视频库 / NO.052
--:--
点击播放，这里会跟随视频显示当前句的中英字幕。
第 52 期 · 核心追问 Ⅰ·05「整体能大于部分之和吗？」
MIT Godel Escher Bach Lecture 6
节目发布 2012-12-02章节 · 点击跳转视频
0:00
回顾：哥德尔编号与不完备定理
5:34
能否把 G 加为公理生成全部数学
9:33
太空服玩笑：未被察觉的同构
12:32
人是图灵机吗：彭罗斯与明斯基
18:08
从齿轮到操作系统：计算机的描述层次
21:40
神经元到灵魂：大脑的层级塔
27:58
从少量构件到涌现：语言与蚁群
31:22
红细胞问题：自我在哪一层
38:10
内省为何困难：演化的设计
44:20
马斯洛金字塔与牛顿的反例
56:00
物理学家、工程师与计算机科学家
61:48
程序演示：简单规则的涌现行为
71:33
缠结的层级：软件能控制硬件吗
77:25
观察者与被观察者：我存在吗
87:55
宇宙只能模拟自己：描述的必然近似
96:18
鲁棒性、关键节点与自我的连续性
101:54
尾声：无穷的悖论与谁在观察
本期小问 · 档案清单
31:22
什么是「我」？
27:58
整体能大于部分之和吗？
12:32
机器会思考吗？
87:55
我们真的在理解世界吗？
本期讲者
JustinMIT 学生自组织课程《哥德尔、埃舍尔、巴赫》的主讲之一，曾在剑桥求学，负责本讲的哥德尔回顾、描述层次与演化/需求层次讨论。
Curran课程的另一位主讲，编写并演示了库仑力、n 体、交通流、脂质膜等多个涌现现象的模拟程序，主持后半段关于涌现与自我的讨论。
课堂学生参与讨论的 MIT 学生，提出「人是否是图灵机」「软件是否真的控制硬件」「细胞更替之间我在哪」等关键问题，推动了本讲的辩论。
01回顾：哥德尔编号与不完备定理
0:00
the following content is provided under a Creative Commons license your support will help MIT open courseware continue to offer high quality educational resources for free to make a donation or view additional materials from hundreds of MIT courses visit MIT opencourseware at ocw.mit.edu all right welcome back everybody um first of all I want to apologize to the virtual world out there that we didn't get last lecture lecture on tape um so I want to do a quick cap honestly I mean no more than two or three minutes on what happened um and that's why I put all the effort in putting it up on the board first um so remember we're dealing with the formal system here and here I'm just going to denote this formal system as F recently we've been interested in typographical number Theory TNT um and we carried out this process of of encoding symbols of f through this girdle numbering process essentially giving equals you know a number like 555 and things like this um so we have now symbols of f corresponding to a small
以下内容依据知识共享许可协议提供，您的支持将帮助 MIT 开放课程继续免费提供高质量的教育资源。如需捐赠或查看来自数百门 MIT 课程的更多资料，请访问 MIT 开放课程网站 ocw.mit.edu。好，欢迎大家回来。首先我要向线上的朋友们道个歉，上一节课我们没有录到像。所以我想快速回顾一下，说实话，也就两三分钟，讲讲上次发生了什么。这也是为什么我先花了那么多功夫把这些写在黑板上。记得我们这里处理的是一个形式系统，我在这里就把这个形式系统记作 F。最近我们一直关注的是「排版数论」TNT。我们通过哥德尔编号这个过程，对 F 的符号进行了编码，本质上就是给「等号」之类的符号赋一个数字，比如 555 之类的。所以我们现在有了F 的符号，对应于全体自然数的一个小子集。然后我们就得到了这样一个对应关系，实际上是
1:09
subset of all natural numbers um we then have this correspondence this actually exact mathematically precise isomorphism between strings of F and subset of all numbers which is the girdle number of some string the strings that we have over there um we then take our rules aums and rules of f remember these were formal recursive operations on strings in our in our formal system and we arithmetization be equivalent to taking 10 times blah blah blah blah blah this number blah blah blah blah blah divide by blah blah blah Chinese remainder theorem blah blah blah um and we're going to actually be able to do the same symbol shunting we would do with like the Miu system or with the typographical number Theory system system and just do it in the form of numbers and this is why with girdles incompetance theorem we need formal systems strong enough that Encompass number theory in order to do this process so we then have strings which are statements of f which correspond to numbers that are f- producible we then kind of do this leap
一个数学上精确的同构：F 的字符串与某个数字子集之间的对应，这些数字就是某个字符串的哥德尔数——就是我们那边写的那些字符串。接着我们取 F 的公理和规则，记住这些是我们形式系统里对字符串的形式化递归操作，而我们把它「算术化」之后，就等价于乘以 10 再怎样怎样，这个数再怎样怎样，除以什么什么，中国剩余定理什么什么的。然后我们实际上就能做和MIU 系统或者排版数论系统里一模一样的符号搬运操作，只不过是以数字的形式来做。这就是为什么在哥德尔不完备定理里，我们需要足够强、能涵盖数论的形式系统，才能完成这个过程。于是我们就有了 F 中作为陈述的那些字符串，它们对应于「F 可产生」的数。然后我们做了一次跳跃，
2:18
outside of the system we have meta F right we're taking statements about strings of of our formal system and these correspond to this corresponds to number Theory taking statements about numbers which are F producible right which are F numbers um and an example of this was you know in number Theory we say okay so 641 is prime or you know six is a perfect number and we just denote this property perfect number and we can give it we can make it a property with a free variable and he's like all right is five a perfect number no is six a perfect number yes um and just we can create a new prop property called primness which corresponds to provability and then we essentially then have this problem of determining whether a particular string is a theorem of f is equivalent to establishing whether or not a number is Prim so I wanted to quickly recap going back over here so that the essential steps in in girdles proof so we
跳到系统之外，我们有了元 F，也就是我们在讨论关于形式系统中字符串的陈述，而这些对应于——这对应于数论——也就是讨论关于那些「F 可产生」的数的陈述，也就是 F 数。一个例子就是，在数论里我们会说，好，641 是素数，或者六是完全数，我们就把这个性质记下来，「完全数」，我们可以把它做成一个带自由变量的性质，然后就是：好，五是完全数吗？不是。六是完全数吗？是。同样地，我们可以造出一个新的性质，叫做「哥德尔可证性」（primness），它对应于可证明性。于是我们就把「判定某个特定字符串是否是 F 的定理」这个问题，等价转化为「判定某个数是否具有 primness」。所以我想快速回顾一下，回到这边，看看哥德尔证明中的关键步骤。所以我们
3:28
arithmetization inference and deduction operations on numbers so then we make this property of provability and we equate it through this exact precise isomorphism which gerle discovered to a probab property of primness we then turn the operation of coining so coining was you know when preceded by itself in quotations yields a full sentence of course we could play around with this like so snow is white snow is white didn't mean anything but when we fed that operation that function itself into that variable for when preceded by itself in quotations yields a full sentence when preceded by itself in quotations yields a full sentence created a full sentence which was itself self-referential and it was a fix point it was which remember when you have a function which was in our case a property of coining um or it could have been just simply like multiplying by 2x a fix point is something which when you feed it into your function you get the same thing back out so we took coining into an operation on
把推理和演绎算术化为对数的操作。然后我们造出「可证明性」这个性质，并且通过哥德尔发现的这个精确同构，把它等同于 primness 这个性质。接着我们把「引用式自指」（coining）这个操作——所谓 coining，就是「当被置于其自身的引号之前时，产生一个完整的句子」。当然我们可以随便玩这个，比如「雪是白的」「雪是白的」——那没什么意义。但当我们把那个操作、那个函数本身代入那个变量位置，就成了：「当被置于其自身的引号之前时，产生一个完整的句子」「当被置于其自身的引号之前时，产生一个完整的句子」。构造出了一个完整的、自我指涉的句子，而它是一个不动点——回忆一下，当你有一个函数时，在我们的例子里它是哥德尔配数的一个性质，嗯，它也可以就是简单地乘以 2x，不动点就是这样一种东西：当你把它输入进你的函数，你得到的输出还是同一个东西。所以我们把配数变成了对
4:33
numbers our fixed point of coining gave us inherently self-referential sentence and then we were able to describe not directly spell out but describe a formula which felt like this statement here when fed its own gral number yields a non-pr number when fed its own gral number yields a nonpr number which describes this big thing called G and G is essentially the reason why number theory is incomplete it's what breaks The mathematician's Credo this idea of true if and only if provable and with the production of G we were able to establish that although all provable things are true not necessarily all true things are provable and this was a very important idea it's the take home message of the course um we connected this briefly to the halting problem which is just this idea of you can't have a magical machine which you can take in a program and an input and it'll tell you whether or not it'll terminate just like you can't have a magical machine which says give me the girdle number for some statement in number Theory and I'll
数的一个运算，配数的不动点给了我们本质上自我指涉的句子，然后我们就能够描述——不是直接写出来，而是描述——一个公式，它读起来就像这句话：当把它自己的哥德尔数输入进去时会产生一个非素数；当把它自己的哥德尔数输入进去时会产生一个非素数。这描述的就是那个大家伙 G，而 G 本质上就是数论不完备的原因，它打破了数学家的信条，也就是“真当且仅当可证”这个想法。有了 G 的构造，我们就能确立：虽然所有可证的东西都是真的，但并非所有真的东西都必然是可证的。这是一个非常重要的想法，是这门课的核心要点。嗯，我们还简要地把它和停机问题联系了起来，停机问题就是这样一个想法：你不可能拥有一台神奇的机器，你把一个程序和一个输入喂给它，它就能告诉你它是否会终止；就像你不可能拥有一台神奇的机器，你说“给我数论中某个陈述的哥德尔数，我就
02能否把 G 加为公理生成全部数学
5:34
tell you whether or not it's true or false just by detecting whether or not it's Prim or not Prim um these things are inherently impossible and undecidable and this was kind of the shaking of the foundations of number Theory which we which we wanted to do um but that's all I wanted to say about this uh I'm I'm I wanted to move on um and now think about more enjoyable things because number Theory great I wanted more just the take-home message of this idea of having things which are true but not within our system right in the idea that we have to jump outside the system yes Latif we do like something we can have for system then we get all the G statements and then get foral systems in which those holds and so by doing that we generate more formal system and then get like their truth statements true St statements about them that are forc in that system for it to be consistent do this for all the systems you can come up with can we like by that way like create all the mathematics there is right so
告诉你它是真还是假”，仅仅通过检测它是不是素数。嗯，这些东西本质上是不可能的、不可判定的，而这某种程度上就是对数论根基的撼动，这正是我们想要做的。嗯，不过关于这个我想说的就这些了，呃，我我想往下讲了，嗯，现在来想些更有趣的东西，因为数论虽然很棒，但我更想要的只是这个核心要点，这个关于“存在这样一些东西”的想法——没错，但不是在我们这个系统之内，对吧，也就是说我们必须跳到系统之外，是的 Latif，我们确实这么做，比如我们可以先有一个系统，然后我们把所有的哥德尔语句都拿到，然后再去找那些让这些语句成立的形式系统，通过这样做，我们就生成了更多的形式系统，然后再得到关于它们的真语句，也就是在那个系统里为了保持一致性而必须成立的真语句，对你能想到的所有系统都这么做，我们能不能用这种方式创造出全部的数学呢，对吧，所以
6:34
essentially this idea of going ahead and giving girdle numbers for our axioms and our rules of inference and then just being able to essentially generate all of mathematics by just feeding it to a computer and let it carry out these operations on numbers right so it would we could produce a computer which would produce provable things in that direct fashion but it'd be incredibly inefficient right like one of the remarkable things about humans in the human intellect is that we're able to essentially jump several nodes down the tree and not work we don't think on the level of formal deductions on operations of of symbols right um and also once again like that's the whole point of girdles and completeness Theorem is that we can't produce all mathematics I but you can like certainly try to like dve like statements that are almost similar to him to to what he says and then gets systems in which those statements which are for in some of system hold for that system and then try to explore that out
本质上这个想法就是给我们的公理和推理规则都编上哥德尔数，然后只要把它喂给一台计算机，让它对这些数字执行这些运算，就能生成全部的数学，对吧，所以我们确实可以造出一台计算机，用这种直接的方式产出可证明的命题，但那会极其低效，对吧，比如说人类和人类智力最了不起的一点就是，我们能够基本上就是在树上往下跳过好几个节点，而不是——我们思考时并不是在形式推导、符号操作这个层面上进行的对符号做操作，对吧，嗯，而且再说一次，哥德尔不完备定理的整个要点就在于我们没法产生出全部的数学，但你当然可以试着去推导那些几乎类似的陈述类似于他所说的那些东西，然后得到一些系统，在这些系统里那些对某个系统而言的陈述是成立的对那个系统成立，然后再去探索那个系统，看看它能给你带来什么样的数学。好，那么这个思路是不是就是
7:32
system to see which mathematics gives you okay so is the idea then that we essentially take our our our our G here and we let that be essentially an axium of a new system and then we just like get a system in which that thing can be derived to be true and to be speaking about another system but not the system which it right exactly so essentially we would have to it's almost trying to like make our our system complete by considering ing a system where G is derivable right exactly so no I mean that's totally key and mathematicians gave exactly the same point after girdle did this but what girdle showed is that even if you were to tack on G as part of your new system you could then create an analogous G Prime which was incomplete in that system I know I know that that's the problem but since there's a lot of these like um incom things all over mathematics if we can just F them and create new system that like um sort of support them then we can like like have like other mathematics that are not
我们基本上把我们这里的 G 拿过来，让它成为一个新系统的公理，然后我们就得到一个系统，在其中那个东西可以被推导为真，而且它谈论的是另一个系统，而不是这个系统对吧，没错，所以本质上我们几乎是在尝试让我们的系统变得完备，方法是考虑一个能推导出 G 的系统，对吧？没错，不，我的意思是这一点非常关键，数学家们在哥德尔做出这个之后也提出了完全一样的观点。但哥德尔证明的是，即使你把 G 加进去作为新系统的一部分，你也能构造出一个类似的 G'，它在那个系统里同样是不可判定的。我知道，我知道，那就是问题所在。但既然数学里到处都是这类不完备的东西，如果我们能把它们修补好，创造出能够支持它们的新系统，那我们就能得到一些原来那个系统覆盖不了的其他数学，所以这就相当于
8:36
covered by the other system so it's just like a way of exploring the mathematical landscape well no I mean yeah we're doing that all the time right like and but it's it's not necessarily that we can just start out and somehow get the complement of of provable state of not provable statements very easily like I mean that's that's essentially what mathematicians do that's why mathematicians never go out of a job is that we we fundamentally need to think about and create new Concepts and make new definitions and then explore the deductions of our of our new assumptions and axioms which weren't covered by the old you know system involving number Theory we need to we need to add new stuff on there that can't be done in a mechanical process it requires human intelligence um but I mean that's good like uh if you want to if you want to talk about this a little more we should do it after class um because we got a lot to pack in today but let me know if I haven't answered your question fully um we'll do it
一种探索数学版图的方式。呃，不，我是说，对，我们一直都在这么做，对吧，就像但也不一定就是说，我们可以直接从头开始，然后以某种方式得到可证明命题的补集可证明的命题，非常容易，我是说，这基本上就是数学家在做的事，这也是为什么数学家永远不会失业，因为我们从根本上需要去思考、去创造新的概念、给出新的定义，然后探索我们这些新假设和公理所能推出的结论——这些是旧的那套、比如涉及数论的系统所没有涵盖的。我们需要在上面添加新的东西，而这没法用一个机械的流程来完成，它需要人类的智慧。不过我是说，这挺好的，就像，呃，如果你想如果你还想再多聊聊这个，我们下课以后再聊吧，因为今天要讲的内容很多，但如果我没有完全回答你的问题，告诉我一声，呃，我们课后再说。呃，我想快速跑个题，呃，我想稍微退后一步来看，
03太空服玩笑：未被察觉的同构
9:33
afterwards um so I wanted to quickly digress um I kind of take a stand back right and say okay well girdle found this really nice thing um he found this isomorphism essentially between num between PM principia Mathematica and the things it was describing so he was able to get the system to talk about itself but I think that always begs the question was what if what if if he hadn't discovered this isomorphism what if he hadn't discovered uh this kind of Link this analogy um and this actually relates to a story um that kind of kar and I have recently um and you know Kar was coming up to Boston and you know he says all right I'm coming into kindall soon like all right great so I kind of jokingly texted back and said okay have space suit will travel right so Curran texts back he goes ha excellent so I'm like okay great he got my joke right um and then I went I was like oh so we met up and I was like how' you how'd you like the joke and he was like yeah I thought it was funny I was like well you
然后说，好，哥德尔发现了一个非常漂亮的东西，呃，他基本上找到了一个同构，在数（论）之间在《数学原理》（PM）和它所描述的那些东西之间，所以他能让这个系统谈论它自身。但我觉得这总是让人不禁要问，如果……如果他当初没有发现这个同构，会怎么样？如果他没有发现这种联系、这种类比，会怎么样？呃，这其实跟一个故事有关，就是我和 Kar 最近经历的事。你知道，Kar 当时要来波士顿，他说：好，我马上就到 Kendall 了。我说：好啊，太棒了。所以我半开玩笑地回了条短信说：好，有太空服就能走天下。然后 Curran回消息说：哈，太棒了。我就想，好，他get到我的梗了。呃，然后我们见面的时候，我就问：你觉得那个玩笑怎么样？他说：嗯，我觉得挺好笑的。我说：那你懂那个梗对吧？他却说：
10:38
got the reference right and he was like no what are you talking about I'm like well have spacit World Travel is a is a book by Robert heinan um and he was like no I've never heard of it I just thought you were like putting on this metaphorical space suit and that you were going to go meet me at the at the tea and that's why you could travel right um and I was like oh ha I think that's funny that you thought I was funny even though you didn't get my joke um and then so we kind of talked about this a little while and he's like well so what's the book about and I like actually I don't know I haven't read it so then I Wikipedia it today and I was thinking well actually at the end of the book the main character gets a full scholarship at MIT so there was like a third layer of meeting which neither of us knew and it was just this level of isomorphism which we didn't know but everything seemed to be operating just fine right so I think it's kind of interesting you have this idea that what if we had just
不懂啊，你在说什么？我说：《有太空服就能走天下》是罗伯特·海因莱因写的一本书啊。他说：不，我从没听说过。我还以为你是在说穿上那种隐喻意义上的太空服，然后你就能来茶会上跟我碰面，所以你才能'走天下'呢。呃，我就想：哈，真有意思，你根本没懂我的梗，却还是觉得我好笑。呃，然后我们就这个聊了一会儿，他说：那这本书讲什么的？我说：其实我也不知道，我没读过。所以我今天上维基百科查了一下，我就在想，其实在这本书的结尾，主角拿到了麻省理工的全额奖学金。所以这里还有第三层含义，我们俩谁都不知道。这就是一种我们并不知晓的同构，但一切似乎都运转得挺好，对吧。所以我觉得挺有意思的，你提出的这个想法是：如果当初我们就直接顺着《数学原理》（PM）走下去会怎样
11:28
gone straightforward with with PM without knowing any kind of analogies this ability for the system to do selft talk like what would have happened like would have mathematics just marched on like fine thinking it was complete and everything um I just think that's kind of an interesting idea but I want to get a give give a quick signpost to what's happening today um how many of you guys read chapter 10 levels of description of computer system or at least started to read great excellent fantastic uh nobody read the wrong chap chapter which was chapter 16 um that's a good chapter but it's a little Advanced and that uses a lot of Concepts from chapter 13 and 14 which really help hit home the idea of girdles proof um and you should read in your own time so what today I want to talk about today is is this concept of of emergence and emergent properties coming out of uh out of simple descriptions right we've been kind of hitting this home all all the time but um you know today is kind of our last day to just really try to
