What makes ant colonies robust?

在上一期泰勒·马蒂斯的节目中，你了解到复杂系统是如何从一种行为或状态切换成另一种的。

In our last episode with Tyler Marghetis, you heard about how a complex system can tip from one type of behavior or state into another.

今天，你将听到来自德国波茨坦大学物理与天文系的复杂性科学教授卡罗琳·维森纳的分享。

Today, you'll hear from Karoline Wiesner, Professor of Complexity Science in the Department of Physics and Astronomy at Potsdam University in Germany.

卡罗琳会谈谈系统什么时候不会发生跳跃或失效。

Karoline is going to talk about when systems don't tip or fail.

也就是说，你会听到究竟是什么让复杂系统变得坚韧。

In other words, you're going to hear about what makes a complex system robust.

我们还会通过蚂蚁群体来谈这一切。

And we're going to explore all this by talking about ant colonies.

这里是“简化复杂性”播客，我们在这里探讨复杂系统的基本原理。

This is Simplifying Complexity, a podcast where we explore the underlying principles of complex systems.

这些系统看起来好像超出了我们的理性认知，比如经济、生态，甚至你我和整个自然界。

Systems that seem to defy our rational view of the world, like economies, ecologies, or even you or me.

我是法医工程师肖恩·布雷迪，今天由我来主持。

I'm forensic engineer Sean Brady, and I'll be your host.

今天，我很高兴邀请到卡罗琳·维斯纳来到我们的节目。

Today, I'd like to welcome on the show Karoline Wiesner.

你好吗，卡罗琳？

How are you, Karoline?

我很好。

I'm very well.

谢谢你，肖恩。

Thank you, Sean.

太好了。

Fantastic.

我们要讨论的是复杂系统中的稳健性这个概念。

We're gonna talk about the concept of robustness in complex systems.

为了开启这个话题，我们实际上要从蚂蚁群落开始聊起。

And to kick off the discussion, we're actually going to begin by talking about ant colonies.

蚂蚁群落是我最喜欢的稳健系统之一，原因有很多，我想我们会在这一集里一一探讨。

Ant colonies are one of my favorite system of a very robust system for various reasons, which I think we'll go through in this episode.

首先，也许我们应该说说为什么把蚁群当作复杂系统来讨论。

And to begin with, maybe we should say why we're talking about ant colonies as complex systems.

鲁棒性是复杂系统中非常重要的一部分。

Robustness is such an important part of a complex system.

一个蚁群由成千上万只蚂蚁组成，而单个蚂蚁的认知能力非常有限。

So a colony consists of hundreds of thousands or tens of thousands of ants, And an individual ant has very little cognitive capacity.

当然，设计巢穴并不是由某一只蚂蚁完成的，而是由这成千上万只蚂蚁仅通过它们的互动共同完成的。

Certainly to design colonies, that's not done by a single ant, but it's done by these thousands of ants together only through their interactions.

这就像是，如果你愿意，这上千只蚂蚁的蠕动、探索和行走中涌现出来的东西。

So it's something that, if you like, emerges out of the wiggling and searching around, walking around of thousands of ants.

所有涌现出来的这些东西，都是这些互动的结果，没有别的什么因素。

And all these things that emerge from it are the result of these interactions and nothing else.

而鲁棒性正是从这无数互动中涌现出来的一个特性。

And the robustness is one such thing that emerges out of these many, many interactions.

那么，蚁群具备鲁棒性意味着什么呢？

So what does it mean for an ant colony to be robust?

很简单，你拿一个大蚁群，比如从森林地面上舀起一千只蚂蚁，然后继续观察接下来会发生什么。

Very simply, you take a large colony and say, you just scoop up from the floor of the forest, you scoop up a thousand of them, and then you keep watching what happens afterwards.

蚁群基本上不受影响。

And the colony is pretty much undisturbed.

我是说，它们可能会有点困惑，但还是会继续忙自己的事。

I mean, maybe they'll be a bit confused, but they'll just keep on doing their business.

被我们舀走的那些蚂蚁在蚁群里本来有特定职责，比如说它们是负责觅食的。

And the ants that we scooped up will have had a certain function in the colony, say they were foragers for food.

所以突然之间，蚁群就少了一千只觅食的蚂蚁。

And so suddenly the colony is missing a thousand foragers for food.

但这很快就会得到补充，而且速度非常快，方式也各不相同。

But this will be very quickly repaired and really quickly and by different means.

比如说，其他蚂蚁可能会接手觅食的工作。

So for example, other ants might take over the foraging job.

