叫循环矩阵的东西。

And there's something that are called cyclic matrices.

对。

Mhmm.

比如，对于红、绿、黄三种颜色，如果我现在是红色，那么下一步变成绿色的概率就是100%。

So a cyclic matrix for red, green, and yellow would say, for example, if I'm red now, the probability is one that I'll see green next.

接着，变成黄色的概率也是100%。

The probability is one then that I'll see yellow next.

然后，再回到红色的概率同样是100%。

And then the probability is one then I'll go back to red.

这样循环下去，就是状态之间流转的最快方式，这实际上对应着一个与光速相关的矩阵。

So you cycle in, and that's gonna be the fastest way that you loop among all of and that will effectively be that will that will correspond to a matrix that corresponds to the speed of light.

明白了。

Okay.

但如果我们换一个矩阵，比如我现在停在红灯，那我有50%的概率变绿灯，也有50%的概率继续停在红灯。

But if we're, for example, if I'm a different matrix that says I'm stopped at red right now, yeah, I might go to green with the probability of 0.5, but I might stay at red with probability 0.5.

这样的话，我在各个状态之间移动的速度就会慢很多。

Now I'm going to be moving much more slowly around the states.

所以这里存在一个最大速度，也就是你在这个情境下可以称之为‘通行时间’的东西。

So there is a maximum speed at which you can, you know, what would, in this case, you can call the commute time.

有一个最快通行时间，也就是最短的通行时间，相应地，也就有了最快的速度。

There's a fastest commute time, a smallest commute time, and therefore fastest speed.

你可以用这个概念来逐步建立类似时间膨胀这样的想法。

So you can use that for starting to build up notions like on time dilation.

接下来会稍微有点专业，我就简单带过一下，毕竟我们不想深入数学细节。

And then now it gets a little technical, so I'll just hand wave a little bit because we don't want to go into the math.

但在这些矩阵上，存在一种叫作狄利克雷形式的数学结构。

But there are then, on these matrices, there are mathematical structures called Dirichlet forms.

狄利克雷形式？

Dirichlet forms?

是的

Yeah.

D-I-r-I-c-h-l-e-t，叫狄利克雷，是个法国姓氏，狄利克雷形式，能帮我们有效描述状态在马尔可夫链中的扩散方式。

D I r I c h l e t, Dirichlet, French name, Dirichlet forms, that will allow us to use effectively how states diffuse in a Markov chain.

这两个状态的扩散模式非常相似。

The two states diffuse in a very similar pattern.

我们会把它们在时空头显里放得靠近一些。

We're gonna put them close together in our space time headset.

对

Mhmm.

所以，我们也可以用狄利克雷形式来构建这些矩阵上的距离。

So we can build distance from Dirichlet forms on these matrices as well.

所以计数器会膨胀，狄利克雷形式也会变化，因为当你做迹运算时，这些狄利克雷形式所对应的距離会收缩。

So the counters dilate, the Dirichlet forms change, because when you take a trace, the distances that you're going to get on these Dirichlet forms contracts.

因此，我们会从这里得到时间膨胀和长度收缩。

So we're going to get time dilation, length contraction from this.

现在，

Now,

那为什么这些变化会这么可预测呢？

that's Why top would the changes be so predictable?

当接近光速时，事物变化的精确模式可以被非常清晰地描述出来——时间与距离的反应方式截然不同，这使得爱因斯坦能够提出理论。我完全承认他可能是错的，但用‘时空弯曲’来近似描述，确实能让我们精准地测量这些现象。

So there is a precision with which things change that can be mapped out very eloquently when you're talking about the speed of light as we approach it and how time and distance react very differently, which allowed Einstein, and I'm perfectly willing to accept he was incorrect, but his approximation by calling it a curvature of space time allows you to very precisely measure this stuff.