在不知道任何类比、系统也没有这种自我对话能力的情况下，会发生什么呢数学是不是就一路往前走，还美滋滋地以为自己是完备的、什么都齐了？嗯，我就是觉得这个想法挺有意思的。不过我想先给大家快速交代一下今天的安排。你们有多少人读了第10章就是「计算机系统的描述层次」那一章，或者至少开始读了？很好，太棒了。呃，没有人读错章节读成第16章吧？嗯，那一章也不错，但稍微高阶了一点，而且它用到了很多第13、14章的概念，那些概念真的能帮你把哥德尔证明的核心想法吃透，你们应该自己找时间去读一读。那么今天我想讲的，是「涌现」这个概念，以及从简单描述中涌现出来的那些性质。我们其实一直在反复强调这一点，不过呢，今天算是我们最后一次真正去展示复杂性，去展示它是怎么从简单的、从简单的
04人是图灵机吗：彭罗斯与明斯基
12:32
show complexity and show how it's emerging out of out of a simple out of a simple building items out of simple building items sorry um and really what this chapter here starts with and why we I mean why does why do you think we talk about computer systems does anyone have an idea why why do you did anyone find it interesting Sandra like computer use the same way as we think okay down so we use computers in the same way that we think like we in terms of breaking down yeah and that's how we can program and fix it for a logic right exactly so go ahead so it so we Bas it on so we make computer system these programing programs so is it based on the way we think or is there okay all right so let me make sure I understand what you're saying so are you asking like are the programs that we program just based on how we think like essentially it's just our idea hey we want the computer to do this yeah and how is it Universal and and how is comp ah so how is computation kind of universal is that is that what you're
构件中涌现出来的——抱歉，是从简单的构件中。而这一章开篇讲的，还有我们为什么……我是说，为什么你们觉得我们要讨论计算机系统？有人知道为什么吗？有谁觉得这个问题挺有意思的？Sandra？「计算机的运作方式和我们思考的方式一样」，好的。所以我们使用计算机的方式和我们思考的方式是一样的，就是那种拆解……「拆解」，对，而且这样我们才能编程、才能按逻辑去修正它，对吧？没错。你说，请讲。所以我们是基于……我们造出计算机系统、这些程序，那它到底是基于我们思考的方式，还是……好的，那我先确认一下我有没有听懂你的意思。你是在问，我们写的那些程序是不是就是基于我们思考的方式，就是说本质上它只是我们的想法——「嘿，我想让计算机做这个」？对，那它又怎么会是通用的？计算……啊，你是说计算为什么具有某种普适性，你想问的是这个吗？好，对。嗯，那这其实
13:51
getting at okay yeah I mean so that that kind of leads back to a fundamental idea of of a turing machine right where where you have um although we subjectively have this own approach to how we'd solve the problem right it can ultimately be reduced to an algorithm a set of instructions which is universal since as universal as mathematics right so pretty much anybody in any language should understand it right is that kind of what you wanted to say okay nine oh sorry the could we just be on implementation of a touring machine say again could we be just an implementation of a teering machine could we just be yeah like you know us as humans be implementation yeah and like emotions are like different ways of like talking about like some part of a Miss okay all right so you're getting at a really interesting idea here this idea that humans in our thoughts are actually best modeled by a by a turning machine right um and I don't know if you've done this reading and maybe that's why you're being clever and asking this question
又回到了图灵机的一个根本思想，对吧？就是说，尽管我们主观上各自有自己解决问题的思路，但它最终都能被归约成一个算法、一组指令，而这是普适的，就像数学一样普适，对吧？所以基本上任何人、用任何语言都应该能理解它。你想说的是这个意思吗？好。第九……哦抱歉，「我们会不会本身就是图灵机的一种实现？」再说一遍？我们会不会本身就是图灵机的一种实现？我们会不会就是……对，就是我们人类本身就是一种实现。对，而且情绪之类的东西，只是在谈论这台机器某一部分的不同说法。好的，你说到了一个非常有意思的想法，就是人类的思维其实最好用图灵机来建模，对吧？嗯，我不知道你有没有做过预习，也许这就是你这么机灵、问出这个问题的原因——你听说过一个叫罗杰·彭罗斯（Roger Penrose）的人吗？听过？好，罗杰·彭罗斯
14:56
but have you heard of someone named Roger Penrose yeah okay so Roger Penrose is a very prominent mathematical physicist he at Oxford right um and he also wrote uh an pretty interesting book called Shadows of the mind um in which He suggests that girdle's incompleteness theorem and the halting problem has application to human intelligence um but what he argues is that humans are fundamentally not turning machines but that computers are right and that's why artificial intelligence in his mind is impossible is because a machine is always limited to these kind of constraints which we've cooked up here today and through the halting problem but humans obviously aren't we can meta think we can Meta Meta think and we can Meta Meta Meta think always jumping outside the system right but we're never ever bound to this the same constraints of a machine now I'll warn you penrose's ideas are who constrains the machine who constrains the machine the ones who logic logic right the fact that all a machine can do
是一位非常著名的数学物理学家，在牛津，对吧？嗯，他还写过一本相当有意思的书叫《意识的阴影》（Shadows of the Mind），书里他提出哥德尔不完备定理和停机问题可以应用于人类智能。嗯，但他论证的是：人类从根本上不是图灵机，而计算机是对吧。这也是为什么在他看来人工智能是不可能的——因为机器永远受限于我们今天在这儿捣鼓出来的这些约束，还有停机问题带来的约束，但人类显然不受这些限制。我们能元思考，能元元思考，还能元元元思考，永远可以跳到系统外面去对吧，我们永远不会被机器那种约束绑住。不过我要提醒你们，彭罗斯的这些想法……「是谁在约束机器？」谁在约束机器？「是那些……逻辑。」逻辑，对。事实就是，机器能做的
16:01
is you know not and copy jump basic assembly code right right it is just a turning machine but what can a human do what can a human do that's not a turning machine good question right we can think emotionally we can believe we can double think right humans all the time believe in both p and not P yeah but it's like um humans AR like this cerent systems it's like I I think we think like we're like one thing when we're like really many things that can also be contradictory it's like a bunch of little machines going together like communicating with each other they can they can contradict what they can agree okay so you've got essentially this kind of society of the Mind Viewpoint you know Marvin Minsky's book um where human intelligence is best modeled by a bunch of little actors right that kind of almost vote on certain beliefs right and I mean it's a good question right you also want to have to figure out one how do the actors work and how do the actors make decisions um but two do you really think it's just a
无非就是「非」、「与」、复制、跳转这些基本汇编指令，对吧？对，它就是一台图灵机。那人类能做什么呢？人类能做什么是图灵机做不到的？好问题。我们可以情绪化地思考，我们可以「相信」，我们可以「双重思想」，对吧？人类无时无刻不在同时相信 P 和非 P。「对，但就好像，嗯，人类是那种（并发的）系统就是说，我觉得我们以为自己是一个整体，其实我们是很多个东西，而且这些东西之间还可能互相矛盾，就像一堆小机器凑在一起、互相通信，它们可以互相矛盾，也可以互相一致。」好，那你基本上是持一种「心智社会」的观点，就是马文·明斯基（Marvin Minsky）那本书里讲的，人类智能最好用一堆小小的「行动者」（actor）来建模，它们差不多是在对某些信念进行投票，对吧？我是说，这是个好问题，但你也得搞清楚：第一，这些行动者是怎么运作的，它们又是怎么做决策的；第二，你真的觉得你脑子里就是少数服从多数吗？就好像有一群小人、小小的
17:00
majority kind of rules inside your brain like or a little a little guys little homunculus is up in your head like voting on like no I think today we do believe in God and you know no I don't believe in God and then you know all right everyone raise your hand if you do if you don't okay today we don't believe in God right I mean does that really is that really how human intelligence works right we don't know right but that's the important uh important thing about studying layers of systems layers of description and that's what this chapter is all about um so I want to give someone else like an opportunity to explain essentially why do we study computer systems and trying to understand intelligence trying to understand the Mind does anyone have an idea because we can like we from things that we think we know we can do we can try to make a computer do similar things in some way just to maybe suggest maybe this is how we would be the same thing okay so possibly right I'm not sure um but here's here's what I kind of came up
「侏儒」（homunculus）在你脑袋里投票，说「我觉得今天我们信上帝」，然后另一个说「不，我不信上帝」，接着就是「好，信的举手，不信的举手，好，今天我们不信上帝」，对吧？我是说，人类智能真的是这么运作的吗？我们不知道，对吧。但这正是研究系统层次、描述层次的重要之处，而这一章讲的就是这个。嗯，所以我想给别的同学一个机会来解释一下：我们为什么要研究计算机系统，来试图理解智能、理解心智？有人有想法吗？「因为我们可以，从那些我们自认为知道自己能做的事情出发，去试着让计算机以某种方式做类似的事，也许就能提示：也许我们就是同一回事。」好，有可能，对吧，我也不确定。嗯，不过我自己是这么想的，这也是我认为
05从齿轮到操作系统：计算机的描述层次
18:08
with um and it's my motivation for why I think people might want to study computer science um this will kind of launch into a whole little lecture of what do you what do you want to do with your life kind of thing I'm not going to try to solve that problem for you but we can try um so what I think is cool about computers is that um kind of back in the day uh you know if you want to go all the way back to babage and these guys in the you know late 1800s right they were just playing around with gears and things like this right so I mean the physical they were working with just physical level right so we had kind of this physics governing everything um and you know he was playing with gears and wheels but what really makes up today is we have these kind of we have transistors um and capacitors we have electronic things for doing really small kind of computations like babage was thrilled when he could just get like a simple calculating machine like aha wow now you can do logarithms like this is amazing
人们为什么可能想学计算机科学的动机。嗯，这会引出一整段小小的演讲，关于「你这辈子想做什么」那种话题。我不打算替你们解决这个问题，但我们可以试试。嗯，我觉得计算机很酷的一点在于呃，早年间——如果你想一路追溯到巴贝奇（Babbage）他们那一代人，在，你知道的，1800年代末，对吧——他们就是在摆弄齿轮之类的东西，对吧？我是说那个物理层面，他们工作的层次就是纯物理层面，所以有一套物理规律在支配一切。嗯，你知道，他就是在摆弄齿轮和轮子。但今天真正构成我们这一切的，是我们有了晶体管，嗯，还有电容，我们有各种电子元件来做非常微小的运算。像巴贝奇，他当时能搞出一台简单的计算机器就已经欣喜若狂了——「啊哈，哇，现在能算对数了，这太神奇了！」嗯，呃，但他是在这个
19:16
um uh but he was programming at this level right um his levels of description were almost completely based on Gears wheels and then what became in the 1900s early 1900s trans sisters in capacitors um but then we kind of had this uh I would suppose this kind of Revolution and in how we did thing how we do things and we got got up to the level of kind of to speak in modern terms you know motherboards and like video cards and things like this and we went ahead and started like abstracting entire chunks of Hardware to do very specific highly designed things and we just started playing around with those chunks like H you know now I won't have as much lag in my video games I can have a new video card and Pop That in um so we no longer have I mean how many people I mean at least popular computer users ever worry about the transistors and capacitors in their computer like oh no like um it doesn't happen and then we have this kind of Next Step Up of and this is really what where the of computer science lives is
层次上编程的，对吧。嗯，他的描述层次几乎完全建立在齿轮、轮子之上，然后到了1900年代，出现的是……1900年代早期，电容器里的晶体管……不过后来我们经历了一场，我想可以说是一场革命，改变了我们做事的方式，改变了我们做事的方式，我们上升到了用现代话来说就是主板、显卡这一类东西的层次，诸如此类的东西，我们开始把整块整块的硬件抽象出来，让它们去做非常特定、经过高度设计的事情，然后我们就开始摆弄这些模块，比如说，嗯，现在我打游戏不会那么卡了，我可以买块新显卡插上去。所以我们不再有——我是说，有多少人，至少是普通的电脑用户，会去操心自己电脑里的晶体管和电容器？根本不会有这种事。然后我们再往上走一层，这才是计算机科学真正所在的地方，也就是软件和操作系统，
20:31
you know software and operating systems which which going of live on top of this even um and this is amazing right like something we're dealing with things which don't really exist right we're kind of executing spells when when Curran gets up later today and and he he hacks away on his computer you know type type type type he just puts in these little you know images that show up on the screen and then he he hits an enter button and then it suddenly like executes like magic and it's exists in a different kind of Realm but it all boils down to what's happening on this level of transistors and capacitors and the ability to actually tell you if if you went to somebody and you said hi my name's X can you tell me how you know Windows Works in terms of of capacitors and transistors right they were just going to go do you want to enroll in school for the next four years like sure like uh and that's really the the project here but what's interesting is that there's a conceptual framework we have for
它们又运行在这一层之上。这非常神奇，对吧，我们在跟一些并不真正存在的东西打交道，我们某种意义上是在施法。等一会儿 Curran 上来，他在电脑上敲敲打打，噼里啪啦，他就是敲进去一些显示在屏幕上的小图像，然后按一下回车键，突然之间它就像魔法一样执行起来了，它存在于另一个层面的领域里，但归根结底一切都取决于晶体管和电容器这个层面上发生的事情。而要真正说清楚——如果你去找一个人，跟他说，你好我叫某某，你能用电容器和晶体管来给我讲讲 Windows 是怎么工作的吗？他们大概只会说，你是想再读四年书吗？当然啦，呃，这其实就是我们这个项目要做的事，但有意思的是，我们在这里有一套概念框架，它非常接近我们希望用来破解心智之谜的思路，
06神经元到灵魂：大脑的层级塔
21:40
thinking here uh which is very close to how we hope to crack the problem of the mind and that's the idea that we kind of have we have neurons and you know neurotransmitters things ultimately governed govern transmitters sorry ran out of space here uh you know we have we describe things in terms of like cable Theory like we're just sending messages down an insulated cable um we have this kind of layer level of description for the brain at this very small level and this is already you know research worthy but then we then we have a bunch of biologists and such and you know people interested in neuroanatomy you know talking about like oh what about the amygdala I don't know if I'm spelling any of this right the hippocampus you know your your front cortex your cerebral cortex and um I'll just denote by cortices I don't know if that's the correct piz um and then somehow out of these larger structures where you abstract away the details of you know what's going on with the transcription in this cell and
那就是这样一个想法：我们有神经元，还有神经递质之类的东西，最终由这些递质来支配——抱歉，这里位置写不下了——我们用类似电缆理论的方式来描述这些东西，就好像我们只是在一根绝缘电缆里传递信息。我们在这个非常微观的层面上有这样一套对大脑的描述，这本身就已经很值得研究了。但接下来，我们还有一大群生物学家之类的人，还有那些对神经解剖学感兴趣的人，他们会讨论，哦，杏仁核怎么样——我不确定这些词我拼得对不对——海马体，还有你的前额叶皮层、大脑皮层，嗯，我就统称为皮层好了，我不确定这个复数形式对不对。然后不知怎么的，从这些更大的结构出发——在这一层你把细胞里转录之类的细节都抽象掉了，转而只是去想，我的海马体今天状态怎么样——我们就一路进入了
22:56
instead you're just thinking about like how's my hippocampus doing today um we then all the way enter essentially an isomorphic level of software and Os which is you know the psyche the mind um and I'll go ahead and put it up here Soul right and this is kind of fundamentally why artificial intelligence researchers are found in the computer science department um is because we've had a lot of guys who have gone from this level of dealing with very kind of clunky physics to doing really souped up stuff on on like larger physical entities motherboards video cards and then getting up to you know kind of the level of software systems and and operating systems um but I want to take a quick poll like where do your guys interests lie on this kind of Stack you you know kind of in both both camps here but especially focusing on on the brain and going up to the mind right what do you guys find most interesting right this is going to really unve reveal a lot about who you are and what you want to do right what do you find
一个与软件和操作系统本质上同构的层面，也就是心灵、心智，嗯，我干脆也把它写上去：灵魂。这基本上就是为什么人工智能研究者会出现在计算机科学系里，因为我们有很多人就是从处理这种笨重物理层面的工作，一路做到在更大的物理实体上搞出很厉害的东西——主板、显卡——然后再往上做到软件系统和操作系统的层面。不过我想快速做个小调查，你们的兴趣落在这个层级的哪一层？你们可能两边都有点兴趣，但特别是聚焦在大脑往上到心智这条线上，你们觉得哪一层最有意思？这其实会很大程度上揭示你是什么样的人、你想做什么。在这三个框里，你觉得哪个最有意思？如果只能投一票，
24:13
most interesting in these kind of three boxes like if you had to vote for one box what would it be you you want you want two boxes I can't give you two boxes you got to decide I mean otherwise you're going to be in school forever right all right two boxes but that means you automatically have to go to school for seven years after undergrad right um okay go ahead give me some votes I want to see some hands uh who likes studying molecules neurons neurotransmitters maybe even basic physics okay Sandra durus Maya Rishi okay who's kind of the inherent biologist here who likes you know amydala hippocamp ET cortices all right we got Latif all right now who's interested in kind of the psyche the mind the soul who who who wants to like kind of actually sit in an office and deal with people's problems rather than or maybe not I warn you it's it's you don't want to deal okay so we got naven I'm sorry I forget your name again Vivan say again Vivan Vivien Vivian Maya Latif and Felix okay cool um I have another one to
你会投哪个框？你想要两个框？我不能给你两个框，你得做出选择。不然的话，你就要一辈子待在学校里了。好吧，两个框，但那意味着你本科毕业后自动要再读七年书。好，来吧，给我投个票，我想看到举手。谁喜欢研究分子、神经元、神经递质，甚至是基础物理？好，Sandra、Durus、Maya、Rishi。好，这里谁是天生的生物学家，喜欢杏仁核、海马体、各种皮层？好，我们有 Latif。好，那么谁对心灵、心智、灵魂感兴趣？谁想真正坐在办公室里处理别人的问题——或者也许不想，我警告你，那可不好受。好，我们有 Naven，抱歉，你叫什么名字来着？Vivan？再说一遍？Vivan？Vivien？Vivian、Maya、Latif 和 Felix。好，很好。嗯，我还想加一个。好，说吧。谁想理解所有这些层级之间的共性？
25:28
add okay go for it who wants to understand the commonalities between all of these levels okay we're getting a lot of hands there all right so who wants to understand the commonality between all these all right and this is you know ker and I kind of put together this this list um and he'll show you when when he kind of comes up and gets the projector rolling um but it's this idea of we've found ourselves working with problems and systems which are so complex that we can only really occupy ourselves with kind of certain layers of abstraction but the interesting thing is that we can abstract right just like when we're you know modding out our computer like we don't ever have to really care about transistors and capacitors right um and similarly when I'm you know around on mat lab right I don't ever have to worry about like well sometimes I have to worry about memory use but I don't usually have to worry about any physical details yes you can't have one without the other right right exactly Sandra says you can't have one
好，这里举手的人不少。好，那么谁想理解所有这些层级之间的共性？这其实是 Curran 和我一起整理出来的这份清单，等他上来把投影仪弄好，他会展示给你们看。但核心就是这个想法：我们发现自己在处理的问题和系统太复杂了，我们只能把注意力放在某些特定的抽象层次上。但有意思的是，我们是可以做抽象的，就像我们改装电脑的时候，根本不需要真的去关心晶体管和电容器一样。同样地，当我用 MATLAB 的时候，我根本不需要担心——好吧，有时候我得考虑内存占用，但我通常不需要操心任何物理细节。对，你不可能只有其中一个而没有另一个，对，完全正确。Sandra 说你不可能只有一个而没有另一个，这非常重要，
26:42