所以从这个角度来说，蚁群对于被舀走一千只蚂蚁这种干扰是非常有韧性的。

So in that sense, colony is very robust towards the disturbance of somebody scooping up a thousand of them.

而且这种很多复杂系统中都能看到。

And we see this quality in a lot of complex systems.

那鲁棒性到底是怎么定义的？

So what's the definition of robustness?

我们大概会先讲这个，然后聊聊为什么有些东西不鲁棒，有些却鲁棒。

We'll probably jump into that and then we'll talk about why some things aren't robust and why some things are.

那你的定义呢？

So what's your definition?

我最喜欢的鲁棒性定义是：在面对某种干扰时，仍然能维持结构或功能。也就是说，这定义涵盖了非常非常不同的系统。

The definition of robustness that I like best is the ability to maintain structure or function in the presence of some disturbance, which means that, you know, this encompasses robustness for very, very different systems.

说到蚁群，如果我跟人讲蚁群的结构，人家可能会笑，因为蚁群看上去乱糟糟的。

In the case of the ant colony, if I talk about structure in an ant colony, people probably laugh because we look at an ant colony and it looks terribly disorganized.

对啊。

Right.

但里面的功能鲁棒性其实很强。

But there's a lot of functional robustness in it.

我刚刚已经讲过觅食了。

So I already talked about foraging.

觅食是蚂蚁群落中的一项功能。

Foraging is a function of a part of the ant colony.

几千只蚂蚁负责觅食，另外几千只负责照顾幼虫等等。

So a few thousand are designated forages and then another few thousand are designated to take care of the brood and so on.

所以这个功能对干扰非常稳定。

So this functionality is very robust against disturbance.

而在其他系统中，结构本身才是稳定的。

And in other systems, it is the structure itself which is robust.

我们还可以举一个关于结构稳定性的截然不同的例子，就是我们都很熟悉的动物皮毛图案。

And we could take in a very different example of structural robustness, which is something that we all know, which is the pattern of furs of animals.

我想一提到斑马条纹，每个人脑子里都会立刻浮现出那个画面。

And everybody, when I say zebra pattern, everybody has immediately an image, I think.

而且这种图案非常稳定，不管斑马在一天中的什么时候出生，结果都一样。

And it is very robust in the sense that, you know, zebras, it doesn't matter what time of day they're born.

不管它们出生在什么季节。

It doesn't matter what season they're born in.

甚至不管它们出生在草原的哪个地方。

It doesn't matter even which part of the savannah they're born in.

它们都有同样的花纹。

They have the same pattern.

当然，我说同样并不是完全一样，因为每匹斑马的花纹都略有不同。

And of course, when I say the same, I don't mean exactly the same because every zebra has a slightly different pattern.

但总体结构的稳健性是存在的，它由内部非常非常复杂的生化过程来调节。

But the robustness of the general structure is there, and it is regulated by very, very complicated biochemical processes that are happening inside.

事实就是整个结构非常稳健。

Fact is that the structure as a whole is very robust.

你还有一个很好的不稳健系统的例子，就是魔方。

And you have a wonderful example of a system that's not robust, the Rubik's cube.

对的。

Yes.

这是一个绝佳的例子，说明什么是不稳健，而且它本来就不该稳健。

It's a beautiful example of something which is not robust, and it is supposed not to be robust.

这就是游戏的乐趣所在，对吧？

That's the fun of the game, isn't it?

所以核心在于结构的稳健性，或者说不稳健性。

So it's a robustness or the lack of robustness of structure.

因为一开始买魔方的时候，它是很规整的。

Because initially when we buy Rubik's cube, it's very structured.

对吧？

Right?

每一面都是单一颜色。

Every side has a single color.

然后我动手一转，比如转两三下。

Then I go along and I turn, say, I take two or three rotations of it.

我就毁掉了最初那个漂亮的图案。

I've destroyed the initial beautiful pattern.

仅仅转三下就把它弄乱的概率几乎是一。

The probability that I destroyed by just three turns is almost one.

我是说，碰巧转三下就回到原来的结构，那是不可能的。

I mean, by accident, I won't turn it three times and get back the same structure.

然后我把它交给别人，让他们帮我恢复成最初那种每面一色的整齐图案。

And then I give it to someone else and tell them to give me back the original, very ordered pattern of one side, one color.

即使我只转了三次，他们可能也会觉得很难。

And they might struggle even if I've just turned it three times.

要解开它可不是件容易的事。

It's not an easy task to solve.

这正是这个游戏的乐趣所在。

And that's the whole fun of the game.

所以这就是一个结构不稳健的例子，因为一点点扰动就能破坏结构，而要恢复原状却很难。

So this is an example where structure is not robust because a slight disturbance, it destroys the structure and then getting it back is hard.