如果我没理解错的话，天啊，这假设可不小——如果我面对的是这些巨大的矩阵，所有可能的马尔可夫链组合，然后我只看其中的子集，但整个矩阵的内在关联依然存在。如果我比较两个不同的迹子集，距离看起来会收缩，时间似乎也变慢了，但不同的情况表现又各不相同。

If I'm understanding him, Jesus, this is a big if, but if I'm looking at this gigantic matrices all the different Markov chain combinations that there could be, and I'm saying I'm looking at now subsets of that, but there's still the sort of connective tissue of the whole of this matrices, if I'm looking at two different traces subsets, then the distance will appear to be collapsed, the time will appear to slow down, but different ones are going to behave differently.

没错。

That's right.

所以，我不太明白——当然，我承认这是我的无知——但不管它们处于什么状态，不管是从12个变量降到2个，还是降到10个，为什么距离的收缩总能以如此一致、可预测的方式发生，以至于让爱因斯坦误以为自己看到了时空弯曲？

And so I don't see, and this is just my ignorance, fully accept, but I don't see how that, no matter where they are, no matter whether we're going from 12 things down to two or 12 things to 10 or whatever, that it would collapse the distance in the same predictable fashion that would trick Einstein into thinking he was seeing the curvature of space time.

这正是我们需要证明的。

And that's what we have to prove.

所以这就是

So that's what

但你就是在说这个吧？

But that's what you are saying?

我其实明白你的意思。

I actually understand.

我就是这个意思。

I'm absolutely saying.

我们并不是说爱因斯坦错了。

So we're not saying that Einstein is wrong.

我们说的是，爱因斯坦

We're saying that Einstein

时空根本是假的。

Well, space time is fake.

所以你

So you

我不至于说时空是假的。

are saying Well, I won't say that space time is fake.

我会说，时空是一个很棒的数据结构。

I will say that space time is a wonderful data structure.

但它并不完整，时空并不是最终的现实。

It's just not the full it's not all of space time is not the final reality.

它就像一副头戴设备。

It's just a headset.

但这副设备很不错。作为科学家，如果我声称自己在时空之外使用马尔可夫链来构建更深层的软件，我就得向你清楚地展示，我是怎么推导出爱因斯坦的狭义和广义相对论的。

But it's a good headset, and, you know, and what I have to do, as a scientist, if I'm claiming that I'm using these Markov chains outside of space time to get the next layer of software outside of space time, I have to show you exactly how I get Einstein's special and general relativity.

我也得清楚地展示，我是怎么从中推导出量子场论的。

I have to show you exactly how I get quantum field theory out of it.

你

Do you

你觉得马尔可夫链是被有意识地操控，才产生时空弯曲的效果吗？

think that the Markov chain is being consciously manipulated to create the effect of the bending of space time?

其实我根本没必要这么说，不过我以后也可以接受这种说法。

Well, so I don't need to say that at all, but I can allow it later on.

我可以说，我相信某些类型的马尔可夫链能够推导出时空的弯曲。

Should say that I can get the curvature of space time, I believe, from certain kinds of Markov chains.

实际上，我现在的想法是，马尔可夫链的种类非常多，但其中有一小部分——可能是概率为零的特殊子集——才能用来构建我们这种时空模型。

So in fact, the way I think about it right now is there's a wide variety of Markov chains, and I think that there is this very special subset, probably what we call a measure zero probability zero subset of Markov chains that can be used to build the kind of headset that we have.

我们目前用的这种模型非常初级、简单，绝大多数马尔可夫链根本达不到这个效果。

We have such a sort of a low grade trivial kind of headset that the the that most Markov chains just won't do it.

它们都在做其他有意思的事情。

They're off doing other interesting things.

但确实有一类马尔可夫链，可以用来构建平直或弯曲的时空，而另一些则完全无法产生这种平滑的时空结构。

So, but there is a subset of Markov chains that you can use to build flat space time or curved space time, and then there are others that will basically, that kind of smooth space time just won't work for them.

它们就是不行。

They they just won't.

马尔可夫链让你有机会重现爱因斯坦的所有理论，甚至还能超越它们，无限延伸。

So Markov chains give you the chance to build everything that Einstein has, but then infinitely more.

这非常有意思，我可能说得太简单了，因为这是我第一次听到这些，说不定我根本没完全理解。

It's very interesting, and I'm probably I'm very at risk of oversimplifying what you're saying because I'm hearing it for the first time and I may just not understand it.