without the other very important right very very important is you do have a tower here right um but what's scary maybe maybe scary is the idea that some of these things are Universal and are independent right you can have the physical things without the soul or the man the S okay so definitely right these arrows only go One Direction you can have the physical thing without kind of the mental right we have plenty of trees and things who don't have Windows V running on them go ahead arrows what do the arrows represent do you have to ask me that question no but well I know but for most all right so that's actually a very good question that's a very good question um so what do the arrows mean um so a simple answer would be they indicate I guess derivability or dependence right except the dependency is kind of going the other way right you know motherboards and video cards depend upon their underlying transistors capacitors um but the arrow is kind of pointing up towards them that's this idea that you can build things out of it
非常非常重要，你这里确实有一座塔。但有点吓人的地方——也许算吓人吧——是这样一个想法：其中有些东西是普遍的、是相互独立的。你可以只有物理的东西而没有灵魂、没有心智。所以确实如此，这些箭头只指向一个方向：你可以有物理层面而没有心智层面，我们有的是树木之类的东西，它们身上可没跑着 Windows Vista。请说。箭头，箭头代表什么？你非得问我这个问题吗？不，不过，我知道，但对大多数人来说——好吧，这其实是个很好的问题，非常好的问题。那么箭头是什么意思呢？一个简单的回答是，它们表示可推导性或者依赖关系，只不过依赖的方向其实是反过来的，对吧，主板和显卡依赖于底层的晶体管和电容器，但箭头却是朝上指向它们的。这体现的是这样一个想法：你可以用它们搭建出东西来。这是一个很重要的想法：
07从少量构件到涌现：语言与蚁群
27:58
right and this is an important idea is that like once you've got a basic amount of tools available to you you can kind of start building up words right like once you've got and this is I think very interesting and it's it's a good concept to study for for kind of languages right so when you're learning Spanish or any other language what's perhaps the most useful phrase that you will ever learn in Spanish okay Ola why why is that useful you can say hi big deal what if you want to learn more Spanish what's the best phrase you can learn and I'm sorry I don't know many other languages okay okay how do you say this right in sunlight with those three words right you've opened yourself up to an entire world of a language right you know say oh Lio like okay uh K say oh wonderful fantastic so suddenly you have this prompting for the outside world just by learning three words and you could kind of learn the entire language right um and that's an important thing right like once you've got some basic elements some some just
一旦你手头有了基本的一套工具，你就可以开始往上搭建词汇了，一旦你有了——我觉得这非常有意思，而且这是个很适合用来研究语言的概念——比如说你在学西班牙语或者别的什么语言，你这辈子能学到的最有用的一句话是什么？好，Hola。为什么它有用？你可以打招呼，那又怎样。如果你想学更多西班牙语呢？你能学的最好的一句话是什么？抱歉我不太懂别的语言。好，好，就是「这个怎么说」，对吧，用这三个词，指着阳光就能问了。你就等于向一整门语言的世界敞开了大门，对吧，你会说 hola，会说 okay，会说 K，会说 wonderful、fantastic，于是突然之间你就有了跟外部世界互动、发出信号的能力，只靠学会三个词，你就好像能慢慢学会整门语言，对吧，而这是一件很重要的事，就是说，一旦你掌握了一些基本元素，一些最基础的构件，一些词和词汇，
29:13
kind of building blocks of of of words and vocabulary um you can immediately then kind of consider this lofy layer of what happens when you put this word so let's go kind of a b C so what happens if we have just a together you know B to by itself or we could have a or we could have you know AC and then BC right we can we already have this like idea of once you've got some things you can take the power set of them and and get oh sorry forgetting something crucial um and you suddenly have a whole layer of complexity just by considering the metal level beyond your beyond your first first layer here so you like you learn a handful of Spanish words right and then just by considering their kind of combinations and permutations and ability to prompt the outside world you can suddenly feed into a whole another layer of abstraction another step up it's the same thing with here right so say you've got one transistor right okay good for you right but suppose you have like a hundred of them right you can start doing stuff some
你马上就可以去考虑更高一层的东西：当你把这些词组合起来会发生什么。那我们就来看看a、b、c 吧。如果我们只有 a 会怎样，或者只有 b，或者我们可以有 a，也可以有ac，然后 bc，对吧，我们已经有了这样一个想法：一旦你有了一些东西，你就可以取它们的幂集，然后得到——哦抱歉，漏了很关键的一点——于是你突然就多出了一整层复杂性，仅仅是因为你考虑了第一层之上的那个元层次。所以就好比你学了几个西班牙语单词，对吧，然后仅仅通过考虑它们的各种组合、排列，以及用它们跟外部世界互动的能力，你就突然进入了另一整个抽象层次，又往上迈了一步。这里的情况也一样，对吧，所以假设你有一个晶体管，好吧，挺好的，对吧，但假设你有一百个，那你就能开始做点事情了，
30:26
right and you can start doing stuff and have things which emerge which are greater than just the individual parts right um I mentioned briefly this idea of the ant of ant societies right of uh of ant societies um exhibiting these kind of emergent properties which are greater than any single one ant right and this is an important idea especially when you even even think about humans right like humans always do this we one person has an idea and another person has another idea and then you can kind of consider or what's the interface of these two ideas and you've got another idea right like you can suddenly build things out of these simple pieces um and I suppose the Merit of of studying this is not only do you understand when can you forget about the details
对吧，你可以开始做各种事情，而且会涌现出一些大于各个部分之和的东西，对吧。我之前简单提过蚂蚁社会这个想法，对吧，就是蚂蚁社会会表现出这种涌现性质，超出任何单只蚂蚁所能做到的，对吧。这是个很重要的想法，尤其当你甚至去想想人类的时候，对吧，人类一直在做这种事，一个人有一个想法，另一个人有另一个想法，然后你可以去考虑这两个想法的交汇点是什么，于是你就有了第三个想法，对吧，你突然就能用这些简单的零件搭出东西来。我想研究这件事的价值不仅在于让你明白什么时候可以忽略细节，
08红细胞问题：自我在哪一层
31:22
abstraction um but what can you do given the pieces that you have right um so I I think this is kind of a you know important piece tied into this though is this idea of of levels of description right how do you describe something based on what level of even ourselves right I don't I'm sure many of you ran across the part in in chapter 10 where Hoffer talks about the paranoid in the operating system and he talks about this program Perry and he said and he talks about the situation and you know he he kind of plays out and I'll go ahead and elaborate stuff that wasn't in the book um but I think it's an interesting thought experiment so how many of you would identify your bodies as part of yourself Sandra navine dur Latif Felix Rishi okay so wait you guys didn't raise your hand right see don't identify your bodies with yourself okay so Latif changes his mind all right okay so for those of you who did oh is the part of my body is the brain part of your body yeah okay then yeah okay so the brain's part of your
也就是抽象，还在于：给定手上的这些零件，你能做什么，对吧。所以我觉得跟这一点紧密相关的另一个重要部分，就是「描述层次」这个概念，对吧，你要在什么层次上去描述一样东西，甚至包括描述我们自己，对吧。我相信你们很多人都读到了第 10 章里侯世达谈到操作系统里那个偏执狂程序的那部分，他讲到一个叫 Perry 的程序，他描述了那个情境，他把它演绎了一遍。我下面会展开讲一些书里没有的内容，但我觉得这是个很有意思的思想实验。你们当中有多少人认为自己的身体是自我的一部分？Sandra、Naveen、Latif、Felix、Rishi，好，等等，你们几个没举手，对吧，你们不认为身体属于自我。好，Latif 改主意了。好吧，那对于那些举手的人——哦，那我身体的一部分——大脑算是你身体的一部分吗？算，好，那大脑是你身体的
32:42
body anyone I want to change the votes okay fine so if I were to ask you you know how's your leg feeling today how many people would think that's a normal question okay so we got plenty of normal questions all right um so suppose I didn't ask the question which hofstead ask is like so why are you making so many so few red blood cells today right is that a normal question yeah go ahead Sandra you think it's a normal question do you have control over how many red blood cells no no but they're part of yourself it's part of you right so why can't I address you red blood cell count as part of you well it's inside you it's inside of you right but you said that your body is part of okay so you're accepting that as a part of The Logical consequence of the things you've said all right so good for you because you're being consistent does anyone else like suddenly feel like I've I've said something awkward all right maybe we shouldn't even say how does your leg feel today maybe we shouldn't even say how does he look so we never
一部分。还有人想改票吗？好，那行。那如果我问你「你今天腿感觉怎么样？」，有多少人觉得这是一个正常的问题？好，看来大家都觉得挺正常。那假设我问侯世达提的那个问题：你今天为什么造了那么多、或者那么少的红细胞？对吧，这算正常问题吗？好，Sandra 你说，你觉得这是正常问题吗？你能控制自己造多少红细胞吗？不能，但它们是你自身的一部分，是你的一部分，对吧，那我为什么不能把你的红细胞计数当作你的一部分来跟你谈呢？嗯，因为它在你体内。是在你体内没错，但你刚才说了你的身体是你的一部分。好，所以你接受这是你所说的那些话的逻辑推论。行，这挺好的，因为你保持了一致性。还有别人突然觉得我说的话有点别扭吗？好吧，也许我们连「你今天腿感觉怎么样」都不该说，也许我们连「他看上去怎么样」都不该说，这样我们就永远不用
33:54
address which is which right so suppose like I don't know Kar comes in he's just got two black eyes right I'm like Kar you look terrible and he's like that's just not an acceptable question right because I'm referring to his physical appearance right so mind is reflecting his physical feelings his physical thing so he's like almost internalizing that picture and saying oh yeah that's me okay so you're saying Curran's kind of self model is inherently dependent on his body and his interaction with his environment right so then when I say you look terrible he views his own internal model and looks at himself as being terrible right so then that does make sense right but then what about the red blood cell question right the red blood cells I mean usually you do not have like an internal model of your own red blood cells okay so yeah right you don't normally have an internal model of your red blood cells right but suppose you did right and suppose that and this is kind of going to be one of the
区分到底指的是哪一个了，对吧。那假设，比如说 Kar 走进来，他顶着两个黑眼圈，我说「Kar，你看上去糟透了」，然后他说这问题根本不能接受，对吧，因为我指的是他的身体外表，对吧。所以心智反映了他身体上的感受、身体上的东西，他几乎是把那个形象内化了，然后说「对，那就是我」。好，所以你是说 Kar 的自我模型本质上依赖于他的身体以及他跟环境的互动，对吧。那么当我说「你看上去糟透了」时，他会去看自己的内部模型，然后觉得自己很糟糕，对吧，这么说就讲得通了。但那红细胞的问题怎么办呢？对吧，红细胞——我是说，通常你并没有关于自己红细胞的内部模型。好，是的，对，你通常没有关于自己红细胞的内部模型。但假设你有呢？假设——这其实就是这个思想实验最值得带走的有趣结论之一：如果你能在内部直接访问
34:56
take-home interesting consequences of this thought experiment is what if you had internal access to everything that was going on right terrible okay so everyone's going oh God it would be terrible I mean why would it be terrible like I'd be like trying to walk across this room it's some choices you make are not necessarily the right or appropriate ones yeah so how stop wait minute wait a minute it's um first you have to identify what what's making the decisions in first perhaps the conscious thing doesn't make the decisions but some the decisions it is aware that are being made it ascribes to itself okay repeat that last line one more time the decisions that is aware of being made it describes as being made by itself okay so they're you make certain decisions and you think that you're making those decisions you I always ask who's making the decisions my neurons or my so all right and you know I really wish I could have passed out the dialogue like who pushes who around the whom around the crine cranum yes and of
体内正在发生的一切，会怎么样？很可怕。好，大家都在说「天哪那太可怕了」。可为什么会可怕呢？就好像我想走过这个房间时——你做的有些选择未必是正确或合适的。对，那么——等一下，等一下，首先你得先弄清楚是什么在做出这些决定。也许有意识的那部分并不做决定，而是它意识到某些决定被做出了，然后就把这些决定归到自己头上。好，最后那句再说一遍。它所意识到的那些被做出的决定，它就认为是自己做的。好，所以你做出某些决定，然后你认为是你在做这些决定。我总是问，到底是谁在做决定，是我的神经元还是我？好吧。我真希望我能把那篇对话印发给大家，就是关于「谁在颅腔里推来推去」的那篇，对，还有其他几篇对话。你说得
36:06
H that as other dialogues um and you're right and I wanted to kind of read this dialogue but I don't know if we have time for it today but the Prelude ant view gets this idea of symbols right of these kind of larger emerging groups of neurons and little internal representations you have the outside world and to an certain extent they make up who we are our internal models are the ones that then collectively make decisions but on the other hand we then have times where the emerging thing then models I mean manipulates our internal models of the world right so suppose like you're sitting in this class one day and then I'm like telling you well you know Universe actually isn't kind of deterministic it's actually got this probabilistic framework to Quantum Mechanics so you go wow and then you change your like base little mental model of the world right so I mean there's this kind of weird interaction with hierarchies right so this is this is an interesting point right because here there was almost there's very
对，我本来想读一读这篇对话，不过我不确定今天还有没有时间。不过《前奏曲与蚂蚁赋格》里提出了「符号」这个概念，对吧，就是那种更大的、涌现出来的神经元群体，以及你对外部世界的小小内部表征。在某种程度上，正是它们构成了我们是谁，我们的内部模型才是那个集体做出决定的东西。但另一方面，有时候那个涌现出来的东西反过来会建模——我是说会操控我们对世界的内部模型，对吧。比如说，假设某天你坐在这个课堂上，我跟你说，你知道吗，宇宙其实并不是决定论式的，它其实有量子力学那套概率性的框架，然后你「哇」了一声，接着你就改变了自己关于世界的那个基础的小小心智模型，对吧。所以你看，这些层级之间存在一种很奇怪的互动，对吧。这是一个很有意思的点，因为在这里，往下的方向几乎没有什么相互作用，但你知道吗，这种事确实会发生，
37:09
little interplay going down this way but you know what it does happen right like I think one of the best examples is your cell phone right because you have this SIM card right as part of your cell phone um which actually kind yeah exactly I mean you take it out and it might not be operational right but you could also try putting it into another cell phone and it might not work right and that's in part ascribed to this higher service which is getting beam back down so there's there is a weird interplay of levels software and Hardware going on here suppose we can have like a brain you know you have your um perhaps your your lower level neurons and neurons that talk about those and neurons that talk about those so you're not having like this crazy software view you know the uh maybe the uh Consciousness that's not even the physical world is actually like attracting like or maybe like making changes to the physical world it's like the physical world is making changes to the physical world but those higher
对吧。我觉得最好的例子之一就是你的手机，对吧，因为你手机里有一张 SIM 卡，它是手机的一部分，而它其实算是吧，对，没错，我是说你把它取出来，它可能就不能用了，对吧，但你也可以试着把它装进另一部手机里，它可能也不好使，对吧，这在某种程度上要归因于这个更高层的服务，它把信号发回下面来。所以这里面软件和硬件的层级之间有一种很奇怪的相互作用。假设我们可以有个大脑，你知道，你有你的，嗯，也许是你的低层神经元，然后有谈论那些神经元的神经元，还有谈论那些的神经元。所以你不是持有那种疯狂的软件视角，你知道，嗯，也许，嗯，那个意识甚至都不在物理世界里，实际上像是在吸引，或者像是在对物理世界做出改变。其实是物理世界在对物理世界做出改变，但那些更高层的神经元算是代表抽象概念的那些。抽象概念并不真的存在，但
09内省为何困难：演化的设计
38:10
level neurons are sort of the ones representing the abstractions the the abstractions don't really exist but it's like these neurons are like taking something from like the lower level neurons in they see like the essential processes of them and then they're controlling them back so you are like mostly your higher level NE all right so maybe you are mostly your higher level neurons but I think an interesting example and you know we're getting into details with the brain which I'm not sure how how it really works but for example with the ant societies right no ant is higher than another ant right all ants are pretty much created equal um yeah say again the queen except for the queen but still there was this myth right there was this myth that the queen ant was like sitting there with her hyper brain right and all right you know tell forces a through b to go to sector G and attack and then bring over the leaves here right no the queen ant doesn't do that right I mean it's just like all the other ants that
就好像这些神经元从低层神经元那里获取某些东西，它们看到了那些神经元的本质过程，然后再反过来控制它们。所以你基本上就是你的高层神——好吧，所以也许你主要就是你的高层神经元。但我觉得有一个很有意思的例子，你知道，我们现在讲到大脑的细节了，而我并不确定它到底是怎么运作的。但比如说蚂蚁社会，对吧，没有哪只蚂蚁比另一只蚂蚁更高级，对吧，所有蚂蚁基本上生而平等。嗯，对，你再说一遍，蚁后——除了蚁后之外。但即便如此，曾经有这么一个迷思，对吧，有这么一个迷思，说蚁后就像坐在那儿，带着她的超级大脑，对吧，然后好吧，你知道，命令 A 部队到 B 区，去 G 区进攻，然后把叶子搬到这边来，对吧。不，蚁后不干这些事，对吧。我是说，她跟其他所有蚂蚁一样，大概
39:09
has probably three neurons right and can't do much with them um but it's the interaction of these very simple things which can produce complex Behavior yes Li I just realized something the thing is you're not even a way of your own thinking and the thing we like we often think that we're like very self-reflective even though we're not all right yes we are we we are we do think even reason is beyond us I mean I mean think how many times have you ever reason ever since you were a child you most of the time you felt you know speak you don't even know where the worlds are coming from you don't know where your concepts are coming from so I can say we are highly self-reflective if we don't like even like most of the time reflect towards our own thoughts our feelings and whatever you know mhm no I mean right exactly like we we tend to just kind of assimilate data about our outside world and especially kind of the mental models passed on to us from our parents right that immediately start telling us and you know kind of training
就三个神经元，对吧，靠这些干不了太多事。但正是这些非常简单的东西之间的相互作用，才能产生复杂的行为。是的，我刚意识到一件事，问题是你甚至意识不到你自己的思考。而我们常常觉得自己非常善于自我反思，其实并不是。好吧，是的，我们是的，我们确实会思考，甚至连推理都超出我们的掌控。我是说，我是说，想想看，从你小时候到现在，你真正推理过多少次？你大多数时候都是凭感觉，你知道，你说话的时候，你甚至不知道那些词是从哪儿冒出来的，你不知道你的概念是从哪儿来的。所以我可以说，如果我们大多数时候连自己的想法、感受之类的都不去反思，那我们怎么能算高度自我反思呢，你知道吧。嗯，不，我是说，对，没错，就像我们往往只是吸收关于外部世界的数据，尤其是父母传给我们的那些心智模型，对吧，那些从一开始就立刻告诉我们，你知道，算是在训练我们、在我们身上做印刻的东西，真的就像大雁一样，从我们出生第一天起就被印上要去做
40:10
us and printing us really like geese from you know our first day to do certain things right and we can rarely break out of those out of those kind of Loops out of those systems right um but that that's an interesting idea right um it's this idea that what happens when we can't really scrape away at the boxes anymore right like we try to I mean really what happens I mean why is psychology so difficult why is solving Consciousness so hard right because we're kind of like trying to like take out an eyeball and like peer back out at us and like all right what am I like and you know that I I remember having this discussion with one of you after class it's this idea that so Evolution really kind of designed us to do certain things and you know I suppose one reason why humans kind of took off with these abnormally large forebrains was because it was a good method for survival right we suddenly started getting together in little groups and you know planting things in the ground and suddenly we had a really stable