你用的另一个例子我也很喜欢，就是那种我们觉得不应该稳健，但实际上却出奇稳健的系统。

And the other example that you use, which I really like is the example of a system that we don't think should be robust, but it's surprisingly robust.

就是字母稍微打乱一点的时候，我们还能正常阅读文本。

And that's reading texts when the letters are jumbled up a little.

对。

Yeah.

我觉得这真的很神奇，阅读的时候，我们读一段文字或者只是几个词，扫一眼就能很快理解。

I find this amazing that reading is, you know, we read a text and or we read just snippets, a few words, and we're very quick in even just scanning it.

我们很快就能辨认出这些词。

We're very quick in discerning the words.

有时候，当两个辅音字母被调换时，我们甚至都没注意到字母是乱的。

And sometimes we don't even notice that the letters are scrambled when they are swapping two consonants.

我们就像读正确拼写的单词一样直接读过去了。

We just read it as if the word was written correctly.

所以这里有一个巨大的鲁棒性，这是功能的鲁棒性，也就是处理这些信息的功能。

So there's an enormous robustness, and this is a robustness of function, which is the function of processing this information.

视觉系统接收图像，然后在某个地方对图像进行处理，把它转换成文字，再转换成意义。

The visual system takes in the image, and then somewhere the processing happens of the image, turns it into words, turns it into meaning.

而这种功能容量对于扰动非常鲁棒，比如交换字母或漏掉字母。

And that functional capacity is very robust towards the disturbance, which is swapping letters or dropping letters.

我们知道，阅读时即使元音缺失，我们仍然很擅长推断出那个词是什么。

We know that when we read a text and the vowels are missing, we're still very good in inferring what the word is.

所以它对于接受特定扰动是非常鲁棒的。

So it's super robust to accept particular disturbance.

还有其他一些更难处理的扰动，但大脑经过非常好的训练，能应对那些情况。

And there are other disturbances that are much harder to deal with, but the brain is very well trained to deal with that.

有时候甚至我们自己都没注意到。

Sometimes without us even noticing.

是的。

Yeah.

当我们使用鲁棒性这个词时，我还看到其他出现的词就是稳定性和韧性。

When we use the term robustness, the other terms that I've seen crop are up are stability and resilience.

那鲁棒性跟稳定性和韧性之间是什么关系？

So where does robustness fit in with stability and resilience?

在我看来，稳健性和稳定性是一回事。

I would say robustness and stability to me are the same.

我是学物理出身的，你要是去问其他科学家，他们对稳健性、稳定性或者韧性怎么理解，可能会得到不同的答案。

I'm trained as a physicist, and you might ask other scientists what robustness or stability or resilience are to them, and you might get different answers.

比如说，你去问一个工程师什么是稳定性，他们会说，就像建一座桥，桥不动的时候就是稳定的。

So for example, you ask an engineer what stability and they talk about, say, building a bridge and the bridge is stable when it doesn't move.

就是说，最好别动，至少粗略来说，它真的不应该动。

I mean, better doesn't move, at least roughly speaking, it really shouldn't be moving.

然后你去问化学家，你觉得什么是稳定性？

And you ask a chemist, so what's stability for you?

他们会说，比如说，一个化学反应在发生的时候才是稳定的，也就是说事情一直在发生，一直在变化。

And they say, well, for example, a chemical reaction is stable when it does move, which means things happen all the time, things change all the time.

但变化的速率是稳定的或者说是不变的。

But the rate at which they change is stable or is constant.

所以稳定性在不同的科学和不同的语境下，含义差别很大。

So stability has very different meanings in different sciences and different contexts.

你提到的另一个词，弹性，它更多带有动态的含义。

And the other term that you asked about resilience, that has somewhat more of a dynamical meaning to it.

所以弹性，你看，这就更接近我一直说的了，就是系统在受到扰动之后，能够恢复到扰动之前状态的能力。

So resilience is something, you know, here we're getting closer to what I've been talking about is the ability of a system to, after a perturbation, come back to the state that it was in before the perturbation happened.

比如说你从蚁群里带走那一千只觅食蚁，确实需要一些调整，但这种调整会自然发生，然后蚁群又会有足够的觅食蚁继续运转。

So you take out these thousand forages of an ant colony, there is some readjustment necessary, but that readjustment will happen and then there will be enough forages again for the ant colony to continue.

那能不能把稳健性设计进一个系统里呢？

So can you design robustness into a system?

这就是我那些工程师朋友都会问的问题。

That's the question all my engineer friends would ask.