我听过很多这样的访谈，能明显感觉到采访者其实没跟上对方在讲什么。

And I've heard so many of these interviews where you can tell that the interviewer is like, not quite tracking what's being said.

所以我现在也面临同样的风险。

So I run that risk right now.

但你说这些的时候，我想起一句话：对于拿着锤子的人，所有问题看起来都像钉子。

But I will say, as you were describing all of this, again, to a hammer, every problem is a nail.

不过当我看到这一点时，觉得把这看作代码、看作模拟是说得通的，因为你确实想用程序生成的方式构建。

But when I look at this, the metaphor or reality that this is code and it's simulated would make sense because you really want to build procedurally.

但要程序生成，你就得搞清楚概率空间到底是什么，这一点还不太清楚。

But to build procedurally, you have to know what the probability space is not quite right.

所有可能的组合到底有哪些？

What are all the permutations?

于是你就开始想，这会不会是带标记的排列？

So you get into this idea of, is it decorated permutations?

那是一种，所以

That's one So the

我查了下那是什么，才明白你们其实是想考虑更多变量。

when I looked up what that was and was like, Oh yeah, you guys are just trying to account for more variables.

如果你在做程序生成的世界，光是生成地形还不够，还得安排树木之类的元素。

So if you're doing a procedurally generated world, not only do you have to come up with a terrain, but you've got to like place trees and stuff like that.

然后你就得想，这棵树要朝哪个方向？

So then you go, Okay, what orientation is the tree?

这棵树要多高？

How tall is the tree?

翻滚、俯仰、偏航，这些都要定下来。

Roll, pitch, yaw.

所有这些细节都得一一确定。

Like you have to determine all of those different things.

没错。

Exactly.

所以你描述的马尔可夫链，还有那些痕迹元素——虽然你没说‘痕迹元素’这个词，但应该是这个意思。

And so it feels like what you're describing with the Markov chains with the trace elements, didn't call it trace elements, trace something.

追踪逻辑。

Trace logic.

追踪逻辑。

Trace logic.

谢谢。

Thank you.

所有这些都还在我的普通认知范围内，符合这个模拟现实的框架。

All of that stuff is still fitting within my layman's understanding of this simulated reality.

如果我们能成功，就会形成一些定理，明确告诉你该怎么写代码，你就能直接写出这样的代码

It's we when we're if we succeed at this, we will have theorems that tell you exactly how to write the computer code, and you could write computer code

来实现这一点。

to do this.

你说话的时候，我脑子里就在想，如果这真能成为我能用在游戏里的突破性方法，我一定会特别兴奋。

Literally, as you were talking, I was like, man, if this is gonna be a breakthrough that I can use in my game, I'm gonna be very excited.

这么说吧。

That's put it this way.

如果我不能给你一个能用在游戏里的突破性成果，那我就错了。

If I cannot give you a breakthrough that you can put in a game, then I'm wrong.

这话说得挺在理。

This this is fair enough.

所以我们正在建立这个TRACE研究所，目的就在这儿。

So that's what that's what we have this TRACE Institute that we're building.

我们的目标是在未来五年内，把这些定理连同证明彻底搞清楚。

That's the goal is within the next five years to basically nail down these theorems with proofs.

有了这些证明，我们就能明确地知道该怎么构建这些算法，接下来就万事俱备了。

The proofs will then give us effectively the ways to to build these algorithms, and we're off to the races.

这正是我们的目标。

So that's that's the goal.

一旦我们弄明白了怎么做出这个头戴设备，就能开始实际测试了。

And the idea will be that once we figure out how to build this headset, then we can start to play with it.

到时候我们可以反过来去分析它的原理。

We can start to reverse engineer it.

借助追踪逻辑，我们现在能实现一些近乎神奇的效果。

And using the trace logic, now we have the ability to do some kinds of magic.

但我应该把这个放进一个完整的框架里。

But I should just put this in a framework.