某些事情，对吧。而我们很难跳出那些，跳出那些循环，跳出那些系统，对吧。不过那是个挺有意思的想法，对吧。就是这个想法：当我们再也没法把那些盒子一层层刮开的时候会怎样，对吧，就像我们试着——我是说，到底会发生什么？我是说，为什么心理学这么难？为什么解决意识问题这么难，对吧？因为我们有点像是在试图把眼球取出来，然后回过头来盯着自己看，然后说，好吧，我到底是什么？你知道，我记得下课后跟你们中的一位讨论过这个，就是这个想法：演化其实是把我们设计来做某些特定的事情。你知道，我猜人类之所以进化出这种异常巨大的前脑并且发展起来，一个原因是它是个很好的生存手段，对吧。我们突然开始聚成小群体，你知道，把东西种到地里，于是我们突然有了非常稳定的食物来源；或者我们会，你知道，诱骗猛犸象，把它们赶下悬崖，让它们摔到我们自己
41:17
source of food or we would you know trick mammoths and run them off into Cliffs and like have them fall in spikes that we designed and sharpened ourselves like and suddenly like problem solving was a really really important thing for evolution by natural selection right so people who had kind of had this these problem solving capabilities were selected favorably but you know it's really nice to be able to you know run mammoths off cliffs and plant things in dirt um but what about the guy who is like you know kind of in at the P bottom of the pack kind of thinking you know what am I and you know the cheah like attacks right as he's like caught in these thought Loops about what you know what am I um like crazy philosopher right so I mean the philosopher was I mean it's like you first you get like problem solving average you know maybe perhaps above average a little bit and then you get like these monsters like so creepy you know those genuses they come along they like all the time they're like all
设计并削尖的尖桩上。就这样，突然之间，解决问题就成了一件对自然选择演化非常非常重要的事，对吧。所以那些具备这些解决问题能力的人就被优先选择下来了。但你知道，能把猛犸象赶下悬崖、能在土里种东西固然很好，嗯，但那个，你知道，处在群体最末尾、在那儿琢磨“我到底是什么”的家伙会怎样呢？然后你知道，猎豹就扑上来了，对吧，因为他正陷在那些关于“我到底是什么”的思维循环里，嗯，就像个疯狂的哲学家，对吧。所以我是说，哲学家——我是说，就像你先获得解决问题的能力，达到平均水平，你知道，也许稍微高于平均一点，然后你就得到了这些怪物，那些吓人的、你知道的天才，他们时不时就冒出来，他们总是，他们一下子就跑到那么远的地方去了，你知道，我们跳得太快了，我们还没有
42:21
the way over there you know we jump way too fast we still have we don't have like this infrastructure to like sort of like keep yourself from all these like enemies you know you got to slow down a little bit I mean I think it's exactly like you do have these kind of figures which just seem I mean like Archimedes for example like you want to talk about a guy who I mean the way one of my lecturers at Cambridge used to put him you know pib Dr pibel Hall is he would describe these people as nine-dimensional hyper intelligent beings and like their bodies were just you know their four-dimensional protrusion into our worlds and like that's because were just you know they were like practically aliens which were in terms of their intelligence and we just had no idea what they were doing um and that's almost dangerous right like I mean we tend not to like or really go go favorably with people who seem really out there right like um and we we kind of have a sad history of of persecution right like and humans aren't always very
那种基础设施来保护自己免受所有这些敌人的伤害，你知道，你得稍微慢一点。我是说，我觉得确实就是这样，你确实会遇到这些人物，他们看起来就是——我是说，就像阿基米德这样的例子。你想说一个这样的人物，我在剑桥的一位讲师以前是这么形容他的，你知道，皮博士，Piebalgs 博士，他会把这些人描述成九维的超级智慧生物，而他们的身体只是你知道，他们在我们这个世界里的四维投影而已。就是因为，你知道，就他们的智力而言，他们几乎就是外星人，我们完全不知道他们在干什么。嗯，而这几乎是危险的，对吧，我是说，我们往往不太喜欢、或者说不太善待那些看起来非常另类的人，对吧。嗯，而且我们在迫害这方面有一段很可悲的历史，对吧，人类对高于平均水平的人并不总是很友善。我是说，我刚才是在提出这个
43:22
nice to the above average so I mean I was I was bringing up this idea of of of you know having intelligence and having these problem solving capabilities but that this ability to look back at ourselves and kind of be reflective as not being very good for evolution and I've actually heard it suggested that our brains are actually evolved not to really solve the problem of Consciousness because we're spending too much time well because it wasn't favorable in terms of evolution to actually be able to consider deep and understand what's going on internally having internal model of all your thoughts like we don't have a transcript of why did I make that decision like well you know in sector a this happened and this near on fire and that and this one and that you know is this right we don't have access to that that's in part because Evolution kind of said you know we don't want to give you access to that we're going to put this together in a nice little rat brain you're not going to have access to those thoughts right
想法，就是，你知道，拥有智力、拥有这些解决问题的能力，但是这种回头审视自己、进行自我反思的能力，对演化来说并不是很有利。我其实听人提出过一种说法：我们的大脑其实并没有进化成能真正解决意识问题的样子，因为我们花了太多时间——嗯，因为从演化角度看，能够真正深入思考并理解内部正在发生什么，并不是有利的，也就是拥有你所有想法的内部模型。就像我们没有一份记录说明我为什么做出那个决定，比如“嗯你知道，在 A 区发生了这个，这个神经元放电了，然后那个，还有这个，还有那个”，你知道，是这样吗？我们没法访问那些东西。这在一定程度上是因为演化算是说了，你知道，我们不想让你访问那些，我们要把这一切打包塞进一个漂亮的小老鼠脑袋里，你不会有权限接触到那些想法，对吧，你只会得到它们的结果。所以我——我是说，这是个让人不安的想法。Sandra，你刚才想说
10马斯洛金字塔与牛顿的反例
44:20
you're just going to have their consequences so I I mean it's troubling idea Sandra you're going to say something ADV human besides is way Advance is there any way to advance logical thinking to essentially Advance humans human advancements in thinking maybe there's way that's better okay I mean we didol right right so I mean to an extent we are right right now like in this classroom um and the fact that most of us including myself you know we have bad Vision like if I were really stuck out in the field like I wouldn't have a chance to like sit and like look over books right I'd be Fally hunting and like not succeeding and dying right but our society has reached a level of technology and healthare where people can go on and you know we don't have to worry about based survival needs we've kind of and I don't know how many of you have ever heard of this but um maso's hierarchy of needs anyone heard of maslo have any of you guys done debate in high school do you even have debate teams or classes they don't have these programs
什么？“既然人类已经很先进了，有没有办法提升逻辑思维，也就是从本质上提升人类，让人类在思维上取得进步，也许有更好的方式？”好，我是说，我们确实——对，对，所以我是说，在某种程度上我们正在这么做，对吧，就像现在，在这个教室里。嗯，而且事实上我们大多数人，包括我自己，你知道，我们视力不好。如果我真的被丢在野外，我根本没机会坐下来慢慢读书，对吧，我会拼命去打猎，然后失败，然后死掉，对吧。但我们的社会已经达到了这样的技术和医疗水平，人们可以继续生活下去，你知道，我们不用担心基本的生存需求。我们算是——我不知道你们当中有多少人听说过这个，嗯，马斯洛的需求层次理论，有人听过吗？马斯洛。你们有谁在高中参加过辩论吗？你们这儿有辩论队或者辩论课吗？他们这儿没有这些
45:31
here gasp all right um so maslo said we have this kind of hierarchy of needs before we can do anything and kind of at the top here he had uh kind of essentially Transcendence right um I I'm actually blanking on the exact word which you used um this is embarrassing but either way I want you to Google this Abraham maslo we have hierarchy of needs I can't spell either is there an a here no okay it's just a it's just r h i e r hierarchy of needs thank you and we have this this this self-actualization huh uhuh there you
项目吗？（倒吸一口气）好吧，嗯，所以马斯洛说，在我们能做任何事之前，我们有这么一个需求层次。在这个金字塔的顶端，他放的基本上就是，嗯，超越，对吧。嗯，我其实一下子想不起来他用的确切那个词了，嗯，这有点尴尬，但不管怎样，我想让你们谷歌一下这个，Abraham Maslow，需求层次理论。我也不会拼，这儿有个 a 吗？没有，好吧，就是就是 r，h-i-e-r，hierarchy of needs，谢谢。然后我们有这个，这个自我实现。啊？嗯哼，对了，就是
46:39
go and we have kind of like [Music] food you know Water Shelter and this is really what Humanity spent you know most of our early development just staying at this level um ah fantastic uhhuh and you know he roughly described describes this as physiological needs and then we have safety right so we have you know security of body I'm just going to go ahead and write safety in here pull that out there safety um love belonging
这个。然后我们大概还有【音乐】食物，你知道，水、住所，而这其实就是人类在我们早期发展的大部分时间里一直停留的层次。嗯，太好了，嗯哼。你知道，他大致把这个描述为生理需求。然后我们有安全，对吧，所以我们有，你知道，身体的安全保障。我就直接在这儿写上“安全”吧，把那个拉出来，安全。嗯，爱与归属，
47:34
Este steem you have to at least think somewhat POS positively of yourself and then we get to self-actualization which you know Wikipedia describes as morality creativity spontaneity problem solving lack of prejudice acceptance of facts um and really the argument and a lot of people people use this kind of in a scary slippery slope say well look we're going to protect you by constant 24-hour surveillance because before you guys can even think of of doing anything up here like you just need to be safe in your homes right you need to be safe from terrrace um so before you can even do any of this right you got to have this um which is funny because we tend to do things which would infringe on inherent properties of self-actualization like freedom of speech freedom of thought etc etc right um but you can in some ways say that humans have been trying to climb up this ladder and that even us as individuals in our lives kind of you know our parents provide us with food and shelter um and security and safety um and good parents
尊重——你至少得对自己有一定程度的正面看法。然后我们就到了自我实现，你知道，维基百科把它描述为道德、创造力、自发性、解决问题、没有偏见、接受事实。嗯，而真正的论点是——很多人会用一种吓人的滑坡论证说：“看，我们要通过 24 小时不间断的监控来保护你，因为在你们能想到去做上面这些事情之前，你首先得在家里是安全的，对吧，你得免受恐怖袭击。”嗯，所以在你能做上面这些之前，对吧，你得先有这个。嗯，这挺好笑的，因为我们往往会做一些侵犯自我实现固有属性的事情，比如言论自由、思想自由等等等等，对吧。嗯，但你在某种意义上可以说，人类一直在试图爬上这个梯子，甚至我们作为个体，在我们的人生中也是这样，你知道，我们的父母为我们提供食物和住所，嗯，还有保障和安全，嗯，而好的父母应该给我们一种爱与归属的
48:49
should give us a feeling of love and belonging um and this then enables us to do things like you know go to college and you know Think About Emanuel Kant and other philosophers and stuff yes Latif couldn't like being like under 24 hours surveillance sort of take away our feeling of safety right so I mean the argument goes both ways you could also not feel safe by being Sur surveyed but I mean this is a debate for policy makers not necessarily for me right now um but you're right I mean Sandra I'm glad you brought up this point of this idea of like of humans trying to advance um you know and then it's kind of sad to say but most of us here in this classroom are really occupied with like we're we're all kind of up here in this triangle right and we're worried like well what about logic well is computer science going to lead us to and we're worried about this very Tippy Peak which is Enlightenment right um whereas there are plenty of people out in the streets today who who you know are worrying
感觉。嗯，然后这才使我们能够去做一些事情，比如，你知道，上大学，你知道，思考伊曼努尔·康德和其他哲学家之类的。请说，Latif。“24 小时被监控难道不会反而剥夺我们的安全感吗？”对，所以我是说，这个论证是双向的，你也可能因为被监视而感到不安全。但我是说，这是留给政策制定者的辩论，现在不太该由我来讲，对吧。嗯，但你说得对。我是说，Sandra，我很高兴你提出了这一点，这个关于人类试图进步的想法。嗯，你知道，然后说来有点让人难过，但我们这个教室里的大多数人，其实关注的都是——我们都算是在这个三角形的上面这部分，对吧，我们担心的是，比如，逻辑怎么样？计算机科学会把我们带向哪里？我们操心的是这个最最尖端的顶点，也就是启蒙，对吧。嗯，然而今天街上还有很多人，他们，你知道，操心的是这些下面的东西，对吧。他们担心的不是
49:54
about these things right they're not worried about like is girdle's incompleteness the I'm going to give me my final understanding of Consciousness in the brain right I mean just think about how far of a level that is and even think of how far of a level we've historically come as people and kind of escaping the uh the kind of brutish and harsh natural selection on this level um and we've kind of now been engaging in a selection on this level and as you guys watch in Waking Life Next lecture actually get a really pumped up philosopher who who will talk on this film about this idea of a of kind of Neo Evolution and the Neo what happens when when Evolution starts no longer acting in terms of genes and and security and safety here because we as humans we don't usually worry about this anymore like now evolution is happening on a cultural level on a memetic level um for those of you know Who memes what memes are but we're going tangental right now um and I want to pass things over to Karan who's got lots of goodies for you once
“哥德尔不完备定理会不会最终让我理解意识和大脑”，对吧。我是说，想想这中间隔了多少个层次，甚至想想我们作为人类在历史上跨越了多少层次，算是逃离了这个层次上残酷严苛的自然选择。嗯，而我们现在算是开始在上面这个层次上进行选择了。就像你们下节课看《Waking Life》时会看到的，里面有一位非常激昂的哲学家，他会在那部片子里谈到这个想法，一种新演化（Neo Evolution），以及当演化不再作用于基因、不再作用于这里的保障和安全时会发生什么——因为我们作为人类，通常已经不再为这些操心了。现在演化是在文化层面上发生的，在模因层面上发生的。嗯，你们当中有人知道模因是什么的话——不过我们现在扯远了。嗯，我想把话筒交给 Karan，他又给你们准备了一大堆好东西，
51:02
again um and uh you know go through this idea of of climbing up pyramids and this interplay of of of levels and and descriptions so as such you guys get your five minute break while we reorganize things and see you all back here unbalanced so what are some consequences if those things aren't balanced right so what happens if these son they don't feel safe right okay uh well I think some clear consequences based at least on this model is the idea that if we don't feel safe right we're not going to like want to pursue anything higher right like you know we're definitely not going to feel loved if we don't feel safe right and we're not going to feel we're not going to have any kind of self-esteem if we don't feel loved right how about I don't think he felt SI when he was doing B okay so you're right there are these few kind of crazy characters in history which um uh which have been Ultra paranoid not felt loved and just been essentially like schizophrenic right I mean look at Isaac Newton you want to talk about a not nice
嗯，还有，你知道，会讲讲这个爬金字塔的想法，以及层级和描述之间的这种相互作用。所以你们现在有五分钟休息时间，我们重新整理一下，待会儿在这儿见。不平衡。那么如果这些东西不平衡会有什么后果呢，对吧？所以如果这些孩子他们感觉不安全会怎样，对吧？好，嗯，我觉得至少基于这个模型，有几个很明显的后果，就是说如果我们感觉不安全，对吧，我们就不会想去追求任何更高层次的东西，对吧，就像，你知道，我们肯定不会感到被爱，如果我们感觉不安全的话，对吧。而且我们不会有任何形式的自尊，如果我们感觉不被爱的话，对吧。“那——我觉得他在做那个的时候并没有感到安全。”好，你说得对，历史上确实有那么几个疯狂的人物，嗯，他们极度偏执、没有感受到爱，基本上就像精神分裂一样，对吧。我是说，看看艾萨克·牛顿，你想说一个不友善的人的话，
52:17
man he was a very very evil man in fact as his I think 32 years as as head of the mint where he's responsible for for you know know giving people their lives back because in that in those days clipping coins was punishable by hanging right um and he would never never clear anyone's sentence I mean everybody who went up for coin clipping got hung underneath Newton's so back in the day right coins were actually silver right because the metal itself had inherent value so what people used to do is they would clip little chips of silver from it every time they came into possession of a coin and they have a pile of silver and you could turn it into like another coin right so you just kind of had another coins spring from nothing um of course like you can imagine like um they used you would have to write letters to the warden of the mint saying things like oh I'm so sorry Mr Newton sir like I I clipped only two coins and it was so I could buy an extra loaf of bread for my family you know mes has another you know
他就是。他实际上是个非常非常恶毒的人，比如他担任造币厂厂长的那，我想是，32 年间，在那个位置上他负责，你知道，决定人们的生死，因为在那些年代，剪削钱币是要被绞死的，对吧。嗯，而他从不、从不赦免任何人的刑罚。我是说，在牛顿手下，所有因剪削钱币被送上法庭的人都被绞死了。所以在那个年代，对吧，硬币其实是银的，对吧，因为金属本身就有内在价值。所以人们过去的做法是，每次拿到一枚硬币，他们就从上面削下一点点银屑，然后攒下一堆银子，就可以把它做成另一枚硬币，对吧。所以你就算是凭空又变出了硬币，对吧。嗯，当然，你可以想象，嗯，他们过去——你得给造币厂厂长写信，说些像这样的话：“哦，实在对不起，牛顿先生，我，我只削了两枚硬币，那是因为我想给家里多买一条面包，你知道，我妻子肚子里又，你知道，又怀了一个，我有
53:15
another bun in the oven like I've got three kids I need to Fe feed and you know this guy's probably illiterate too and struggling with writing this letter right and you know Newton's like H like hang him so he wasn't a very nice person but he was brilliant right and and yeah he kind of went higher at this pyramid than any of us can really hope to so yeah there are some defects in this model onization didn't you actually say morality was one of them yeah so you can sort of get rid of it don't need morality but I mean ideally we'd have this kind of now you have to also remember maslo was kind of philosopher of the 60s and 70s who believe that you know enlightened people would be loving and happy and all the time um you want this turn yeah can you actually was in that no I don't think you can right but I mean he was if you want to talk about the qualities of of intelligence and problem solving I mean even when he was old and people are trying to solve the the problem of the brist bistone you know
三个孩子要养。”而你知道，这家伙很可能还不识字，写这封信都很吃力，对吧。可你知道，牛顿的反应就是，哼，绞死他。所以他不是个很友善的人，但他很杰出，对吧。而且，对，他在这个金字塔上爬得比我们中任何人真正能指望到达的都要高。所以，对，这个模型确实有些缺陷。“自我实现——你不是刚说道德也是其中之一吗？”对，所以你可以算是把它去掉，不需要道德。但我是说，理想情况下我们会有这种——不过现在你还得记住，马斯洛算是六七十年代的哲学家，他相信，你知道，开明的人会一直充满爱、充满快乐。嗯，你要接这个吗？对。“你能不能其实就在那里面——”不，我不觉得你可以，对吧，但我是说，如果你要谈智力和解决问题的能力这些品质，我是说，就算他上了年纪大家都在试图解决最速降线问题，你知道的，就是什么形状的滑道能让一个球滚下去时
54:31
the shortest what shape slide would enable when you drop a ball down it to go from point A to point B in the least amount of time like this was being circulated through European and British mathematical journals for years right nobody really came up with a good solution so somebody finally passed it off to Newton you know comes in at like 65 right and you just kind of go don't insult me with these easy problems stood up that night and solved it in you know couple hours like I mean that's just that's just the kind of guy that he was you say I don't think so I don't think you ever let anyone off for 32 years of being one of the me so Karen do you want to talk about this you can start talking about it okay okay all right so this is actually um kind of this this gives you an idea of how current and I plan for lectures um actually this is the only time we've ever used a whiteboard um so we had this idea of is um I wrote this question on the board for occur and I said if he had to administer a test to decide um whether
从A点到B点用时最短，这个问题在欧洲和英国的数学期刊上流传了好多年，对吧，没人真拿出一个好解法，最后有人把它转给了牛顿，那时他大概65岁了吧，他大概就是那种「别拿这种简单题来侮辱我」的态度，那天晚上站着几个小时就把它解出来了，我是说，他就是这样的一个人你说我不这么认为，我觉得你不会放过任何人，32年来一直是那种呃，那么Karen你想讲讲这个吗，你可以开始讲了，好，好，那么这其实呃某种程度上，这能让你们了解我和Curran是怎么备课的，呃其实这是我们唯一一次用白板，呃我们当时有这么个想法，我在板上给Curran写了这个问题，我说如果他要出一场考试来判断，呃，判断（这有点写到边上去了）某个人是不是天生该当计算机
11物理学家、工程师与计算机科学家
56:00