如果我想让我的系统更稳健，比如像蚁群那样被碰来碰去还能存活，这能做到吗？

If I wanna make my system more robust, like an ant colony that can survive getting nudged around, can that be done?

我给出的答案可能对工程师来说不算特别满意，但我觉得即使是工程师，也能从中有所收获。

The answer I'll give is maybe not quite satisfying to an engineer, but still even an engineer, I think, could take away something from it.

如果我们去思考稳健性的内在机制，就像蚁群那个例子，关键在于蚂蚁之间那数不清的相互作用。

If we think of the underlying mechanisms of robustness, and that was in the case of the ant colony, it's the many, many interactions that we have between these ants.

而且这些互动都是随机的。

And these interactions are stochastic.

它们会随机选一个方向开始走，然后碰到另一只蚂蚁，这次互动就会产生某种作用。

They take some random direction and they start walking and then they run into another ant and this interaction does something.

比如，这只蚂蚁找到了食物，那我就会跟着它往同一个方向走。

For example, this ant has found food, so I'll walk in the same direction.

所以这个系统本身就带有随机性，同时也有一定的多样性。

So there is this stochasticity built into the system and also a certain amount of diversity.

不是每只蚂蚁都在做完全一样的事情。

Not every ant is doing the exact same thing.

所以从这个角度来看，可以把鲁棒性当作寻找解决方案的方式，构建进系统里。

So in that sense, one can think of building in robustness into a system just in the sense of solution finding.

这可能是工程师们能够理解的一种方法。

So maybe that's an approach that an engineer can relate to.

我该怎么找到解决一个问题的好办法？

How do I find a good solution to a problem?

如果大家都朝同一个方向走，那我肯定找不到答案。

Well, I certainly don't find it if everybody starts walking in the same direction.

但如果大家分散到不同的方向，去尝试不同的东西，然后再把结果带回来，那就能找到解决方案，因为系统里加入了这种噪音。

But if people walk into different directions, try out different things and then bring them back, I will find solutions just because I'm adding this noise into the system.

所以系统里的噪音反而有助于找到新的稳定状态或者解决方案。

So a noise in the system helps the system find new stable states or find solutions.

这就是复杂系统的特点。

This is what complex systems do.

它们不会像桥一样静止不动。

They never stand still like a bridge.

它们总是在运动。

They are always moving.

正因为这种无序的运动，当出现干扰时，系统往往能因为这种内置的噪音而找到新的解决方案，从而化解问题。

And because of this movement being disordered, a disturbance can be resolved because a new solution will be found just because of this built in noise.

听着，我觉得有意思的是，你说的本质上是一种自下而上的行为。

And I think I mean, the fascinating thing about that is that you're talking about a bottom up behavior, essentially.

你知道，我觉得这是系统涌现出来的一种特性。

It's you know, I presume it's an emergent property of the system.

这种稳健性也是一种涌现特性。

The robustness is is an emergent property.

因为从工程学角度，如果想用自上而下的方式来设计稳健性的话，那在设计系统时，怎么让它更稳健呢？

Because in engineering terms, if we would be thinking of designing in robustness by taking a top down approach, when we design the system, how do we make it more robust?

而我们的做法就是增加冗余，在一套备份系统之上再加备份系统，以确保安全。

And our approach would be add redundancy, add backup systems upon backup systems to make it safe.

但你现在说的，其实是完全不同的一件事。

But what you're talking about there is really quite a different thing.

这区别很大。

It is very different.

不过这也不意味着冗余不重要，因为蚁群的强健也来自于成千上万只蚂蚁的冗余。

That doesn't mean that redundancy isn't also important because part of what the ant colony what makes it robust is also the redundancy of these thousands and thousands of ants.

我意思是，比如说如果你只抓几百只行军蚁的话，行军蚁基本上都是瞎的。

I mean, for example, if you take just a few 100 army ants, army ants are pretty much blind.

假设你把这100只行军蚁放在一架飞机上，周围几乎什么都没有。

And you put these, say, 100 army ants on a plane and there's not much else there.

那里没有拓扑结构。

There's no topology.

比如，那只是一张平坦的纸面。

That's just a flat piece of paper, for example.

当然，它们之间还是会相互作用的。

Of course, they will interact.

它们会互相撞到一起。

They will bump into each other.

但很快，它们就会排成一个圈，然后沿着这个圈一直走。

But eventually, very quickly, they will just form a circle and they will walk around in the circle.

一只跟着另一只，最后首尾相接形成一个闭环。

One end is following another end, which is following another end and circle closes.

它们会一直走下去，但因为只是在原地转圈，所以永远找不到食物。

And they will keep walking, but they never find any food because they're just walking in the same circle.