目前，卡尔·弗里斯顿等人在自由能原理方面做了大量出色的工作，致力于减少意外、区分自我与外界的边界等等。

Right now, there's a lot of work that's being done on that's very, very good work by Carl Friston and others on the free energy principle and trying to minimize surprise and maintain markup blankets for self versus outside and so forth.

这和我们讨论的思路非常相似。

And this is very much similarly related to that kind of thing.

追踪逻辑与最小意外之间的关系在于：

The trace logic is related to the minimum surprise in the following sense.

追踪逻辑本质上就是零意外的逻辑。

The trace logic is the logic of zero surprise.

所以我所主张的是，生命所追寻的前定和谐，其实就是这种零意外的追踪逻辑。

This so what what I'm claiming is that the pre established harmony that liveness was looking for is this logic of zero surprise, the trace logic.

由此，我们对自我与外界的区分方式，也与那种‘标记毯’式的思路大不相同。

And then it leads to a very different approach to maintaining self versus outside than the mark of blanket kind of thing.

我们得到的东西比所谓的‘标记毯’更广泛。

We get something that's broader than the mark of blanket.

不过，其中关键的一点是，如果你感兴趣，我们还可以深入更多细节。

So but one of the key and we can go into the there's more more details if you want to.

我可以给你演示，我们是怎么通过递归运用追踪逻辑来构建能动性的。

I can show you how we build agency by recursion of the trace logic.

在继续之前，我想先说一句。

Well, before we we go

不过在那之前，我有点担心，非技术背景的听众可能跟不太上。

there, because I do worry that we're at risk of losing some nontechnical people.

我想聊一聊人工智能。

One thing I want to talk about is AI.

当你说到神经元其实不是真实的，这一切都是虚构的，我就想，那为什么AI代理就不能有意识呢？

So when I think about the way that you describe like the neurons aren't real, this is all fake, I go, well, then why can't an AI agent become conscious?

为什么AI代理现在就已经没有意识呢？

Why isn't an AI agent already conscious?

如果意识是根本的，那是什么阻止了人工智能拥有意识？

Like, if consciousness is fundamental, what is it that stops AI from being conscious?

这么说吧。

Okay.

这取决于你怎么去理解意识是如何产生的。

So it depends on how you want to frame the question of of the emergence of consciousness.

如果你是从一个本身没有意识的物理系统出发，想知道怎么通过操作这个系统让它产生意识，

If if you're framing it in this in the case where you're saying, I'm gonna start with a physical system that's not conscious, and my question is, what can I do with that physical system to make consciousness emerge?

那我的回答是：做不到。

Then I say, you can't do that.

意识永远不会凭空出现。

It will never come.

所以你的观点是，一切事物本来就有意识？

But your take is that everything is already conscious.

没错。

That's right.

所以我希望你能明白，为什么我们在做假设时必须格外谨慎。

So so I'm giving a so you can see why we have to be very careful about our assumptions.

我的大多数同事都把空间和时间当作基本前提，他们认为这些东西本身没有意识，于是就想弄清楚，到底需要什么样的神经结构、神经活动或功能特性，才能产生意识。

Most of my colleagues are are starting with space and time as fundamental, and they're saying that, of course, they're not conscious, and so we're trying to figure out what neural structures or neural activities or properties are going to be required, what functional Is that

一个物体有意识吗？

a lender conscious?

那种问题正是他们才会问的。

Well, that's the kind of question they would have.

你怎么看？

What do you think?

我先说一句，我完全不认同这种框架——意识并不是从物理系统里冒出来的，也不是从神经元或者类似的东西里产生的。

I'll first say I reject that entire framework, that consciousness does not emerge from physical systems, it doesn't emerge from neurons or anything like that.

但那个唯一的意识，会出现在唐身上，也会出现在汤姆身上，对吧？

But does the one consciousness, the one consciousness does pop into Don, it does pop into Tom.

没错。

Right.

那为什么它不会出现在搅拌机里呢？

Why doesn't it pop into the blender?

所以呢，我

So so I

不管我看到什么，这个杯子、这张桌子，都是如此。

would say that no matter what I see, this cup, this table Mhmm.