or not it kind of runs off the edge uh someone was destined to be a computer scientist what would it be computer yeah to be a computer yeah I don't I I guess it got clipped somehow um but either way like you know just by forcing you guys to vote by those boxes um I went ahead and I had this I was trying to get at what levels do you are you guys most most interested in um and you know I talked about like if you're a physicist right and you're sitting in kind of high school biology right someone might tell you about organs and things like this and if you were a physicist you would be like okay great whatever tell me more and then like okay well these organisms are made out of tissues and cells and physicist would say uhhuh fine yeah what what else um well those are made out of like you know these big proteins and these kind of organic molecules this this they be like okay um and those are made of you know chemicals which are then made of atoms and then they' be like oh now you're starting to make me interested um
科学家，那这场考试会是什么，「computer」，对，「to be a computer」，对，我不知道，我猜是不小心被截掉了，呃不管怎样，就像你们知道的，通过让你们用那些方框投票，呃我就想借此弄清楚你们最感兴趣的是哪些层级，呃你知道，我讲过比如说你是个物理学家，然后你坐在高中生物课上，对吧，别人会跟你讲器官之类的东西，如果你是个物理学家，你会说，行啊挺好，无所谓，再往下讲，然后是，好吧，这些生物体是由组织和细胞构成的，物理学家会说，嗯哼，行，还有呢，嗯，那些又是由那些大分子蛋白质和有机分子构成的，他们会说，好吧，嗯，而那些又是由化学物质构成的，化学物质又由原子构成，然后他们就会说，哦，现在你开始让我有兴趣了，嗯
57:13
and so then a physicist really occupies themselves with pushing downwards on this level as as as far as possible Right particles forces quirks strings um so you can characterize it as a process of reductionism um and this is even true when you're thinking about large things like uh Supernova galaxies you're trying to reduce these large scale phenomenon in terms of explaining them in fundamental forces f equals ma things like this um but then like what about you know I I forget what's the statistic but I want to say something like 50% plus of all people who for undergraduate Majors do like human ities things like law history creative writing things like this um I mean if you're one of these people like you don't ever even look below this line right like you're really caring about like you if what happens here right what happens when you take your basic building blocks of of say fear hunger desire and selfishness and you take linear combinations of them and like well I'm going to give like 75%
所以物理学家真正投入的，就是尽可能地往这个层级的下方推进对吧，粒子、力、夸克、弦，嗯所以你可以把它概括为一个还原论的过程，嗯而且这一点即使在你思考大尺度的东西时也成立，比如超新星、星系，你是想把这些大尺度的现象还原成基本力来解释，F等于ma之类的，嗯但是那么那些，你知道，我记不清具体数据了，但我想说大概有50%以上的本科专业学生学的是人文类的东西，比如法律、历史、创意写作之类，嗯我是说如果你是这类人，你可能从来都不会往这条线以下看，对吧，你真正关心的是这里发生的事，对吧，当你拿恐惧、饥饿、欲望、自私这些基本构件，然后对它们做线性组合比如说，我要来75%的恐惧加一点点欲望，这就成了一种新的情绪，叫做
58:24
fear with a little bit of desire and that's going to be a new emotion called apprehension right and then as you kind of build up here you just you start going with levels of description starting on the levels of emotions then building up to things especially if you're like kind of a literature major interested in what your language can do um so then I you know we keep asking so like what what's an engineer right I mean who here thinks they like engineering okay great so like you guys see or you might see um just you know everyday materials like well you know you're kind of like your own little mcgyver like give me give me a a toothpick a a a rubber band and you know a piece of dough and I can you know pick somebody's lock right like I mean like that's just what mcgyver does um and you're always kind of or you know I hate to generalize but you're interested in how you can take kind of Base readyy efficient materials around you and then create practical effective Solutions and you're always kind of working in this domain
忧惧，对吧，然后你在这上面不断往上搭，你就开始用一层层的描述层级，从情绪的层级开始，然后往上搭建，尤其如果你是文学专业的，关心语言能做到什么嗯所以我们就一直在问，那工程师又是什么呢，对吧，我是说这儿有谁觉得自己喜欢工程？好，很好，所以你们看到的，或者说你们可能看到的，就是日常的材料，比如说，你有点像你自己的小马盖先，给我一根牙签、一根橡皮筋，再来一块面团，我就能撬开别人的锁，对吧，我是说马盖先干的就是这个，嗯而且你总是，或者说，我不太想一概而论，但你感兴趣的是怎么利用身边现成、高效的基础材料，然后创造出实用有效的解决方案，而你总是在这个领域里干活，甚至可能从小项目做到全城规模的项目，嗯那么我们还是要问
59:30
here maybe even going from small projects to Citywide ones um so then still like we ask so what what makes a computer scientist and I say that really fundamentally a computer science is a lot like a mathematician who doesn't really explore the physical world in its layers of abstraction but instead live in kind of a platonic world and their their atoms are kind of sequences and series and Logics and grammars and calculus and algebra and geometry um and you you just get to play around with with these things um but I would say that inherently kind of what makes a mathematician and computer scientist is looking at just patterns um and kind of abstracting away any details uh I was really sad to when I stopped really being interested in biology because I used to love biology I'm like oh isn't this great look what evolution produces but really once I learned about the theory of evolution I real is well that pretty much explains it right you have couple roll that die you have you know genetic mutation
是什么造就了一个计算机科学家，我会说，从根本上讲，计算机科学非常像数学家，他们并不真的去探索物理世界的各个抽象层级，而是活在一种柏拉图式的世界里，他们的「原子」是序列、级数、逻辑、文法、微积分、代数和几何，嗯你就是可以摆弄这些东西，嗯但我想说，本质上造就数学家和计算机科学家的就是只看模式，嗯并且把一切细节抽象掉，呃我当年挺难过的，我不再真正对生物学感兴趣了，因为我以前特别喜欢生物学，我会想，哦这多棒啊，看看演化造出了什么但其实一旦我学了演化论，我就意识到，这基本上就解释完了，对吧，你有几次掷骰子，你有基因突变、表型、选择、重复，对吧，一旦这个理论摆在我面前，我就
60:41
phenotypes selection repeat right once the theory was there in front of me I said well I'm glad you know Darwin solved that problem I don't have to worry about it and suddenly biology became less interesting to me just because and what I guess ultimately boils say again Sandra it's not my so you're not curious right I mean like the theory is there and and that kind of like I guess puts me at this kind of theorist and philosopher is that I don't care about the details like and you know people spend their entire lives like studying the action of this protein on thing X and Y right um but I don't really care about the details right just the conceptual process um but I don't want to you know bias any of you guys I mean everybody has their own kind of take at things um and you know even Kar and I had kind of differ opinions about what was interesting and what you guys find interesting um so I'm going to turn things over to him so something that was really cool that we noticed about these is the universality of like things being
说，好吧，我很高兴达尔文解决了这个问题，我不用操心了，然后生物学突然就对我没那么有意思了，就因为，最终归结起来，你再说一遍，Sandra，这不是我的——所以你不好奇了，对吧，我是说理论就在那儿了，而这大概就把我归到了理论家和哲学家那一类，就是我不在乎细节，你知道，有人一辈子都在研究某种蛋白质对X和Y的作用，对吧，嗯但我真的不在乎细节，只在乎概念层面的过程，嗯不过我不想影响你们任何人，我是说每个人都有自己的看法，嗯而且连Curran和我对什么有意思都有不同意见，也跟你们觉得有意思的东西不一样，嗯那我就把话筒交给他了，那么我们注意到一件特别酷的事，就是这种普遍性：东西存在于较低层级，而更高层级的东西
12程序演示：简单规则的涌现行为
61:48
on lower levels and things emerging out of those lower levels into higher levels of things like um electrical engineering like you said transistors and stuff you have all these different layers of things that can happen up to software and with neurons and the brain and thoughts and with Biology like you know DNA and RNA and replication and whatnot reproduction and then that leads to Evolution which is an emerging property so I'm going to show you some programs that I wrote that exhibit some emerging properties so this is um sort of a physics simulation but it's discrete it's not continuous and physics is modeled you know as as continuous you know you integrals and stuff like that so here I'm sort of doing integration but it's discret so you have this sort of weird error you know that makes it not exactly like physics but it's still really cool so um the blue and the red things have different charges and there are kic forces acting between them so I'll just play with it um colic forces is like
从这些低层级中涌现出来，比如电子工程，像你说的晶体管之类的，你有各种不同的层次，一路可以上升到软件；神经元、大脑和思维也是这样；生物学也是比如DNA、RNA、复制之类的，繁殖，然后这又导致了演化，而演化就是一种涌现属性。所以我要给你们看几个我写的程序，它们展示了一些涌现属性。那么这个是呃，算是一个物理模拟，但它是离散的，不是连续的，而物理通常是被建模成连续的，你知道的，积分之类的。所以我在这里算是在做积分，但它是离散的，所以你会有一种奇怪的误差，让它跟真正的物理不完全一样，但还是挺酷的。所以呃蓝色的和红色的东西带不同的电荷，它们之间有库仑力在作用。我就随便玩一下，嗯库仑力就是
63:09
plus and minus you know opposites are tract and when they're the same they repel each other and so we can change you know all these properties and the color reflects how charged they are so this things that's really weird like this this would never happen in like in real physics like where's this strange energy coming from and I think it's because it's um discrete it's discretized and not continuous but it's still pretty cool pretty interesting uh yeah repolish and now it's stable so what what I want you to keep in mind in looking at all these is that with every one of these examples it's very simple rules or pretty simple rules that govern the behavior of each individual one and the rules that govern one of these particles one of these balls is no different than the rules that govern all the other ones so these simple rules that are local lead to Global phenomenon emergent Behavior this is what emergence is all about things at higher levels emerging from you know simpler things on lower
正和负，你知道的，异性相吸，同性就互相排斥。我们可以改变所有这些属性，颜色反映的是它们带电的多少。所以这个东西特别奇怪，像这样，在真实物理里绝不会发生，这股奇怪的能量是哪来的？我觉得是因为它是离散的，被离散化了而不是连续的，但还是挺酷的挺有意思的，呃对，重新抛光一下，现在它稳定了。所以我希望你们在看这些的时候记住的是这每一个例子里，支配每一个个体行为的都是非常简单、或者说相当简单的规则而且支配其中一个粒子、一个小球的规则，跟支配其他所有粒子的规则没有任何区别。所以这些局部的简单规则导致了全局的现象，也就是涌现行为，这就是涌现的全部要义：高层级的东西从低层级更简单的东西中涌现出来。所以举个例子，自然界中的
64:38
levels so for example like crystallization in nature is an example of an emerging property so here's a set of presets that a set of settings that leads to crystallization like it's like tuned to crystallize that that's an interesting thing to point out is that we have this level of description which we just described we call this a crystal right we don't say it's the arrangement when particle I acts on particle J and has the following charge right like that level description is too fundamental but this higher level description of a crystal is much more convenient right yeah yeah if we do say crystal we mean exactly the same thing it's like we're just dodging around the exactly what the thing is so you you said that when we say crystal we're not really saying the thing we're saying we're sort of you know beaing around the bush right saying yeah it's like okay so we have these crystals there they're like like uh compositions of different particles that usually attct each other to like certain
结晶就是涌现属性的一个例子。这里有一组预设，一组会导致结晶的参数设置，像是被调好了专门用来结晶的。有一点值得指出，就是我们有了这个描述层级，我们刚才描述的，我们把这个叫做晶体对吧，我们不会说这是粒子i作用于粒子j、带有如下电荷时的排布，对吧，那个层级的描述太底层了，而晶体这个更高层级的描述要方便得多，对吧，对对，如果我们说晶体，我们指的其实是完全一样的东西，就好像我们在绕开这东西到底是什么，所以你刚才说，当我们说晶体的时候，我们并没有真的在说那个东西本身，我们算是在绕圈子，对吧，说，对，就是说，好，我们有这些晶体，它们像是呃由不同粒子构成的组合，这些粒子通常互相吸引，形成呃某种容易辨认的形状，���就是晶体的定义，但
65:56
like uh easily discernable shapes that's the definition of c but once you like see it as it is it's like uh it's like the set of all these uh like different um configurations of matter such that these properly holds you know you just like yeah I mean just put giving it a name for some that's exactly you're exactly right so we say crystal we mean a very sort of high level thing and this is what I mean when I say low level and high level low level means like describing the exact way that the particles interact and different you know things that they do but crystal is a higher level and so the lower level details could be different you know crystals form in you know out of all kinds of different substances in nature and this we can call it a crystal because it sort of resembles the crystals in nature that it has this regular structure but yeah you're right it's sort of glossing over all the details it's existing at a higher level it's a higher level of description so it's like is this mathematical theory of
一旦你如实地去看它，它其实就像是呃，就是所有这些呃不同的物质构型的集合，使得这些性质成立，你知道的，你只是——对，我是说只是给它起了个名字，没错，你说得完全对。所以我们说晶体，我们指的是一个非常高层级的东西，这就是我说低层级和高层级时的意思。低层级是指描述粒子相互作用的确切方式和它们做的各种事情，而晶体是更高的层级，所以底层细节可以不一样，你知道，自然界中各种不同物质都能形成晶体而这个我们也能叫它晶体，因为它有点像自然界中的晶体，具有这种规则结构，但对，你说得对这算是掩盖掉了所有细节，它存在于一个更高的层级，是一种更高层级的描述，所以这就像是数学里的范畴论，你可以说，好，我们有这些不同的领域
66:58
category Theory so you can say okay we have like these different fields everything but what happens over here can be mapped to exactly what happens over there so like when you're like uh describing things um as crystals or whatever you just like almost speaking like category Theory language like it's just like normal speech Justin knows more about category Theory I have no idea about category Theory yeah oh I mean one want to just tell you to be careful but two yes you're kind of right like you're just looking for General features and kind of systems and and then kind of ascribing to that universality right exactly you know I wanted to point out really quickly that what I wrote up there on the board you know fi Jal K Qi QJ over r² I mean that's just your your your rule of Attraction for cool right like be um and this is how and we have this kind of level of description for the interaction between two particles but what happens when we have you know in particles right something the equations
什么的，但这边发生的事情可以精确映射到那边发生的事情，所以当你呃把东西描述成晶体之类的时候，你几乎就是在讲范畴论的语言，只不过它是日常语言。Justin对范畴论懂得更多，我对范畴论一无所知，对，哦我是说，第一是想提醒你小心一点，但第二，对，你说得有点道理，你就是在寻找系统的一般特征然后把这种普遍性归到它上面，对吧，没错。你知道我想很快地指出一点就是我在板上写的那个，你知道，F_ij等于k乘以Qi乘以Qj除以r平方，我是说那只是你的库仑吸引定律，对吧，嗯这就是——我们对两个粒子之间的相互作用有了这样一个描述层级，但当我们有n个粒子的时候会怎样呢，对吧，那些方程
68:07
become really hard to solve but we start getting really interesting geometric Behavior which is easier to describe about t right um but we're going to kind of hit your idea um later because we're going to see examples where we see the similar concept of force but there's no forces going on right we we'll talk about that in traffic flow and things like that we don't actually have cars ramming into each other but we still get the same behavior but we'll save that so this this is another set of you know parameters that leads to this droplet sort of forming which is pretty cool and here's the the end body problem being simulated oh there it is so it's just things orbiting around each other based on pretty much that equation that he wrote on the board yeah except you replace the charges with masses because q i and QJ represent charges you masses something like that yeah but that's it it's just that's one that's unpredictable right yeah this is it is predictable but you have to remember the it's deterministic not predictable it's
就变得非常难解，但我们开始得到非常有趣的几何行为，而这些反而更容易描述对吧，嗯不过我们待会儿会回到你的想法，因为我们会看到一些例子，里面有类似力的概念，但其实并没有真正的力在起作用，对吧，我们会在交通流之类的例子里讲到，我们并没有真的让车撞在一起，但我们还是得到了同样的行为，不过这个先留着。所以这个这是另一组参数，会导致形成这种液滴，挺酷的。然后这里是n体问题的模拟，哦出来了，就是些东西绕着彼此转，基本上就依据他在板上写的那个方程，对只不过你把电荷换成质量，因为Qi和Qj代表电荷，你换成质量之类的，对，但就是这样。这个是不可预测的，对吧？对，这个——它是可预测的，但你得记住它是确定性的但不可预测，它是确定性的而不可预测，它是混沌的，因为它变成非线性的了，对吗？所以唯一
69:17
deterministic and not predictable it's chaotic because it's it becomes nonlinear is that correct so the only you know what's going to happen is to run it yeah and even then you're approximating it's not continuous so you're going to get you know an approximation of what might happen so yeah you can't really solve the EMB body problem so what if you you got like your um your mind right so what if if if it is like a different bunch of programs like even in that thing you only like talking about party CU those are like simple stuff but when you have Minds you have like a bunch of these stuffs going all together when you have what when you have Minds mind yeah so when you consider a mind to be one of these particles right in society interacting with all the other Minds around it right this is called agent-based modeling this is Agent based modeling flocking is when you have but the problem is for the party you have definite rules but for societies even the minds themselves are hard to prodict
知道会发生什么的办法就是把它跑一遍，对，而且就算跑了你也是在近似，它不是连续的，所以你得到的是可能发生的情况的一个近似，所以是的，你没法真正解出n体问题。那如果是你的呃你的心智呢，对吧，那如果如果它像是一堆不同的程序呢，就算在那个东西里你也只是在讲粒子，因为那些都是简单的东西，但当你有心智的时候，你就有一大堆这样的东西同时在运作，当你有——什么，当你有心智，心智，对，所以当你把一个心智看作这些粒子之一，对吧，在社会中与周围所有其他心智相互作用，对吧，这叫做基于主体的建模，这就是agent-based modeling，群集行为就是当你——但问题是对于粒子你有明确的规则，而对于社会，连心智本身都很难预测，对，没错，所以你是说
70:20
yeah exactly so you're saying you know if we if we model society as a this agent based model where each agent consists of a mind it's even you know more impossible to predict because the mind itself is is not really predictable the Mind itself is emerging out of the things that comprise it so yeah I mean that's that's a idea of going backwards right starting with behavior and then trying to figure out what rule governs it right so it's kind of like how did Newton figure out this Law of Attraction right you start from the behavior try to or infer better yet what the law between two things is and what we're going to see today both cars and I guess even with people is we don't know the people we don't know the interaction between two people but we see the overall behave so here's another version of that program where you can get these uh molecules to form which is really fascinating these like stringy molecules so it's another kind of emergent Behavior here see there they go it's totally a molecule pick it