那就意味着它们很快就要死了。

That means they will pretty soon die.

哇。

Wow.

这里缺了两个东西，但我们在说的是冗余。

And there are two things missing here, but we're talking about redundancy.

如果蚂蚁更多的话，这个系统会更稳健。

And if there had been more ants, then the system would have been more robust.

我能从一个蚁群中捞起一千只蚂蚁，但对蚁群伤害不大，因为至少还有一万只蚂蚁能顶替它们的位置。

The fact that I can scoop up a thousand ants from a colony doesn't hurt the colony too much is because there are at least another 10,000 that could take that place.

所以冗余确实起作用，但并不是唯一的因素。

So redundancy does play a role, but it's not the only thing that plays a role.

光是冗余是远远不够的。

Redundancy alone is simply not enough.

接下来我们就说说你真正想聊的那两个关键点，它们才是真正能带来大不同的地方。

So let's get on to the two of the key points you want to talk about that really does make a big difference.

第一个就是反馈，另一个就是你之前提到的，噪音的作用。

And the first of those is feedback, and then the other is one you've alluded to already, which is the role of noise.

那我们说的反馈是什么意思呢？

So what do we mean by feedback?

反馈就是一次互动会影响未来的互动。

Feedback is when an interaction influences future interactions.

其实很简单。

So it's very simple.

比如说，我在街上遇到一个陌生人，他们看起来很友好，那下次我再遇到他们，可能会请他们喝杯咖啡什么的。

Say, I meet a new person on the street and they seem friendly, then next time I meet them, maybe I'll invite them for a coffee or something.

然后也许从那里，一段友谊就产生了。

And then maybe from that, a friendship emerges.

所以这就是之前互动的反馈。

So this is a feedback from previous interactions.

而当我遇到一个人，他们看起来很不友好，甚至有点敌意，那这会影响我未来的行为，比如我可能会避开那条街，因为我知道他们住在那里，诸如此类。

Whereas when I meet someone and they seem very unfriendly, maybe hostile, then that will influence my future actions in the sense of, well, maybe I'll try to avoid that part of the street because I know they live there, something like that.

所以反馈确实一直存在于我们的生活中，我们总是在根据过去的经验行事。

So certainly feedback is present constantly in our lives, and we're constantly acting on previous experiences.

而且反馈对复杂系统来说至关重要，因为我们之前说过，蚁群的所有结构、功能等等，都是从蚂蚁之间无数次的互动中涌现出来的。

And feedback is so essential for complex systems because we said earlier that ant colonies, all the structure, functionality, and so on emerges from these many, many interactions between the ants.

而这些互动的一个关键就是蚂蚁的化学信号。

And one key to these interactions is chemical signaling in ants.

所以当它们遇到另一只蚂蚁时，能闻到对方的气味。

So when they meet another ant, can sort of smell the other ant.

它们能闻到，这里说“闻到”其实是个比喻，但它们确实能感知到其他蚂蚁。

They can smell, and smell is more metaphorically here, but they can detect the other ants.

比如，是不是同一个蚁群的？

For example, is it of the same colony or not?

它是不是刚去过食物源？

Has it just been to a food source or not?

如果这两个问题的答案都是肯定的，那么这只蚂蚁很可能就会跟着它刚闻到的这种化学物质——信息素——走。

So when both of these are answered with yes, then this ant is probably going to follow that pheromone, this chemical that it has just smelled.

这样一来，蚂蚁在互动的过程中也会沾上这种化学物质，也就是食物源留下的气味。

And that way, from this interaction, the ant also takes on the chemical, the food source.

于是越来越多的蚂蚁会闻到这种信息素，然后循着气味找到食物源。

And so more and more ants will smell this pheromone, will go to the food source.

就这样，形成了一条完整的蚂蚁路径，它们往返于食物源和巢穴之间。

And that way, a whole trail of ants is coming about that is moving towards the food source and back to the nest.

这就是反馈机制。

So that's feedback.

而反馈正是连接蚂蚁这种看似杂乱无章的随机游走与实际涌现出的结构之间的纽带。

And the feedback is the link between this maybe noisy random walking around of ants and the actual structure that emerges from it.

在这些互动中，你会区分正反馈和负反馈吗？

And do you distinguish between positive feedback and negative feedback in these interactions?

当然会。

Very much.

这两者带来的后果截然不同，虽然现象相似，但名字起得挺误导人的。

These are two very different consequences, similar phenomena, very misleading names, by the way.

正反馈听起来好像是好事。

Positive feedback sounds like it's a good thing.

可能确实是好事，但有时也可能是很糟糕的事。

And it can be, but it can also be a very bad thing.