就连一只蚂蚁，每一种情况，我本质上都在与一个有意识的现实互动。

An ant, In every case, I'm fundamentally interacting with a conscious reality.

我的头戴设备让我对这个有意识的现实有了或多或少的洞察。

And my headset gives me more or less insight into that conscious reality.

和汤姆在一起时，我的头戴设备让我相当清楚地了解他在想什么、在回应什么。

With Tom, my headset has given me quite a bit of insight into what Tom is thinking, what he's, you know, reacting.

从你的角度看，我是有意识的，还是你在创造我这样的存在？

Am conscious so from your perspective, or are you materializing I would

我想说的是，我所体验到的只是一个化身，它让我得以洞察意识。

say that what I'm experiencing is an avatar, that's giving me insight into consciousness.

我把这个化身叫做汤姆。

I'm calling that avatar Tom.

汤姆把这个化身叫做唐，我能部分但不完全地感受到汤姆的意识体验。

Tom is calling this avatar Don, and I am getting some but not complete insight into the conscious experiences that Tom is having.

我当然不指望能完全了解你脑子里在想什么、有什么感受，但……

By no means do I expect to be, you know, knowing everything that's going on in your mind and your feelings and so But

如果意识是进入我们体内并广播某种信息，那生物学是不是让它得以实现的关键？

if the consciousness is stepping inside of us broadcasting whatever that is, is biology the thing that makes it viable?

神经系统是不是让意识愿意驻留的必要条件？

Is a nervous system the thing that makes it viable as a place the consciousness would want to go?

我觉得你这样表述其实是不对的。

So I would say that that way of putting it is the wrong way to put it.

这又预设了我们是从无意识的物理成分出发，试图找出意识在哪里涌现。

That's assuming, again, that we're in the game of starting with nonconscious physical ingredients and trying to figure out where consciousness can emerge.

但你之前说过，我观察情绪或类似体验的那个空间，正是真正意识所在的地方。

But you said that that space where I'm observing the emotions or whatever is the indication of where the one true consciousness is.

对，因为我的体验是这样的

Right, because my experience

在我头戴设备里的那个汤姆的虚拟形象，细节足够丰富，能让我对你的感受有些了解。

in my headset, the avatar in my headset of Tom, is detailed enough to give me some insights into the experiences you might have.

至于我的猫，我了解得少一些，但也能猜到它可能想要什么。

With my cat, I have less insight, but some insight into what that cat might be wanting.

比如我知道它饿了，会蹭我的腿，诸如此类。

Like, I know when it's hungry and it's it's rubbing against my leg and so forth.

我能猜得出来。

I I can guess.

对于蚂蚁，我得到的信息就更少了。

With an ant, I'm getting less information.

如果我认真研究，或许能理解当一只蚂蚁是什么感觉。

If I really studied, I might be able to understand what it might be like to be an ant.

但如果反过来想，我觉得蚂蚁对我的世界、对我的意识又能了解多少呢？

But but it's if if I turn it around and ask, how much do I think the ant knows about my world, my consciousness?

我猜，如果我走到蚂蚁跟前，抬脚就要踩死它，它根本察觉不到我正在做什么。

My guess would be, I I can go up there to an ant, and I can go like that and be ready to kill it, it wouldn't even know that I'm doing that.

所以，我的世界对它来说几乎等于零。

So how much of my world almost nothing.

我觉得我自己挺复杂的，但蚂蚁眼中的我，大概连个简单模型都算不上，甚至压根没有。

I would imagine that, I mean, as complicated as I feel I am, the ant, its model of me is probably trivial, if it has any model of me at all.

真的非常简单。

It's pretty pretty trivial.

把这个想法反过来想想。

So, take that insight and turn it around.

那些在我看来微不足道的事物。

Things that look trivial to me.

也许是因为，在它们眼里，我就像一只蚂蚁。

Maybe it's because I'm an ant to them.

对吧？

Right?

我对蚂蚁来说微不足道，不是因为我真的微不足道，而是因为它根本看不见我。

I'm trivial to the ant, not because I'm trivial, but because the ant can't see me.