如果我们把社会建模成这种基于主体的模型，每个主体都是一个心智，那就更加不可能预测了，因为心智本身就不太可预测，心智本身是从构成它的东西中涌现出来的，所以对，我是说这就是反过来做的思路，对吧，从行为出发，然后试着弄清楚是什么规则支配它，对吧，所以这有点像牛顿是怎么发现这个引力定律的，对吧，你从行为出发，试着或者说更准确地说，推断出两个东西之间的定律是什么。而我们今天要看到的，无论是车还是我想连人也一样，就是我们不知道人，我们不知道两个人之间的相互作用，但我们能看到整体的行为。那么这里是那个程序的另一个版本，你能让这些呃分子形成，特别有意思，这些像链条一样的分子，所以这是另一种涌现行为。看这儿，它们出现了，完全就是一个分子，把它挑出来，规则基本上是
13缠结的层级：软件能控制硬件吗
71:33
out and the rules are pretty much the same I don't know exactly what particular rules they are and here it is in 3D also so if we wait for a minute there these molecules are going to form in 3D so I mean yeah emergence is really a cool concept so here here's this molecule in 3D that the world is constantly describing itself in like different levels of description could it be that the world is describing itself in levels different levels of description yeah because you do have that and then you also have like something higher like a geometrical shape on yeah something like that but you can also like describe it as a bunch of um different par DED by different forces you can also like describe it as somebody actually looking at that thing itself yeah who's doing the describing yeah yeah I mean it's it's what Douglas hofs sat calls a tangled hierarchy hierarchy right where it's not really a hierarchy because things on low lower levels are related to things on higher levels and can have influence back and
一样的，我不太清楚具体是哪些规则。这里还有3D版本，所以如果我们等一会儿这些分子会在3D里形成。所以我是说，涌现真的是个很酷的概念。那么这里就是这个3D的分子。世界一直在用不同的描述层级来描述它自己，会不会世界就是在用层级描述它自己呢不同层次的描述，对，因为你确实有那个层次，然后你还有更高一层的东西，比如说一个几何形状之类的，对，差不多是那样，但你也可以把它描述成一堆被不同的力所支配的不同粒子你还可以把它描述成有个人正在看着那个东西本身，对，也就是那个正在做描述的人对对，我是说，这就是道格拉斯·霍夫斯塔特所说的"缠结的层级"，对吧，它其实并不是一个真正的层级，因为低层次的东西和高层次的东西是相关的，而且可以来回互相
72:50
forth so just like you said you know the world is constantly describing in itself interacting with itself you know between different layers different levels and it's just like we higher level also like influences the lower level or can it was just like lower to Upper so you said can we can we say that the higher level influences the lower level or does it always go upwards well no it definitely does not go only upward because think of software right if I run this program this next program which we'll just watch for a while this program itself is controlling the transistors and whatnot that's operating but it's not really controlling the transistors it's not really controlling it no no no no look look you got the code you put it into the ram this something like all it's like represented by like different electrical activities and those electrical activities are like computed and they like different like Hardware is attach those and those like different codes like match like different hard behavior and then you
影响。所以就像你说的，世界一直在描述它自己、和它自己互动，在不同的层、不同的层次之间。就好像高层次也会影响低层次，还是说它只是从低到高？所以你刚才说，我们能不能说高层次会影响低层次，还是说它总是往上走？嗯，不，它绝对不是只往上走的，想想软件嘛，对吧，如果我运行这个程序就是接下来这个程序，我们先看它一会儿，这个程序本身在控制着晶体管之类的东西，也就是在运行的那些但它其实并没有在控制晶体管，它其实并没有在控制——不不不不，你听我说，你有那段代码，你把它放进内存里，它就是被不同的电活动所表示然后那些电活动被计算，然后不同的硬件跟这些绑在一起，那些不同的代码对应不同的硬件行为，然后你就有了你那个小小的软件程序，对吧，并没有什么更高层的东西在控制低层，就好像
73:56
have your little soft program right there's no higher thing con the low like the lower one always yeah it's all one it's all one thing it's just we think about it in terms of high level and low level so you're right I mean the hardware is controlling itself right but it's based on what we put into it and what we say tell it to do it at a higher level so I mean the software that I wrote which came from my mind which CHS if it's at a higher level in a sense than the hardware like this projector that's actually projecting these pixels I mean so that's that's what I mean when I say the higher level things influence you know affect control even in this case the lower level things yeah it goes both ways and like if a if a transistor would have crap out right now it would control the higher level things because it would just stop working it's of causality I guess causality go that doesn't make any sense okay how can a software program control Hardware so this software program is running on hardware and it's controlling
总是低层在起作用——对，它们是一体的，它就是一个整体，只是我们习惯用高层次和低层次去理解它。所以你说得对，我是说硬件在控制它自己，对吧，但那是基于我们放进去的东西、基于我们在更高层次上告诉它要做什么所以我是说，我写的那个软件来自我的头脑，从某种意义上说它处在比硬件更高的层次比如这台正在投出这些像素的投影仪。所以这就是我说的意思，当我说高层次的东西影响、作用于、甚至在这种情况下控制低层次的东西时。对，它是双向的。而且如果一个晶体管现在坏掉了，它就会控制高层次的东西，因为它会直接停止工作。我猜这算是因果性吧，因果性——那说不通啊，好吧，一个软件程序怎么可能控制硬件呢？所以这个软件程序运行在硬件上，它在控制这台投影仪，对吧，它在控制
75:04
this projector right it's controlling the hardware physical tell you what to do or is the hardware telling itself what to do the hardware is telling itself what to do but only after I've told it what to do right but even if we were to remove current right and we just had the software executing by itself um you're right in the sense that it's always just the hardware it's just the hardware right software doesn't really like exist right in this you know ethereal realm right it's software is still just what's going on on the level transistors and stuff it's just that we as you know as humans use these levels of description and and we talk about the software as its own entity even though it's fundamentally still just caused by the interactions of electrons and you know transistors and such so here's another example which is some traffic flow so traffic flow is is an perfect example of um emergence and what it means so like these these little bars it's just like one lane of traffic that's just going and it and it repeats
硬件，物理地告诉你该做什么，还是说是硬件在告诉它自己该做什么？是硬件在告诉它自己该做什么，但那也是在我告诉了它要做什么之后，对吧。但就算我们把电流去掉，只剩下软件自己在执行嗯，你说得对，从某种意义上讲它一直都只是硬件而已，它就只是硬件，对吧，软件并不是真的存在于某个虚无缥缈的领域里，对吧，软件说到底还是晶体管这一层上发生的事情，只不过我们作为人类会使用这些描述层次，我们会把软件当成一个独立的实体来谈论，尽管它归根结底仍然只是由电子和晶体管之类的相互作用造成的。那么再举个例子，交通流。交通流是涌现的一个完美例子，能说明涌现是什么意思。这些小条就是一条车道上的车流，一直在跑，而且是循环的
76:16
itself but so each each one of these cards is like an agent right it's it's a low lower level of description than the wave waves that we that we'll see right here so say there's a red light right so the traffic gets backed up a little bit and then the light turns green again and then they go and so you see this thing on a higher level it's this wave which is propagating back so we say the wave you know is a thing and we can describe it as an entity but it's at a higher level than the cars themselves even though it's comprised of the cars it affects the cars and the cars affect it that's the same thing in with software and Hardware still don't you still don't get it I guess it's really just an issue of reductionism versus holism yeah which is exactly you feel like everything can be understood reductionist thinking sometimes have well at times you just have to like think okay how how does like uh me for instance how do I come from like interactions of particles you have to like do synthesis in that
但每一辆车都像是一个智能体，对吧，它是比我们接下来会看到的"波"更低层次的描述。比如说有个红灯，对吧，车就堵起来一点然后灯又变绿了，它们就往前开。于是在更高的层次上你看到了这个东西，就是这个波，它是向后传播的。所以我们说这个波是一个东西，我们可以把它描述成一个实体，但它比汽车本身处在更高的层次尽管它是由这些汽车构成的。它影响汽车，汽车也影响它。这跟软件和硬件是一回事还是不明白？你还是没搞懂？我猜这其实就是还原论和整体论之争的问题，对，这正是——你觉得一切都能用还原论的思路来理解，有时候你确实得想想，比如像我这样一个人，我是怎么从粒子的相互作用中产生出来的？你得在这里做某种综合
14观察者与被观察者：我存在吗
77:25
but that that does that explain you right and H ask another question the same chapter is so the guy who runs the 100 meters in 9.3 seconds right where is the 9.3 stored it's not right CED in his brain and he's he why he calls those brain signals is like 9.3 that's why he say right no but 9.3 the fact that he ran the 100 meters in 9.3 seconds was the results of you know training him getting good traction as a star I mean it was really an emerging thing right it's an epop phenomenon HS it so here's here's a good question to ask yourself regarding things that are emergent you know do you exist because what are you you ask yourself like what am I uh it it's it's it's an area where it's easily confusing because like you know you're an emerging property from the things that you're made of the problem is it's not to say do I exist I mean the answer to that question is almost the same answer to that to the question what am I if I describe myself as like a some there's like two almost two different parts of myself now myself as
但那真的能解释"你"吗？对吧。再问一个同一章里的问题：那个跑100米用了9.3秒的家伙，那个9.3存在哪儿？它并没有编码在他大脑里，他——为什么他把那些大脑信号叫作9.3？所以他才说，对吧。不，但9.3——他用9.3秒跑完100米这个事实，是训练、起跑抓地良好、成为明星等等的结果，我是说这真的是个涌现的东西，对吧，它是一种副现象。所以关于涌现的东西，这里有个很好的问题可以问问自己：你存在吗？因为你到底是什么？你问自己：我是什么？呃，这是一个很容易让人困惑的领域，因为你知道你是构成你的那些东西的一种涌现属性。问题不在于"我存在吗"，我是说那个问题的答案几乎就等同于"我是什么"这个问题的答案。如果我这样描述我自己的话我自己好像有两个不同的部分。一个是由大脑过程所描述的我
78:55
described by like um brain processes like um maybe like some part of my brain has like a description of what like I've done I've always done and also my beliefs about myself and also another part of the brain has like almost a way of me like going back and also like maybe having this self-conscious awareness of like seeing the world like leaving my description of myself yeah so which one am I well probably I am the thing that the brain describes I am the self I am not the thing that's aware of the self because the thing that's aware of the self-esteem is only exists to to be aware of that self so I cannot be the awareness I must be the one that's been aware them exp so you're basically dividing yourself into the Observer and The observed right the thing that I'm not the Observer I'm the one that's being observed so you're the one that's being observed so what is it that's observing that it's not you what it is it's like this like added structure of like awareness so that's a that's a you know
比如我大脑的某个部分保存着我做过什么、一直以来怎么做的描述，还有我对自己的信念；另一个部分则像是有一种让我回过头去、并且或许拥有这种自我意识的觉知，去看这个世界、去脱离我对自己的描述，对吧，那么我到底是哪一个？嗯，我大概是大脑所描述的那个东西，我是那个"自我"，我不是那个觉知到自我的东西，因为觉知到自我的那个东西存在的唯一目的就是去觉知那个自我。所以我不可能是那个觉知，我一定是被觉知的那一个。所以你基本上是把自己分成了观察者和被观察者，对吧，我不是观察者，我是被观察的那一个。所以你是被观察的那一个，那么正在观察的又是什么？它不是你？它是什么？它像是一种额外附加的觉知结构。所以这是——你知道，这些正是佛教在苦苦思索的
80:08
these are the issues that Buddhism grapples with and it well I don't know I'm not you know enlightened I don't know the answers so I'm still as confused as you are like I think one of the main problem is that like you think too much if if humans just stop thinking everything will just be fine so Latif just said if humans just stop thinking everything will be fine and that's what the Zen Masters say also well it's also saying we should just you know be trees right which also just says like let's give everybody you know mandatory frontal lobotomies right so no one can like think any AB thoughts or get worried about anything we'll just be reduced to basic hle hunting surviving if we can yeah it's like what G says what once you like stop start talking to yourself you're speaking nonsense basically start talking to yourself or about talking about yourself you know it's like I'm okay here's me if that makes any sense of course who's speaking um talking about myself yes the my self-reflective process that
问题。而且，嗯，我不知道，我又没开悟，我不知道答案，所以我跟你一样还是很困惑。我觉得主要问题之一是你想太多了，如果人类干脆不想了，一切就都会好起来。Latif刚才说，如果人类干脆不想了，一切就都会好起来，禅宗大师们也是这么说的。不过这也等于说我们应该干脆当树，对吧，也等于说给每个人强制做个额叶切除手术，对吧，这样就没人能想任何抽象的东西、或者为任何事焦虑了，我们就退化回最基本的采集狩猎、活下去如果还能活下去的话。对，就像G说的那样，一旦你开始自言自语，你基本上就是在胡说八道开始跟自己说话，或者说谈论你自己，你知道那种感觉，就像"我没事，这就是我"，如果这说得通的话。当然了是谁在说话呢？嗯，在谈论我自己？对，就是我那个自我反思的过程在谈论我自己，它来自我心智中最抽象的那些地方
81:22
I'm talking about myself like some of the most abstract places of my mind that means they don't know the details therefore they should not really be talking about myself so you're essentially arguing that self- reference isn't like a well formed thought no not no yeah okay yeah I mean that that's true like when you when you talk about yourself the thing that's talking is not well informed about what it's actually talking about right so back to the complexity of the brain because the brain is so complex that it can't even talk about itself that talking about nonsense you know it's just like almost like looking around like on a weird fract on a weird pattern it's like oh this may be this may be this may be it's like talking nonsense that going random yeah so essentially you're saying talking about yourself is just sort of rambling nonsense that's like sort of like intuition like our mathematical intuition you see like some structure here some structure there you say oh yeah that that may that may like refle
这意味着它们并不了解细节，因此它们其实不该谈论我自己。所以你本质上是在说，自我指涉不算是一个构造良好的想法？不，不是——对，好吧，对，我是说这确实没错，就像当你谈论你自己的时候，那个在说话的东西对它实际在谈论的东西并不了解对吧。所以回到大脑的复杂性，因为大脑太复杂了，复杂到它甚至没法谈论它自己那种谈论就是胡说八道，你知道，就好像在一个奇怪的分形上、在一个奇怪的图案上东张西望，就像"哦这可能是、这可能是、这可能是"，就是在胡说八道，随机乱走。对，所以你本质上是说谈论你自己就是一种漫无边际的胡言乱语，有点像直觉，就像我们的数学直觉，你在这儿看到一点结构、那儿看到一点结构，你说"哦对，那可能反映了
82:26
some mathematical laws or something but it's just a hypothesis it's not really it doesn't have to be true yeah I mean so so you're getting at some really deep questions that we all have to face if we choose to face them so I mean I encourage you to to keep say stop yeah I mean that that's totally it because if we keep thinking about these things get stuck either we'll get stuck and go and and not be able to function or we'll become enlightened maybe I mean I don't know what that is even no I mean be bold right but not too bold right and this gets back to what Justin was saying about earlier like you know introspecting and asking these questions it's not uh favorable to our survival you so we in a sense we've been programmed to just ignore them you know and just live our lives but we're hoping to evolve but yeah we're hoping to transcend that perhaps one would build a conscious machine or something like that we shouldn't like give it the ability to analyze itself maybe like the worst thing any
某种数学规律"之类的，但那只是个假设，不一定是真的。对，我是说，你触及了一些非常深刻的问题，如果我们选择去面对，我们都得面对它们。所以我鼓励你继续——你说要停下来。对，我是说这完全对，因为如果我们一直想这些事情，我们要么会卡住没法正常生活，要么可能会开悟吧，我是说我都不知道那到底是什么。不，我是说要大胆对吧，但也别太大胆。这又回到了Justin之前说的，你知道，内省、问这些问题这对我们的生存并不有利，所以从某种意义上说我们被编程为忽略它们，你知道，就好好过日子。但我们希望能进化——对，我们希望能超越这一点。也许有人会造出一台有意识的机器之类的，我们不应该给它分析自己的能力。也许一个人能做的最糟糕的事，就是真的知道
83:38
person can ever do is to actually know how they think how who they actually really are I think HP left once said something about that you know it's like the most the best thing that we women we I think ever have is that we in any given time we don't ever have like a complete Motel of the world because if we ever did was so frightening I mean but exactly on the other hand though we're inherently curious right and we inherently want to understand bits of things um and I don't know if we can ever have a complete model of the world of as it is um but we can sure as heck try right and you know the bottom line is you should just do whatever makes you feel good understanding part of the brain does that go for it right you can never fully understand the brain don't let that stop you from trying look it's like one of those existentialists you know you got like a man like holding a stone up a mountain or something once they get there the the stone is dropped and then go back again you know the goal is pointless yet
自己是怎么思考的、自己究竟是谁。我记得H.P.洛夫克拉夫特好像说过类似的话，你知道，就是说最——我觉得我们所拥有的最好的东西，就是在任何时刻我们都不曾拥有一个关于世界的完整模型因为如果我们真有了，那会太可怕了。我是说，但另一方面，我们天生就好奇，对吧，我们天生就想理解一点一滴的东西。嗯，我不知道我们是否能拥有一个关于世界本来面目的完整模型，但我们当然可以拼命去试，对吧。而且说到底，你就应该去做任何让你感觉好的事情，理解大脑的一部分能让你感觉好，那就去做，对吧。你永远不可能完全理解大脑，但别让这一点阻止你去尝试。你看，这就像存在主义者说的那样，你知道，有个人推着一块石头上山之类的，一旦推到山顶，石头就滚下来，然后又得重新来过，你知道那个目标是没有意义的，但
84:49
they're like doing the process so I mean you're essentially arguing that understanding the brain is go now why do something that you never what what's the next step oh you make yourself better okay so yeah I don't know I Sandra's idea what what was Sandra's idea like if you like there's puzzles right if you finish the puzzle you're done what else is you make yourself smarter so you can be harder to like understand your
他们仍然在做这个过程。所以你本质上是在说，理解大脑是——那为什么要去做一件你永远做不完的事？下一步是什么？哦，让你自己变得更好。好吧，所以，我不知道，Sandra的想法。Sandra的想法是什么？就像，如果你有些谜题，对吧，你把谜题解完了就结束了，那还剩什么？让你自己变得更聪明，这样你就更难被理解
85:30
so what you're getting at actually you know cuz what what you're saying is sort of um like this like once you understand a certain amount of yourself you've added another part of Another Part of Yourself which you don't understand you mean the part that actually understands itself is the part that you actually don't understand yeah so once once you um once you come to terms with yourself or so you think that part of you which has just come to terms with itself you don't understand that is analogous to adding G to number Theory because it can be galzed again and be proved incomplete yet again and then you can say oh well it's not incomplete now so I can just add that in this is a big part of uh some chapter in gerbach I forget which one but it's essentially you can never understand yourself I mean it's a dangerous use of girdles and completeness theem it's probably not well uh you know based but it's I don't know something to think about always do like the complexity thing you know I think there was some