而负反馈反过来，实际上可能反而是好事。

And negative the other way around could be, you know, actually a good thing.

负反馈在很多情况下甚至是工程设计中常被内置的一种机制。

Negative feedback is often a phenomenon which is even built in at times when you think of engineering things.

负反馈是一种稳定机制。

Negative feedback is stabilizing feature.

比如拿工程上的桥梁来说，当火车经过时，桥会产生振动。

So of course when we take an engineering example of a bridge, you have a vibration comes about because say a train moves across that bridge.

我们希望这种振动能逐渐减弱，甚至完全消失。

You want that vibration to subside, to disappear.

有些案例比如军队齐步走过桥，步伐一致，就会引发共振，这其实是一种自我增强的正反馈，振动越来越大，可能对桥造成灾难性后果。

And there are these examples of, say, armies marching across bridges and they march in step, and so they're causing these vibrations and then there's positive feedback in the sense of self enhancing feedback that increases the vibrations and that can be catastrophic consequence to the bridge.

更准确的说法是自我增强反馈，而负反馈则是起抑制作用的。

A better term is self enhancing feedback and negative feedback is dampening.

所以系统会自我调节，如果你愿意这么理解的话。

So the system self regulates, if you like.

这两种反馈在复杂系统中都存在。

And both of these are present in complex systems.

蚁群就同时具备这两种反馈。

The ant colony has both of them.

正反馈，也就是自我增强反馈，在觅食行为中很典型。比如，少数蚂蚁偶然发现了好的食物源，然后它们带着信息素回来，这些化学气味会吸引更多蚂蚁沿着同样的路径前往，带回更多食物和来自食物源的化学物质，然后更多蚂蚁前往，如此循环往复。

The positive feedback, self enhancing feedback is in the case of foraging, for example, where a few ants happen to hit on a good food source, and then they come back with these pheromones, with these chemical smells, more ants come, follow the same trail, bring back more of the food and also the chemicals that come from that food source, more ants go, and so on and so forth.

这就形成了一种正面的自我增强现象，最终导致蚂蚁形成庞大的行进队伍。

So this is leading a positive self enhancing phenomenon that leads to these big, big trails of ants.

那在蚂蚁的情境中，负反馈是什么样的呢？

And the negative feedback in the ant situation?

你能举个例子吗？

Have you got an example of that?

任何东西太多了通常都不好。

Too much of anything is usually not good.

比如说，如果所有蚂蚁都被派去觅食了，那就没有蚂蚁留下来照顾幼蚁了。

So for example, if all the ants were recruited to do foraging, then there wouldn't be any ants left to take care of the brood, let's say.

所以需要一些负反馈来控制，不能所有的蚂蚁都离开巢穴去找食物。

So there needs to be some negative feedback that not all the ants are recruited away from the nest to go towards food source.

这个负反馈确实存在，因为有其他机制在干预，比如，如果所有蚂蚁都出去觅食，巢穴所在地——也就是幼蚁所在的地方——就会空出来。

And that is present because other mechanisms are interfering with it, which is, for example, if all the ants were to go away to forage, the site of the nest, which is the brood site, would be empty.

而蚂蚁是随机移动的，空出来的空间相当多，它们很可能会误打误撞进入幼蚁区，然后开始帮忙干活。

And because ants are randomly moving around, there will be lots of space for them to accidentally move into the brood site and they would start helping out.

这说法有点粗略，但确实是个负反馈的例子，能平衡整个蚁群的需求。

So that's somewhat hand waving, but it is an example of a negative feedback, which will then balance out the needs of the colony.

除了这个正负反馈，我们再看一个非常不同的例子，比如经济领域。负反馈就是传统的供需关系，对吧？

This positive and negative feedback to go to a very different example like the economy, the negative feedback is that traditional sort of supply and demand, isn't it?

也就是当供应增加时，价格就会下降，因为东西变多了。

That as supply goes up, then the price decreases because there's more of it around.

我们就会减少生产量。

We stop producing it much.

这样就消除了系统中的波动。

So it just takes out that movement in the system.

但随后我们也会遇到正反馈，比如大家都变得有点疯狂，出现房地产泡沫或者其他大宗商品的泡沫，因为觉得它更有价值就都想买，价格就涨了，但长期来看这是不可持续的。

But then we get those positive feedbacks when we all go a bit crazy and we get a housing bubble or a bubble in some other commodity where we all wanna buy it because we perceive it's got more value, so the price goes up for us, and that's not sustainable in the long term.

没错。

Correct.

这正是正反馈自我强化的两个典型例子。

Those are two prime examples of positive self enhancing feedback.