所以，我觉得那些东西很无聊，可能并不是它们真的无聊。

So, the things that I think are trivial, that may that's not because they're trivial.

而是因为我的认知局限太深，没法看清那种意识真正的复杂性。

It's because my headset is just not smart enough for me to see the true complexity of that consciousness.

我这话听起来可能有点极端。

So what I'm saying is is quite radical here.

当我看到一块石头觉得它蠢，其实是我自己太蠢了。

When I see something looks as dumb as a rock, that's because I'm dumb.

不是石头蠢。

It's not that the rock is dumb.

我只能用‘石头’这个词来勉强描述它。

It's the best I can do is come up with a rock.

如果我真的能理解它究竟是什么，我可能会跪下来，惊得说不出话，甚至想膜拜它。

What I'm actually interacting with, if I could actually understand what it was, I might fall down in in amazement before it and and worship it.

所以我看到的只是一个石头，就像蚂蚁看到的我，也几乎完全无法体现我的复杂性。

So but all I can see is a rock, just like all the ant can see of me is almost nothing of of of my complexity.

对。

Mhmm.

所以我们不能搞错。

So we have to not mistake.

我要慢慢说，因为这一点非常关键。

And this is the I'm gonna say this slowly because this is the key point.

别把头戴设备的局限性，当成对现实本质的洞察。

Do not mistake the limitations of the headset for an insight into the nature of reality.

这仅仅是我设备的局限，就这么简单。

It's just a limitation of my headset, period.

我若因为某物看起来笨拙，就认定它真的笨拙，那真是我犯傻了。

It's a dumb move on my part to think that because it looks dumb to me, therefore it is dumb.

不对。

No.

这只是因为你还不够聪明，没意识到那不过是个头戴设备，而任何设备都有它的局限。

That's just you're not smart enough to realize that that was just a headset, and your headset's got limitations.

真正傻的是，你以为自己看到的就是真相。

What's really dumb is to think I'm seeing the truth.

我看到的，只是透过头戴设备的滤镜。

I'm seeing through a filter of a headset.

我需要保持极度的谦逊和低调，去问自己：我的设备有哪些局限？

What I need to do is be very, very humble, very modest, and ask, what are the limitations of my headset?

我该怎么理解我这个设备的局限？

How can I understand the limitations of my headset?

我觉得我看到的是一个有空间和时间的世界。

I think I'm seeing a world of space and time.

哦，原来空间和时间是根本的。

Oh, space and time is fundamental.

真的吗？

Really?

我凭什么相信这个？

Why should I believe that?

我应该保持谦逊，承认空间和时间只是我最初感受到的东西。

I should be very, very modest and say space and time is what I'm first given in my my impressions.

这不过是最初的印象罢了。

That's just the first give.

我们来倒推一下。

Let's reverse engineer.

这背后真正的是什么？

What's really behind that?

再说，如果你总是想想蚂蚁眼中的我是什么样子？

So, and again, I think if you always think about what can the ant see of me?

它几乎什么也看不到。

Almost nothing.

换个角度想想。

Just turn it around.

当我看到某样东西似乎微不足道时，那很可能我就是那个对它而言微不足道的蚂蚁。

When I see something that looks like almost nothing, then I am probably the ant to that thing.

这才是看待问题的正确方式。

That's that's the way to look at it.

所以，对这张桌子来说，我就是一只蚂蚁。

So I'm an ant to this table.

因此，我所接触的任何事物，本质上都是超越我的认知的。

So whatever I'm interacting with is transcendent.

我看到的，仅仅是一张桌子而已。

I'm just getting a table.

这并不是因为现实本身很肤浅。

It's not because the reality is stupid.

而是因为我自己的认知有局限。

It's because I have my limits in my headset.

所以，这就是关键所在。

So that's the the key difference here.

所以，那种试图从愚蠢的组件开始，逐步构建出意识的思路，根本就是走错了方向。

So the whole game of saying, how do I start with stupid components and then build up consciousness is just the wrong game.