所以你其实想说的是，因为你说的差不多是这个意思：一旦你理解了自己的某一部分你就又给自己添加了另一个你不理解的部分。你是说那个真正理解自己的部分，恰恰是你实际上不理解的部分。对，所以一旦你、嗯，一旦你与自己达成和解，或者你以为如此，那么你身上那个刚刚与自己达成和解的部分，你并不理解它。这类比于把哥德尔句子加进数论里因为它可以被再一次哥德尔化，再一次被证明是不完备的，然后你可以说"哦那现在它不是不完备的了所以我把那条也加进去就好了"。这是《哥德尔、埃舍尔、巴赫》里某一章的重点，我忘了是哪一章，但它本质上是说你永远无法理解你自己。我是说，这是对哥德尔不完备定理的一种危险用法，可能站不住脚，但它，我不知道，是值得想一想的东西。总是像那个复杂性的说法一样，你知道，我记得有个
86:44
guy that he did some calculations and he came out you know once his system becomes so complicated he reaches a point that he can't understand so if you could pull that reference that it' be good good to see but yeah tell me if you I think somebody already I don't know who it is yeah I mean once a system gets to a certain point of complexity I if somebody else can't understand it how can understand yeah I mean we can't understand ourselves in the sense of everything that's going on because we can't introspect on our own neurons right you know our neurons in in the case of agent based models like we're talking about like check this one out this is pretty cool um the the balls on top are being pulled upwards and the the ones here are being pulled downwards so there's a sort of stability and I think there are the same number of balls um I'm not sure but somehow I made it so they equal out and this is just agent based modeling applying these Force equations but only between the balls that are connected to each
家伙做了一些计算，他得出的结论是，一旦他的系统变得足够复杂，就会达到一个他自己也无法理解的临界点。所以如果你能找到那个出处就好了，很想看看。对，如果你——我觉得已经有人做过了，我不知道是谁。对，我是说一旦一个系统复杂到某个程度，如果别人都理解不了它，那它怎么可能理解自己。对，我是说我们没法在"了解正在发生的一切"这个意义上理解自己，因为我们没法内省自己的神经元对吧。说到我们的神经元，在基于智能体的模型这方面，就像我们刚才说的，看看这个这个挺酷的，嗯，上面这些球被往上拉，这边这些球被往下拉，所以形成了某种稳定状态，而且我觉得两边球的数量是一样的，嗯，我不太确定，但不知怎么我把它做成了两边正好平衡。这就只是基于智能体的建模，套用这些力的方程，但只在彼此相连的球之间计算。所以我是说它挺接近
15宇宙只能模拟自己：描述的必然近似
87:55
other so I mean it's pretty close to physics and it's pretty cool so what was I going to say about agent can someone remind me what we were talking about so I mean I wanted to kind of comment about this idea of modeling the Universe um if you're really interested I recommend a book called program in the universe by Seth ly um and essentially it says well the universe at the bottom uses quantum mechanics right and we can't model quantum mechanics effectively on a class classical computer and in fact we need a quantum computer but the universe itself is a quantum computer and the only thing which could model the universe is the universe itself right so we can't build I mean we'd have to build another Universe to model the other one but why do that when we already have the universe which is Computing itself right so I mean it's a very interesting book and Seth ly is a very prominent um physicist and he's a mechanical engineer here at MIT who works with this Santa Fe Institute complexity science um and it's
物理的，也挺酷的。那我刚才想说关于智能体的什么来着？谁能提醒我一下我们刚才在聊什么？嗯，我本来想评论一下"给宇宙建模"这个想法。如果你真的感兴趣，我推荐一本书，叫《编程宇宙》作者是Seth Lloyd。它基本上说，宇宙在最底层使用的是量子力学，对吧，而我们没法在经典计算机上有效地模拟量子力学，实际上我们需要一台量子计算机，而宇宙本身就是一台量子计算机，唯一能给宇宙建模的东西就是宇宙自己，对吧。所以我们造不出——我是说我们得再造一个宇宙来模拟这个宇宙，可我们既然已经有了这个正在计算自己的宇宙，何必再去造呢，对吧所以我是说这是一本非常有意思的书，Seth Lloyd是一位非常著名的物理学家，他也是MIT这里的机械工程教授，和圣塔菲研究所的复杂性科学有合作。嗯，这是一本非常非常好的书，但我
89:04
a it's a very good very good book but I think it gets kind of at the Heart of Heart of the problem that really the tools that we need to describe the universe is really the bits of the universe itself right but don't you have like some of those weird properties of infinites you know when like the thing like the whole thing is infinite right but even in that thing itself there's another thing that's also infinite so you can have like a Quantum mechanic computer inside a universe that's also how sort of like emulating the universe yeah but it can only emulate pretty much part of the universe of equal size right a quantum computer involving seven Quantum bits kind of like essentially model something that big right but the information con the universe really is equal to the amount of information the universe can compute with right so you need a computer as big as the universe the model Universe down to every detail down to every detail so all we can do is approximate on higher levels and do this chunky descriptions
觉得它触及了问题的核心：我们用来描述宇宙所需要的工具，其实就是宇宙自身的那些比特，对吧。但你不是会遇到那些关于无穷的奇怪性质吗？你知道，就是整个东西是无限的，对吧，但在那个东西内部还有另一个东西也是无限的，所以你可以在一个宇宙里放一台量子力学计算机，它也在某种程度上模拟这个宇宙。对，但它最多也只能模拟个大概宇宙中同样大小的一部分，对吧？一台包含七个量子比特的量子计算机，基本上可以模拟差不多那么大的东西，对吧？但信息……宇宙真正的信息量等于宇宙所能计算的信息量，对吧？所以你需要一台和宇宙一样大的计算机，才能把宇宙的每一个细节都模拟出来，每一个细节，所以我们能做的只是在更高层次上做近似，做这种粗粒度的描述，用抽象的方式忽略细节，试图提取出
90:11
abstract way details try to to pull out Salient features so I remember what I was going to say before uh so this is an agent based model where you know each one of these is considered an agent agent and this emergent Behavior happens where you know this structure comes together and just moves around so I was I remember I was going to say if you think of your brain as an agent-based uh complex system in a sense which is really what it is neurons are just interacting with each other and outside stimulant uh so roughly speaking it could sort of be considered this um agent based system and the neurons are analogous to the balls here and the the formations and the actions of what's going on at a higher level that are emergent is analogous to your thoughts so when your thoughts try to introspect on well I don't know I I can't really go many much further of like a sea of to you can't get back you can't get to like the other Island so I mean you can't understand each of your neurons because you like
显著特征。所以我想起我刚才要说什么了，呃，这是一个基于智能体的模型，你知道，这里的每一个都被看作一个智能体，然后就出现了这种涌现行为，你知道，这个结构聚合到一起，就这么四处移动。所以我记得我要说的是，如果你把大脑想成一个基于智能体的、呃，某种意义上的复杂系统，而这其实就是它的本质——神经元只是彼此之间以及和外界刺激相互作用，呃，所以粗略地说，它大致可以被看作这种基于智能体的系统，神经元就类似于这里的这些小球，而在更高层次上发生的那些结构和行为，那些涌现出来的东西，就类似于你的思维。所以当你的思维试图内省时，呃，我不知道，我真的没法走得更远，就像一片海……你回不去，你到不了另一座岛，所以我的意思是，你没法理解你自己的每一个神经元，因为你就像……呃，我不知道，这个结构没法内省它
91:21
like uh I don't know this this Str structure can't like introspect on what it's made out of because then it wouldn't be itself anymore I yeah quantum mechanics again I don't know so one more example which is pretty cool which um which exhibits sort of optimization um I can click and add balls and they connect to each other and each one of them have has this each pair every pair has this sort of optimal distance away from one another and it sort of evolves to this this optimal distance so nowhere in my program do I say if you have three of three three of these form a triangle it's an emerging property due to these various local forces trying to achieve local Optimal Solutions and this is this happens in chemistry a lot you know particles and mo ules always try to find the lowest energy State and that's how that's how molecules exist but it's not as simple I mean so here here's another one you add four this is the optimal configuration and it just evolves into it and this is how physics works and with five it makes
自己是由什么构成的，因为那样它就不再是它自己了。我……对，又是量子力学，我不知道。那再举一个例子，挺酷的，呃，它展示了某种优化。呃，我可以点击添加小球，它们会相互连接，而每一个……每一对，每一对之间都有这样一个最优距离，它会演化到这个最优距离。所以在我的程序里我从来没有说过，如果有三个……三个……三个这样的就组成一个三角形，这是一种涌现属性，源于这些各种各样的局部力试图达成局部最优解。这在化学里经常发生，你知道，粒子和分子总是试图找到最低能量态，分子就是这样存在的。但没那么简单，我是说，这儿还有一个，你加到四个，这就是最优构型，它就自己演化成这样，物理就是这么运作的。到五个的时候它会形成
92:43
this star shape but um this this program is not as simple as I as I said it is there are some strange rules that say when and when not to make edges between them edges are connections between the balls so I can do something like this I can just drag and make a ton of things I I think I I don't remember the the rules exactly but I tried to make it so that it was only local and things that are far apart don't interact with one another they only go through intermediaries it's like in resonance like it's forming breaking bonds yeah so it's like resonance sort of resonance I don't remember exactly what resonance is but it's yeah resonance just when you're oscillating between kind of two superimpose two possible States right and I mean really that's what's going kind of what's going on here right is you have these different possible stable arrangements and just kind of this equil that diffusing back and forth between right so there are these two stable Arrangements that lead into one another so it just oscillates between
这种星形。不过，呃，这个程序并不像我说的那么简单，里面有一些奇怪的规则，规定什么时候该、什么时候不该在它们之间连边，边就是小球之间的连接。所以我可以做这样的事，我可以直接拖动，弄出一大堆东西。我……我想我……我记不清具体的规则了，但我当时试着让它只考虑局部，让相距很远的东西彼此不发生相互作用，它们只能通过中间体产生联系。——这就像共振一样，像是在形成和断开化学键。——对，这有点像共振，某种共振。我不太记得共振到底是什么了，但确实，呃——共振就是当你在两个……两个叠加的可能状态之间来回振荡的时候，对吧？——我是说，这里发生的其实差不多就是这么回事，对吧？你有这些不同的可能稳定构型，然后就是这种平衡……在它们之间来回扩散，对吧？所以这里有两个稳定构型，可以互相转化，于是它就在
93:51
them so I mean it's are roughly uh molecules in the world and stuff so I can I can if I can if I can make it all the way around I can make a cell I've done this before where you have this sort of um membrane which forms which is exactly like what happens in real biology oh yes there it is yeah more or less so in biology you have these hydrophilic and hydrophobic ends of lipids and the hydrophobic ones go on the inside and they all pair up with with another and the hydrophilic wait Hy hydrophobic hydrophilic ones go on the outside hydrophilic can interact with water and they're soluble with water but hydrophobic ones they're a fear you know fear of water so they try to go to one another and so you get this sort of structure I mean it's not not exactly analogous to this but it's sort of closed and uh it just sort of assumes this opposite configuration in ourselves and you know if it weren't for this sort of emergent this fundamentally lowlevel you know lower than the level of cells this is membranes you know this
它们之间振荡。所以我的意思是，这大致就是现实世界里的分子之类的东西。所以我可以……如果我能……如果我能一路围成一圈，我就能做出一个细胞。我以前做出来过，你会得到这样一种，呃，膜结构，这和真实生物学里发生的完全一样。哦对，出来了，对，差不多是这样。在生物学里，你有脂质的亲水端和疏水端，疏水的那一端朝内，它们两两配对，而亲水的……等等，疏……疏水，亲水的那端朝外，亲水基团能和水相互作用，能溶于水，而疏水基团则，你知道，怕水，所以它们就往彼此那边靠，于是你就得到这样一种结构。我是说，这跟这个并不是完全类比，但它也是某种封闭的，呃，它就自然形成了这种相反的构型。在我们体内，你知道，如果没有这种涌现，这种从根本上讲比细胞层次更低的，你知道，这就是膜，你知道，
95:05
without this emerging property of lipids we wouldn't exist you know it's a fundamental thing that holds us together all the time ourselves so I mean us as humans and every all biology is just built up of these layers and layers of emerging properties so I mean layers I mean like so these are cells and no no these are like lipids and one layer level above that is these these sort of stable spherical membranes and then a level above that is the interaction of cells with one another to form organs and then a level above that is the interactions of organs with with one another and the bloodstream and a level above that is you know the brain interacting with that you know in a positive way so we just built up all these crazy layers everywhere you know every everywhere in biology you'll find the sort of thing so it's emergence you know it's it's really cool to grock you know emergence so any any thoughts anybody questions or just about out of time what I what I see is that almost like every single little thing can like
如果没有脂质的这种涌现属性，我们就不会存在。你知道，这是把我们凝聚在一起的一个根本性的东西，无时无刻不在，我们自己也是。所以我是说，我们人类，以及所有生物，都是由这一层又一层的涌现属性搭建起来的。我说的层，我的意思是，这些是细胞……不不，这些是像脂质，比它高一层的是这些稳定的球形膜，再高一层是细胞之间相互作用形成器官，再高一层是器官彼此之间以及和血液循环的相互作用，再高一层是你知道，大脑与之相互作用，以一种良性的方式。所以我们就搭起了所有这些疯狂的层次，到处都是，你知道，在生物学里到处都能找到这种东西。所以这就是涌现，你知道，真正领会涌现是件很酷的事。那么，有什么想法吗，有人有问题吗？还是差不多该结束了？——我看到的是，几乎每一件小事都可以归结到最小的、最
16鲁棒性、关键节点与自我的连续性
96:18
be put down to like the smallest like basic physical like processes so the the only problem I think we have is like to sort of develop a threeory of organization like U like sort of how can I invariance of properties like to say you know even though like I have like almost like a thousand of like my brain cells have been like destroyed in like maybe like the past day or something I've been drinking or something you know I'm still me yeah what about my brand organization still says okay this is still a it's like it's still the same thing you know right I me robustness systems right yeah yeah that's that's that's totally what you're hitting at robustness robustness means you know you can change parts of it like for example with a wireless mesh Network or something it's robust because if you knock out you know a bunch of nodes here and there it still exists as a network and so what you're saying in your brain like even if you get completely wasted you still know that you're yourself so it's robust to these changes you know
基本的物理过程。所以我觉得我们唯一的问题是，怎么发展出一套关于组织的理论，呃，比如说，怎么描述属性的不变性，比方说，你知道，即便我大脑里差不多有上千个细胞已经被破坏了，比如可能是过去这一天里，我一直在喝酒什么的，你知道，我还是我。——是啊。——那我大脑的组织方式还是说，好，这仍然是……它仍然是同一个东西，你知道。——对，我是说，系统的鲁棒性，对吧？——对对，这完全就是你想说的，鲁棒性。鲁棒性的意思是，你知道，你可以改变它的一部分，比如说，就像无线网状网络什么的，它是鲁棒的，因为你就算敲掉，你知道，这儿那儿的一堆节点，它作为一个网络依然存在。所以你说的在你大脑里，就像即使你喝得烂醉，你依然知道你是你自己，所以它对这些变化是鲁棒的。——你知道，假设，好，一些脑细胞被破坏了，好，另一些新的被……被造出来，来确保我的
97:26
suppose okay some brain cells are destroyed okay other new ones are bu like made up to like make sure like my self concept is to preserve but between make those brain cells being destroyed and new ones being like created where where am I well okay so you're saying so say if some brain cells get destroyed and your concept of yourself is temporarily yeah temporarily is dissolved and then new new new Pathways or whatnot are formed that get your sense of self back so some time I don't even exist anymore so this is this is I mean hinting at a very important point that you know what are you if you know like it's it's you you are an emerging property your sense of self is not the actual neurons it's this thing which above exists on a higher level yeah but it has to be supported by the ne you know have to create that structure of yep it's a supporting structure but yeah you're saying so if you temporarily dissolve then yeah you just don't exist I mean I don't think there's anything more to it and you just
自我概念得以保留。但在那些脑细胞被破坏和新的被创造出来之间，我在哪儿呢？——嗯，好，你是说，就说如果一些脑细胞被破坏了，而你的自我概念暂时……——对，暂时消解了，然后新的……新的通路之类的形成，让你的自我感重新回来。所以有那么一段时间我甚至不存在了。——所以这……这其实是在暗示一个非常重要的点，你知道，你到底是什么，如果你……就是说，你……你是一种涌现属性，你的自我感并不是那些实际的神经元，而是这个存在于更高层次之上的东西。——是啊，但它必须由神经元来支撑，你知道，得先造出那个结构……——对，它是一个支撑结构，但，对，你是说，如果它暂时消解了，那你就是不存在了。我是说，我不觉得还有更多别的东西，然后你就重新出现——你就消失、再出现，而这种事一直都在发生，好吧，这
98:38
reappear you just disappear reappear and things do that all the time okay that could that could be the case if I was de you know I could Conant I could be knocked out and then I could be rebuilt and okay it be the same me but but is that really mean do I have to have continuity to be really there I even like you can't even have continuity because it has to take some time for the neur like to travel so you know even that the fact that you know I see a continuous SW maybe that's even an illusion you know that's signals have to travel right so I mean fundamentally right you have the cells in your body are replaced but every seven years right so you 7 years ago are not made of the same cells You're Made Of now right so there's just this idea that what constitutes you has to be stored at a higher level of description right that like you can take entire chunks out at a time and still think have things okay and what makes us different than a computer is that we can be running us right just like we would
这有可能是真的。如果我被……你知道，我可以被打晕，然后再被重建起来好吧，那还是同一个我，但这真的意味着……我一定要有连续性才算真的存在吗？我甚至觉得，你根本就不可能有连续性，因为神经信号传递是需要时间的，所以你知道，就连你看到的那种连续感，也许都是一种错觉，你知道，信号是需要传递时间的，对吧？所以我是说从根本上说对吧，你身体里的细胞每七年就会被替换一遍，对吧？所以七年前的你和现在的你并不是由同样的细胞构成的，对吧？所以就有这么一种想法：构成“你”的东西必须存储在更高层的描述层面上对吧，就是说你可以一次性拿掉一大块，你还是能思考、还是拥有那些东西。好吧，那我们和计算机的区别在于，我们能就像我们运行计算机那样运行我们，能够换掉几个晶体管，
99:50
like to run a computer and be able to know swap out a few transistors and maybe even like you know like a stick of memory right and still have everything running smoothly up top right because that's what happens all the time with us right but you can also kick out your memory you're not going to be you yeah so then there there does seem to be certain points of view which are really critical right that there are certain like key nodes that if you knock those out like and this is really kind of almost a problem in graph Theory and this is really something that we've used um in kind of understanding terrorist networks is like sure you can knock out a few lower guys but you're not going to destroy the whole network but if you knock out that guy is it the whole thing going to come to the ground right it's just like if you like you know get a pull to the head or something right like suddenly you could be you know Done Right whereas if someone cracks you in the leg right you're going to be okay
甚至可能换掉一条内存条，而上层的一切依然运转正常，因为这种事在我们身上一直在发生嘛。但你也可以把记忆拿掉，那你就不再是你了，对吧。所以看来确实存在某些真正关键的点，某些关键节点，如果你把那些敲掉——这其实几乎算是图论里的一个问题，而这也正是我们在理解恐怖分子网络时用到的东西：当然，你可以干掉几个下层的人，但你并不会摧毁整个网络；可如果你干掉那个人，是不是整个网络就垮了会倒在地上，对吧，就像……你要是被打中头之类的，对吧，突然之间你可能就完蛋了；而如果有人打你的腿，对吧，你还是没事的，
100:46