当然，通货膨胀也是其中一个例子，还有供需带来的稳定负反馈，以及价格作为其中的一部分。

Inflation is another one, of course, and the stabilizing negative feedback of supply and demand and then the prices that are part of it.

那接下来谈噪音。

So then noise.

所以反馈之后，你想讨论的第二件事是噪音。

So noise is the second thing you want to talk about after feedback.

那是什么意思呢？

So what do we mean?

大概先讲一下复杂系统里的噪声是什么意思。

Start probably what we mean by noise in complex systems.

噪声，有很多不同的术语来描述它。

Noise, lots of different terms that are used for it.

噪声、随机性、波动，它们指的是同一个东西，就是如果我们回到蚁群，蚂蚁很少走直线。

Noise, randomness, fluctuations, they all refer to the same thing, which is that if we go back to the ant colony, ants rarely walk in a straight line.

它们总是走左一点，走右一点，然后直走一会儿。

They always go a bit left and a bit right, go straight for a while.

我们看这条路线，会觉得它是随机的。

We look at this path and we perceive this as random.

而且这种随机性不只是个缺陷，而是一种特性。

And the randomness is not just there as a bug, but it's actually a feature.

我的意思是，如果我们回到我之前提到的那个行军蚁的例子，把一百只蚂蚁放在一块平面上，它们会绕圈走，最后死掉。

I mean, if we go back to this example I mentioned on these army ants that you put them on a plane, and if you take just a 100 of them, they will walk around in a circle and die.

它们有反馈机制，所以会跟着另一只蚂蚁走，那只蚂蚁又跟着别的蚂蚁走，但环境噪声却被忽略了。

Because what they're missing, they have the feedback, so they might follow another ant, which is following another ant, but they're missing the noise.

这里的噪声，就是大量蚂蚁互相碰撞产生的。

And the noise here is related to lots of ants bumping into each other.

这几乎就像一种干扰信号。

It's almost like a disturbance.

一只蚂蚁正沿着路径走，突然撞上另一只带有信息素的蚂蚁，而那个信息素突然变得很有意思。

An ant is following a path and it bumps into another ant that has some pheromone that is suddenly interesting.

这就意味着第一只蚂蚁会偏离原有路径，做点不同的事。

And that means the first ant will leave its original path and do something different.

那后果还不明确。

The consequence of that is not clear.

可能是积极的，也可能是消极的。

It can be positive or negative.

说不定它从一个真正好的食物源，被吸引到了一个很不好的食物源，做了个糟糕的决定。

Maybe it's being drawn away from a really good food source to a really bad food source and it's making a bad decision.

但问题是，当这些稍微随机的相遇导致稍微随机的决策不断发生时，如果系统足够大，总体上系统最终会找到一个好的解决方案，对蚁群来说，就是让蚁群活下去。

But the thing is, when many of these somewhat random encounters leading to somewhat random decisions are being taken, if the system is large enough, on the whole, the system will actually find a good solution to its problem, which in this case is keeping the ant colony alive.

那蚁群的目标是什么？

What is the ant colony's purpose?

目标就是活下来。

It's to stay alive.

通过这些随机互动，它们找到了出奇稳妥的解决方案。

And through these random interactions, they're finding surprisingly good solutions to problems.

比如说，有一个食物源，但蚂蚁得跨过一道沟，或者地面上有道裂缝才能过去。

So for example, there is a food source where they ideally would have to cross some gorge or some gap in the floor.

一只蚂蚁过不了裂缝，但两只蚂蚁就可能开始纠缠在一起了。

A single ant can't cross a gorge, but two ants will maybe start to entangle.

那就算是一座桥的开端吧。

That would be the beginning of a bridge.

然后更多的蚂蚁会踩在它们身上过河，这种事它们经常干。

And then more ants will come and walk over them, which they do all the time.

然后另一只蚂蚁也会开始和这个纠缠在一起。

And then another ant will start to get entangled with this.

就是通过这些看似随机的互动，很多蚂蚁并不会这么做，但有些会，最终它们会搭成一座桥，而这座桥实际上会缩短通往食物源的路径。

And just through these somewhat random interactions where many ants will not do this, but some will, and eventually they will have a bridge, and that bridge will actually shorten the path towards the food source.

同时，这也会占用劳动力。

At the same time it will also take up labor.

所以，桥上的那些蚂蚁就不能再去觅食了。

So some ants that are in the bridge can't do the forging anymore.

所以，这里需要找到一个最佳平衡点，在觅食和通往食物源的短路径之间。

So there is an optimum balance to strike here between forging and having a short path towards the food source.