那些所谓的愚蠢组件，不过是头戴设备放弃了努力。

The stupid components are just the headset giving up.

而不是现实本身的根本性质。

Not the fundamental nature of reality.

这只是头戴设备在说：我实在没办法了。

That's the headset saying, I give up.

我们完全可以专门做一期节目，就讲这个。

We could do a whole episode just about

就讲这个。

Just about that.

唐纳德，这真是太精彩了。

Donald, this was incredible.

大家怎么才能持续关注你的内容呢？

Where can people keep up with you?

我这边呢，很快六月就要推出一个叫TRACE研究所的机构。

I have, well, pretty soon we're gonna have an institute called the TRACE Institute coming out in June.

几周后，我会在X平台上公布TRACE的官方网站。

So I'll be announcing on on, X, in a few weeks the the the website for for TRACE.

到时候就会上线了。

So that that will be out.

我还出过几本书。

And I do have papers.

我写过一本叫《视觉智能：我们如何创造所见之物》的书，还有一本叫《现实的反面：为什么进化让我们的眼睛看不见真相》。

If you know, I've got a book called Visual Intelligence, How We Create What We See, and also The Case Against Reality, Why Evolution Hid the Truth from Our Eyes.

这两本书，一本是《现实的反面》，另一本是《视觉智能》，另外还有我的个人网站。

So there's two books, The Case Against Reality and Visual Intelligence, and my ex account.

如果你是科研人员，直接去Google Scholar搜我的名字，就能找到几十篇已发表的论文。

And then there if you just Google, you know, on Google Scholar, if you're if you're a scientist, then, of course, you'll know that you can just Google my name, and and there's dozens and dozens of papers that are published.

太棒了。

It's awesome.

太棒了。

I love it.

再次感谢你抽出时间，兄弟。

Thank you again for taking the time, man.

谢谢你，汤姆。

Thank you, Tom.

真的非常感谢。

Really appreciate it.

各位朋友，如果还没订阅的话，记得去订阅一下。

Boys and girls, if you have not already, be sure to subscribe.

那咱们下回再见，朋友们，活出传奇。

And until next time, my friends, be legendary.

保重。

Take care.

再见。

Peace.

咱们来聊聊一个肯定在拖你后腿的习惯。

Let's talk about a pattern that is guaranteed to be killing your progress.

你知道自己该做什么。

Know what you need to do.

你需要保持规律的饮食。

You need consistent nutrition.

每个人都是这样。

We all do.

你需要维生素、益生菌，还有绿色蔬果。

You need vitamins, probiotics, greens.

我们都清楚，自己应该多注意这些。

We all know that we should be doing more of it.

早上一乱，你就把这事儿给忘了。

When your morning gets chaotic, you skip it.

一出门旅行，你就干脆不做了。

When you travel, you skip it.

一旦你的日常被打乱，所有事情都会跟着乱套，这种不规律每天都在拖你的后腿。

When your routine breaks, everything tends to break, and that inconsistency compounds against you every single day.

AG1就是为了解决执行难的问题而设计的。

AG1 is designed to solve the execution problem.

一勺粉，兑八盎司水，搞定。

One scoop, eight ounces of water, and you're done.

一勺里含有七十多种成分：维生素、矿物质、益生元和益生菌、营养丰富的超级食物，过去你需要六七种不同的补充剂，还得完美规划，现在这一切都浓缩在一杯饮料里，三十秒就能搞定。

You're getting 75 plus ingredients, vitamins and minerals, pre and probiotics, nutrient dense superfoods, everything that used to require six, seven different supplements, and perfect planning now happens in one drink that takes about thirty seconds to make.

现在，购买AG1的首次订阅，还能免费获得价值87美元的赠品。

Right now, AG One is giving you $87 worth of free gifts with your first subscription.

包括欢迎礼包、旅行装、维生素D3加K2，还有几种口味的试用装。

You get a welcome kit, travel packs, vitamin D three plus K two, and flavor samples.

点击节目说明里的链接，或者访问drinkag1.com/impact，立即领取这个优惠。

Click the link in the show notes or visit drinkag1.com/impact to claim this offer.