蚂蚁们找到的最佳平衡点， surprisingly 接近数学上的最优解。

And the optimum ants find is surprisingly close to a mathematical optimum.

这不是因为有人设计了这座桥，而是因为蚂蚁之间大量的互动。

It's not because someone has designed this bridge, but it is because of these many, many interactions between these ants.

事实上，这是噪声中涌现出的一个非常令人惊讶的特征。

It's a very surprising emergent feature of noise, in fact.

我喜欢你说的这一点：这些系统的稳健性，几乎就是一场拔河比赛，一边是对结构的需求，因为反馈机制驱动着结构，另一边是噪声带来的足够多的变化，防止系统陷入循环，防止它进入一个再也出不来的状态。

What I love about what you're saying is that the robustness of these systems is this almost this tug of war between the need for structure because of the drive for structure, because of the feedback you have happening, but you're just keeping enough variation in there from the noise to stop it essentially getting into a loop, to stop it getting into something that it can no longer get back out of.

本质上就是噪声——如果我们不研究复杂性科学，可能会觉得噪声是坏事——但它恰恰提供了系统适应所需的变化。

It's essentially the noise, which we probably would think of as a bad thing if you're not in the complexity space, that's providing the variation you need for the system to adapt.

噪声的这种双重特性，它似乎在拉扯着完美的结构，比如我们可能想建造的通往食物源的完美路径，但你永远不会看到那样的路径。

This double feature of noise, it's sort of tugging away from this perfect structure which maybe we would like to build, perfect trails towards a food source, you'll never see them.

原因在于，虽然噪声看起来像是个缺陷，但它既是最初产生结构所必需的，也是当结构受到轻微干扰时——比如被其他动物攻击，比如说食蚁兽喜欢吃蚂蚁——修复结构所必需的。

And the reason is that although the noise seems like a bug, it is required both to generate the structure in the first place, and then also if the structure is slightly disturbed, attacked by some other animal, let's say an ant bear that likes to eat ants.

或者你抓走一千只蚂蚁，把它们带走。

Or you grab a thousand ants and pull them out.

没错。

That's right.

或者一个复杂性科学家过来，舀走一千只蚂蚁。

Or a Complexity Scientist comes along and scoops up a thousand ants.

噪声就是用来修复的。

The noise is there to repair.

所以噪音的存在是为了纠正错误，因为漏掉一千个觅食点算是系统里的一个失误。

So the noise is there to correct errors because missing a thousand forages is an error in the system.

因此，系统必须具备纠正这种错误的能力。

So the system needs to be able to correct for that error.

而这正是噪音和反馈这对组合能够做到的事情。

And that is what this tandem of noise and feedback is able to do.

这真是太神奇了。

And that's fascinating.

所以你的意思是，如果没有这些噪声，我们建出来的结构会变得过于僵化。

So you're saying that if we didn't have the noise, we build a structure and that structure would become too solid for whatever better word.

这样一来系统就缺乏韧性了，因为如果不去破坏它或者从中取出一部分，它就缺少足够的变化来适应和修复自身。

And then that leads to the lack of robustness because if we didn't break or pull a piece out of that, it doesn't have enough variation to adapt and fix it.

没错。

Correct.

因此，缺乏噪声就意味着缺乏适应能力，因为这种适应并不是来自某个中央控制指令，比如‘你往左走，你往右走’，然后一切都会好起来的。

So the lack of noise is then a lack of ability to adapt because the adaptation is coming not from some central control element that says, you go left, you go right, and then this is all gonna be okay.

但适应性正是从所有这些分散的行动中产生的。

But the adaptation is coming out of all these decentralized actions.

那真是令人着迷。

That is fascinating.

Karoline，非常感谢你来做客我们今天的节目。

Karoline, thank you very much for being a guest on the show.

这是我的荣幸。

My pleasure.

谢谢你，肖恩。

Thank you, Sean.

感谢收听《为你简化复杂性》，我们与来自世界各地的专家一起探讨复杂性科学的核心概念。

Thanks for listening to Simplifying Complexity, where we look at the key concepts of complexity science with expert minds from across the world.

比如：涌现、自组织、适应性、网络、标度、临界点，等等。

Concepts like emergence, self organization, adaptation, networks, scaling, tipping points, and much more.

本期播客由Brady Heywood和Wavelength Creative制作。

This podcast was produced by Brady Heywood and Wavelength Creative.

为了确保你不会错过任何一集，记得在你的播客应用里订阅或关注这个节目。

To make sure you don't miss an episode, be sure to subscribe to or follow the show in your podcast app.

我是肖恩·布雷迪，我们下期节目见。

I'm Sean Brady, and I'll see you in our next episode.