right so there yeah it's almost like a graph Theory problem fundamentally so a lot of the things that we're talking about are problems of complex systems and you you hinted at before you said like well so maybe what we should do is you know try to come up with a a general understanding of the kinds of systems and you know theories about them and this is complex systems so this is what like New England complex systems Institute does or Santa Fe Institute in California New Mexico Oh New Mexico yeah yeah sorry but yeah no I mean there's a lot of interesting stuff to go on and whatever interests you like I hope we can plug that interest um and please talk with us after class and we can you know kind of point you in the right direction um and we completely open we hope that's open ah you're attached to me I am um no columbic forces here but uh yeah so I mean please feel free to come talk to us after class and we can try to suit you up with your particular interests and even if you don't know what your interested in or
对吧。所以，是啊，这本质上几乎就是个图论问题。所以我们讨论的很多东西其实都是复杂系统的问题。你之前也暗示过，你说，也许我们该做的是试着对这类系统建立一个总体的理解，提出关于它们的理论，这就是复杂系统，这也就是新英格兰复杂系统研究所在做的事，或者加州的圣塔菲研究所——哦，是新墨西哥州，对，对，抱歉。但是啊，我是说，有很多有意思的东西可以继续深入，不管你对什么感兴趣，我都希望我们能接上你的这个兴趣点。课后请一定来找我们聊聊，我们可以给你指个方向。我们是完全开放的，希望这是开放的——啊，你粘到我身上了。我，嗯，这儿可没有库仑力，不过，是啊，所以我是说，欢迎课后来找我们聊，我们可以尽量帮你对接到你自己特定的兴趣上。就算你还不知道自己对什么感兴趣，或者你心里想的是「那部分真的很无聊，但这部分还行」，嗯，我们也许可以
17尾声：无穷的悖论与谁在观察
101:54
you have an idea of well that was really boring but this was okay um we can maybe kind of cater to your interests um but you know this was really kind of the last lecture and I wanted to thank all of you guys for coming here for bringing your questions for bringing your open minds um and you know I think we've had a really good semester so far and I I hope you guys enjoy Waking Life in the next lecture but uh other than that any last questions before I say goodbye in Bon Voyage buts that I've been struggling like for the whole day here you know okay so you have like Zer to one right and you have Zer to two right both of them like have an infinite amount of elements right but it's it could be proven that like as many points from 01 as to 0 to two yeah so why is two still good than one if there's as much stuff between both because we don't use as a well ordering operation the number of points between this and that right we just start off with kind of here are the integers here's a well ordering on them
照顾到你的兴趣。不过呢，这基本上算是最后一讲了，我想谢谢你们所有人来上这门课，带着你们的问题来，带着开放的心态来。我觉得我们这学期过得相当不错，也希望你们喜欢下一讲的《半梦半醒的人生》。除此之外，在我说再见、一路顺风之前，还有什么最后的问题吗？我这一整天都被这个问题困扰着，你知道的。好，比如说你有 0 到 1 这个区间，还有 0 到 2 这个区间，对吧，它们两个都有无穷多个元素，对吧，但是可以证明，0 到 1 里的点跟 0 到 2 里的点一样多，对，那既然两者之间的东西一样多，为什么 2 还是比 1 大呢？因为我们并不是用这个东西之间有多少个点来当作良序关系的依据对吧，我们一开始就是说，这是整数，这是它们上面的一个良序
103:06
and that's how we we we go so so just put like this traction it doesn't matter how much sta right I mean you know exactly I mean it's an uncountable amount of stuff in the unit interval and there's also an uncountable the same uncountable amount of stuff in the entire real line right these are the paradoxes of Infinities but they're not really paradoxes they're just because we're not used to thinking about Infinities convience so maybe it's just for convenience well I mean and partly but the other thing is like I mean mathematicians are pretty kind of retentive creatures and like we we make these not just for convention but we hope they're rigorously defined and um you can actually kind of create your real numbers just out of your rational numbers um by by using see this process of completion and there I mean you have to go do an undergraduate level course and Analysis really um but Justus just like draw like a square on like one ofare another I say this has a bigger than the other yeah but you're still kind of
我们是这么来的，所以就把这个当作一种抽象，不管里面有多少东西对吧 我的意思是，你知道的，没错，我是说单位区间里是不可数那么多东西，而整条实数轴上也是不可数、而且是同样多的不可数那么多东西，对吧，这些就是无穷的悖论，但它们其实不是悖论，只是因为我们不习惯思考无穷罢了 那也许这只是为了方便 我是说，一部分是这样，但另一方面是，我是说数学家是相当讲究、相当较真的一群人，我们定这些不只是为了约定俗成，我们希望它们是严格定义的，嗯其实你可以只用有理数就构造出实数来，嗯，通过所谓的完备化这个过程，那个嘛，我是说你得去上一门本科水平的分析课才行，嗯 但是就好比在纸上画两个正方形，一个比另一个大，我说这个比那个大，对 但你还是在绕开这个问题，因为正如
104:16
getting away at at the idea because just as you said right there just as many points and R2 right the two- dimensional plane it's r on the line right you can have space filling curves that visit everything which can be put into one to one correspondence with just the real line it's the same cardinality right so then why is two greater than one why is two greater than one it's because we start out that way basically right but then we gu sign that's conflicting no it's not conflicting it's just you can still if I give you if you give me a number and another number I can tell you which one's bigger than the other right it's not a problem but this okay how do you define Biggers the amount of stuff that's in it or like the what does that mean what ises like the amount of stuff in it mean right okay like there's there's as many points between Z and one and z and two right but see that's not a yes I mean that's a true statement but it's also true that there's you know as many odd numbers as there are natural numbers
你刚才说的，点一样多，对吧，R² 就是二维平面，跟实数轴上的 R 一样，对吧，你可以有填满空间的曲线，能访问到每一个点，能跟整条实数轴建立一一对应，是同样的基数，对吧，那为什么 2 还是大于 1 呢 为什么 2 大于 1？基本上是因为我们一开始就是这么规定的，对吧，但那我们不就得到一个矛盾了吗 不，这不矛盾，只是说，你还是可以——如果你给我一个数和另一个数，我能告诉你哪个更大，对吧，这没问题，但是 好，那你怎么定义「更大」？是里面东西的多少还是什么？「里面东西的多少」到底是什么意思，对吧？好，比如说 0 到 1 之间的点跟 0 到 2 之间的点一样多，对吧，但是你看，这不是一个——对，我是说这话是对的，但同样对的是，你知道，奇数跟自然数一样多，对吧，偶数跟
105:22
right there's as many even numbers as there are you know integers right but it's not a definite number it's an Infinity right just as you said it can be put into one1 correspondence with a subset of itself like that picture yeah but then I I just don't know how you can get from like things being that way and then you're saying okay but this is still better than that one well yeah you just don't use that as your metric for counting number of points because it's not a well-defined idea right I mean it's well defined but it doesn't give you any info right what it would tell you is that the reals are bigger than the integers right but that's not that's the you want to know that two is bigger than one right just so just on like this number line like put like a metric that's like almost the same as the integers on it I'm not sure but if you want to know more about metrics and things like we should talk about it after class um just real quick quickly I feel any other questions that people are burning deep down inside of
整数一样多，对吧，但这不是一个确定的数字，这是一个无穷，对吧，正如你说的，它可以跟自身的一个子集建立一一对应，就像那个图那样 对，但我就是不明白，你怎么能从「事情是这样」推到你说的「但这个还是比那个大」 是啊，你就是不要拿那个当作你数点的多少的标准，因为那不是一个良定义的概念 对吧，我是说它是良定义的，但它给不了你任何信息，对吧，它能告诉你的是实数比整数多，对吧，但那不是——你想知道的是 2 比 1 大，对吧 那就在这条数轴上，放一个跟整数上那个几乎一样的度量 我不确定，不过如果你想更多了解度量之类的东西，我们可以下课后再聊，嗯 就快速问一下，我看看还有没有人心里憋着别的问题
106:29
them all right well then yes let's call it a wrap and then uh um you know feel free to stick around and hang out after class even if you like explain like yourself as like almost this data thing mhm how you going to explain the Observer of that data yeah it's a tangle and even even in then you you have the Observer what the Observer just observe like being also put into the they're not different from each other they different or they like the same the same not different so that okay you can say the de is observing itself or it's recording things about itself because I mean like even in in quantum mechanics you need an observer like The Observer is just kind of really can be a measurement that just happens by the interaction of two particles so the particles themselves can be kind of observers of themselves right they kind of did I say that no Quantum I was just thinking Quantum yeah Quantum okay fair enough
想问的 好，那我们就到这里吧，然后呃，嗯，欢迎大家留下来一起聊聊下课以后也行 就算你把自己解释成几乎就是个数据之类的东西 嗯 那你要怎么解释那个观察者呢数据 是啊 它是纠缠在一起的 甚至 甚至那时候 你有观察者 观察者只是观察 就好像自己也被放进了它们彼此之间并没有区别 它们有区别还是它们像是一样的 一样的 没有区别 所以那样 好 你可以说这个 de 在观察它自己 或者说它在记录关于它自己的东西 因为我的意思是 就算在量子力学里你也需要一个观察者 观察者其实真的可以就是一次测量 而这次测量只是由两个粒子的相互作用产生的 所以粒子本身可以算是它们自己的观察者 对吧 它们某种程度上 我是那样说的吗 不 量子 我刚才只是在想量子 是啊 量子 好吧 有道理
视频总结 · 一句话概括与核心要点
一句话概括
这是 MIT《哥德尔、埃舍尔、巴赫》系列课程的最后一讲：先用几分钟回顾哥德尔不完备定理的证明骨架（哥德尔编码→算术化推理→"可证性=素性"同构→自指不动点 G），然后转向本讲主题"层次与涌现"——从晶体管到操作系统、从神经元到心灵，讨论描述层级、自指的极限，并用多个粒子/交通/分子模拟程序演示简单局部规则如何产生复杂全局行为。
核心要点
- 哥德尔证明的四步骨架可以压缩成一条同构链：把形式系统 F（此处为 TNT）的符号编码为数字（如 "=" 对应 555），字符串对应哥德尔数，公理与推理规则对应数的算术运算（乘以 10、中国剩余定理等），于是"某串是 F 的定理"精确等价于"某数具有 prim（可证性）性质"。这就是为什么不完备定理只适用于强到足以包含数论的系统。
- G 的构造依赖"引用（quining）"的不动点：类似"当前面加上自身的引号形式时产生完整句子"这一操作喂给自身得到自指句，把这一操作算术化后，其不动点便是"当喂入自身哥德尔数时产生非 prim 数"的公式 G。G 打破了"真 ⇔ 可证"的数学家信条：所有可证的都真，但并非所有真的都可证。
- 把 G 加为公理不能修补系统：学生提出"把 G 作为新公理生成新系统，反复迭代能否穷尽全部数学"，讲师答：哥德尔已证明新系统会产生新的 G'，仍不完备；这与停机问题同构——不存在能判定任意哥德尔数真假的"魔法机器"。这正是数学家永远不会失业的原因：创造新概念、新公理需要非机械的人类智能。
- 描述层次在多个领域呈现同构的"塔"：计算机方面是 晶体管/电容 → 主板/显卡 → 软件/操作系统；大脑方面是 神经元/神经递质 → 杏仁核/海马/皮层 → 心智/灵魂。这解释了为什么 AI 研究者出现在计算机系。课堂投票显示大多数学生对"各层之间的共性"最感兴趣，即复杂系统研究。
- 箭头只单向"可构建"，但影响是双向的（缠结层级）：有身体无心智可以（树不会跑 Windows），反之不行；但高层确实影响低层——SIM 卡与运营商、软件控制投影仪、"量子力学不是决定论的"这类知识改变你的底层心理模型。学生反驳"其实全是硬件在控制自己"，讲师承认软件本体上只是晶体管活动，但人类必须用高层描述来思考，这即侯世达的"缠结层级"。
- 自我模型无法覆盖自身的低层实现：借霍夫施塔特的思想实验——问"你的腿感觉如何"正常，问"你今天为何造这么少红细胞"却荒谬，说明"自我"只包含内部模型能触及的部分。讲师进一步提出，进化可能刻意没有给我们访问自身神经过程的能力（沉迷内省的"哲学家"会被猎豹吃掉），所以我们只得到决策的结果而非其"神经元级日志"。
- 理解自己类似给数论加 G：一名学生提出"理解了自己的一部分，那个做理解的部分又成了新的未知"，讲师指出这与 G 的迭代不完备形式上类似，但也警告这是对不完备定理的"危险用法"，理论上并不牢靠。另有学生引用类似 Penrose《心灵的影子》的论点：人能不断跳出系统做元思考，因而不是图灵机，而机器受限于逻辑；对此讲师提示 Penrose 的观点有争议，并给出 Minsky"心智社会"（一群小 agent 投票）作为替代模型。
- 模拟演示的统一信息：局部同质规则 → 全局涌现结构：库仑力 F=kQiQj/r² 的离散模拟出现晶体、液滴、链状"分子"（2D 和 3D）；N 体问题是确定性但不可预测（非线性混沌，只能运行才知道）；交通流中红灯造成的"回传波"是由车辆组成却反过来约束车辆的高层实体；带最优距离的节点自动排成三角形、五角星、以及类似磷脂双层的"细胞膜"——程序里从未写"形成三角形"。每一层（脂质→膜→细胞→器官→大脑）都是前一层的涌现。
- 蚁群与"女王神话"说明涌现无需中枢：每只蚂蚁只有寥寥神经元，蚁后并不下达指令，复杂群体行为完全来自简单个体的交互；类比人脑中神经元与"符号"的关系（引用《前奏与蚁赋格》），也类比社会的基于主体建模（agent-based modeling）——但因每个"主体"本身是不可预测的心智，社会预测更加困难。
- 宇宙只能被自身模拟：引 Seth Lloyd《为宇宙编程》：宇宙底层是量子力学，只有量子计算机能有效模拟，而宇宙自身就是量子计算机；要逐细节模拟宇宙需要同样大小的计算机，因此人类只能在高层做"块状"近似描述、抽取显著特征。
结论与值得注意的细节
- 课程的核心 take-home：存在系统内为真却不可证的命题，必须"跳出系统"；而这一讲把同样的思想延伸到心智研究——我们无法从内部完整内省自身，只能在更高层用近似描述工作。
- 讲师讲了一段轶事说明"未察觉的同构也能正常运作"：他发短信引用海因莱因《穿上太空服去旅行》，对方没有识别典故却仍觉得好笑；后来查到书的结尾主角获得 MIT 全额奖学金——第三层意义两人都不知道。类比：若哥德尔从未发现同构，PM 也许会"以为自己完备"继续前行。
- 穿插了马斯洛需求层次（生理→安全→归属→尊重→自我实现），指出这间教室里的人都在金字塔顶端担心"逻辑能否解释意识"，而街上很多人还在为下层需求奔忙；并借此引出下一讲电影《半梦半醒的人生》中"进化已转移到文化/模因层面"的观点。
- 牛顿被作为"模型反例"：他极不友善（任铸币厂长 32 年从不赦免剪币犯，剪币当时可判绞刑），却在 65 岁一晚解出最速降线问题——说明自我实现未必伴随"道德与爱"。
- 关于"自我"的鲁棒性：人体细胞约七年全部更换、醉酒损失神经元后"你仍是你"，说明自我存储在高层描述而非具体神经元；但存在关键节点（如头部致命伤 vs 腿伤），类似图论中恐怖网络"去掉核心节点整个网络崩溃"的问题。学生追问"旧神经元死亡与新通路形成之间，我在哪里？"——讲师坦言可能"暂时不存在，然后重新出现"。
- 课后一名学生纠结于 [0,1] 与 [0,2] 点数相同（同基数）为何 2 仍大于 1；讲师解释序关系来自整数的良序及实数由有理数完备化构造，"区间内点数"不是良定义的大小度量（它只能区分实数与整数的基数差异）。
- 讲师推荐：Roger Penrose《Shadows of the Mind》、Marvin Minsky《心智社会》、Seth Lloyd《Programming the Universe》；提到新英格兰复杂系统研究所与圣塔菲研究所（新墨西哥州）作为深入复杂系统的去处；并提醒本讲应读第 10 章"计算机系统的描述层次"，第 13、14 章有助于理解哥德尔证明，第 16 章较进阶。
核心句型 · 9
1. although all X are Y, not necessarily all Y are X
“Although all provable things are true not necessarily all true things are provable”
用于精确表达单向蕴含、否定逆命题。写论证时可用来纠正「A 即 B」的粗糙说法，仿写：Although all experts are experienced, not necessarily all experienced people are experts.
2. what if … hadn't …? would … have just …?
“What if he hadn't discovered this isomorphism … would have mathematics just marched on like fine”
反事实假设句，虚拟语气过去完成时。适合用来引出思想实验或强调某发现的关键性。
3. it all boils down to …
“But it all boils down to what's happening on this level of transistors and capacitors”
「归根结底是……」，口语中高频的总结性表达，用来把复杂现象压缩到一个根本原因。
4. you can have X without Y, but not Y without X
“You can have the physical things without the soul … these arrows only go One Direction”
表达非对称依赖关系的句式，适合讲清「前提」与「结果」谁依赖谁。
5. once you've got …, you can start …
“Once you've got a basic amount of tools available to you can kind of start building up”
「一旦有了……就可以……」，描述从基础构件到高阶能力的生成过程，常见于讲解学习曲线、系统搭建。
6. nowhere in X do I say …; it's an emerging property
“Nowhere in my program do I say if you have three of these form a triangle it's an emerging property”
否定词前置引起倒装（nowhere … do I），强调某结果并非显式设定而是自发产生，适合描述涌现或意外结果。
7. it's deterministic but not predictable
“It's deterministic not predictable it's chaotic because it becomes nonlinear”
用一对看似矛盾的形容词并列来精确区分概念，仿写：It's simple but not easy. / It's legal but not ethical.
8. so you're essentially arguing that …
“So you're essentially arguing that self- reference isn't like a well formed thought”
讨论中复述对方观点以确认理解的句式，既礼貌又能推动辩论；也可用 so what you're getting at is …
9. don't let that stop you from …
“You can never fully understand the brain don't let that stop you from trying”
承认限制后给出鼓励的转折式表达，先让步再激励，适合演讲结尾。
生词精讲 · 141 · 按出现顺序
denote
/dɪˈnoʊt/
v.
0:00
表示，记作（数学中用符号代表某物）
isomorphism
/ˌaɪsəˈmɔːrfɪzəm/
n.
1:09
同构；两个结构之间保持关系的一一对应
arithmetization
/əˌrɪθmətaɪˈzeɪʃən/
n.
1:09
算术化（把逻辑推理转化为数的运算）
symbol shunting
phr.
1:09
符号搬运，机械的符号操作（贬义/戏谑）
Encompass
/ɪnˈkʌmpəs/
v.
1:09
包含，涵盖
free variable
phr.
2:18
自由变量（逻辑学术语）
self-referential
/ˌselfˌrefəˈrenʃəl/
adj.
3:28
自我指涉的
fix point
phr.
3:28
不动点（fixed point：f(x)=x 的 x）
Credo
/ˈkriːdoʊ/
n.
4:33
信条，信念
take home message
phr.
4:33
核心要点，最该记住的结论
halting problem
phr.
4:33
停机问题（计算理论）
undecidable
/ˌʌndɪˈsaɪdəbəl/
adj.
5:34
不可判定的
shaking of the foundations
phr.
5:34
动摇根基
tack on
phr. v.
7:32
附加，额外加上
analogous
/əˈnæləɡəs/
adj.
7:32
类似的，可类比的
complement
/ˈkɑːmpləmənt/
n.
8:36
补集（数学）；补充物
digress
/daɪˈɡres/
v.
9:33
跑题，离题
begs the question
phr.
9:33
（此处口语用法）引出/令人不禁要问某问题
signpost
/ˈsaɪnpoʊst/
n./v.
11:28
路标；（讲话中）预告要点
hit home
phr.
11:28
击中要害，使人深刻理解
emergence
/ɪˈmɜːrdʒəns/
n.
11:28
涌现（复杂性从简单成分中产生）
reduced to an algorithm
phr.
13:51
归约为一个算法
prominent
/ˈprɑːmɪnənt/
adj.
14:56
著名的，杰出的
cooked up
phr. v.
14:56
（口语）编造、捣鼓出来
assembly code
phr.
16:01
汇编代码
double think
n.
16:01
双重思想（同时相信两个矛盾信念，源自奥威尔）
homunculus
/həˈmʌŋkjələs/
n.
17:00
侏儒，「脑中小人」（哲学中的谬误式解释）
abstracting
/æbˈstræktɪŋ/
v.
19:16
抽象化，把细节封装起来
lag
/læɡ/
n.
19:16
延迟，卡顿
hacks away
phr. v.
20:31
埋头敲代码/持续干活
boils down to
phr. v.
20:31
归结为
Realm
/relm/
n.
20:31
领域，界
neurotransmitters
/ˌnʊroʊˈtrænsmɪtərz/
n.
21:40
神经递质
amygdala
/əˈmɪɡdələ/
n.
21:40
杏仁核
hippocampus
/ˌhɪpəˈkæmpəs/
n.
21:40
海马体
transcription
/trænˈskrɪpʃən/
n.
21:40
转录（DNA→RNA）
psyche
/ˈsaɪki/
n.
22:56
心灵，精神
clunky
/ˈklʌŋki/
adj.
22:56
笨重的，不灵便的
souped up
phr.
22:56
（口语）经过改装提速的，加强版的
commonalities
/ˌkɑːməˈnælətiz/
n.
25:28
共性
modding out
phr. v.
25:28
改装（电脑/游戏）
derivability
/dɪˌraɪvəˈbɪləti/
n.
26:42
可推导性
power set
phr.
29:13
幂集
lofty
/ˈlɔːfti/
adj.
29:13
高层次的；崇高的（原文拼作 lofy）
permutations
/ˌpɜːrmjuˈteɪʃənz/
n.
29:13
排列
Merit
/ˈmerɪt/
n.
30:26
价值，优点
paranoid
/ˈpærənɔɪd/
n./adj.
31:22
偏执狂（的）
thought experiment
phr.
31:22
思想实验
Logical consequence
phr.
32:42
逻辑推论
internalizing
/ɪnˈtɜːrnəlaɪzɪŋ/
v.
33:54
内化
ascribes to
phr. v.
34:56
把……归于
Prelude
/ˈpreljuːd/
n.
36:06
前奏曲
deterministic
/dɪˌtɜːrmɪˈnɪstɪk/
adj.
36:06
决定论的
interplay
/ˈɪntərpleɪ/
n.
37:09
相互作用
self-reflective
/ˌselfrɪˈflektɪv/
adj.
39:09
自我反思的
assimilate
/əˈsɪməleɪt/
v.
39:09
吸收，同化
scrape away at
phr. v.
40:10
一点点刮开/剥离
forebrains
/ˈfɔːrbreɪnz/
n.
40:10
前脑
natural selection
phr.
41:17
自然选择
bottom of the pack
phr.
41:17
队伍末尾，垫底的
protrusion
/proʊˈtruːʒən/
n.
42:21
突出部分，投影
persecution
/ˌpɜːrsɪˈkjuːʃən/
n.
42:21
迫害
transcript
/ˈtrænskrɪpt/
n.
43:22
记录，文字稿
troubling
/ˈtrʌbəlɪŋ/
adj.
44:20
令人不安的
hierarchy of needs
phr.
44:20
需求层次（马斯洛）
Transcendence
/trænˈsendəns/
n.
45:31
超越
self-actualization
/ˌselfˌæktʃuələˈzeɪʃən/
n.
45:31
自我实现
physiological
/ˌfɪziəˈlɑːdʒɪkəl/
adj.
46:39
生理的
spontaneity
/ˌspɑːntəˈneɪəti/
n.
47:34
自发性
slippery slope
phr.
47:34
滑坡论证/滑坡效应
infringe on
phr. v.
47:34
侵犯（权利）
Tippy Peak
phr.
48:49
最尖端的顶点（口语）
brutish
/ˈbruːtɪʃ/
adj.
49:54
野蛮的，残酷的
memetic
/məˈmetɪk/
adj.
49:54
模因的，文化复制层面的
tangental
/tænˈdʒenʃəl/
adj.
49:54
离题的，扯远的（标准拼写 tangential）
schizophrenic
/ˌskɪtsəˈfrenɪk/
adj.
51:02
精神分裂的
clipping coins
phr.
52:17
剪币（削取硬币边缘贵金属）
punishable by hanging
phr.
52:17
可判绞刑
bun in the oven
idiom
53:15
（口语）怀孕
illiterate
/ɪˈlɪtərət/
adj.
53:15
不识字的
defects
/ˈdiːfekts/
n.
53:15
缺陷
circulated
/ˈsɜːrkjəleɪtɪd/
v.
54:31
流传，传阅
destined
/ˈdestɪnd/
adj.
56:00
注定的
reductionism
/rɪˈdʌkʃənɪzəm/
n.
57:13
还原论
linear combinations
phr.
57:13
线性组合
apprehension
/ˌæprɪˈhenʃən/
n.
58:24
忧惧，担忧
generalize
/ˈdʒenrəlaɪz/
v.
58:24
一概而论
platonic
/pləˈtɑːnɪk/
adj.
59:30
柏拉图式的（此处指抽象理念世界）
phenotypes
/ˈfiːnətaɪps/
n.
59:30
表型
bias
/ˈbaɪəs/
v.
60:41
使有偏见，影响
discrete
/dɪˈskriːt/
adj.
61:48
离散的
discretized
/dɪˈskriːtaɪzd/
v.
63:09
离散化的
repel
/rɪˈpel/
v.
63:09
排斥
crystallization
/ˌkrɪstələˈzeɪʃən/
n.
64:38
结晶
beaing around the bush
idiom
64:38
拐弯抹角（标准写法 beating around the bush）
discernable
/dɪˈsɜːrnəbəl/
adj.
65:56
可辨别的
glossing over
phr. v.
65:56
掩盖，一笔带过
category Theory
phr.
66:58
范畴论
droplet
/ˈdrɑːplət/
n.
68:07
液滴
chaotic
/keɪˈɑːtɪk/
adj.
69:17
混沌的
nonlinear
/ˌnɑːnˈlɪniər/
adj.
69:17
非线性的
agent-based modeling
phr.
69:17
基于主体的建模
flocking
/ˈflɑːkɪŋ/
n.
69:17
群集行为
infer
/ɪnˈfɜːr/
v.
70:20
推断
tangled hierarchy
phr.
71:33
缠结的层级（侯世达术语）
propagating
/ˈprɑːpəɡeɪtɪŋ/
v.
76:16
传播
holism
/ˈhoʊlɪzəm/
n.
76:16
整体论
traction
/ˈtrækʃən/
n.
77:25
抓地力，牵引力
epop phenomenon
phr.
77:25
副现象（epiphenomenon，附带产生而无因果作用的现象）
grapples with
phr. v.
80:08
努力应对，苦苦思索
frontal lobotomies
phr.
80:08
额叶切除术
well formed
adj.
81:22
构造良好的（逻辑术语）
rambling
/ˈræmblɪŋ/
adj.
81:22
漫无边际的
introspecting
/ˌɪntrəˈspektɪŋ/
v.
82:26
内省
transcend
/trænˈsend/
v.
82:26
超越
existentialists
/ˌeɡzɪˈstenʃəlɪsts/
n.
83:38
存在主义者
come to terms with
idiom
85:30
与……达成和解，接受
Salient features
phr.
89:04
显著特征
chunky
/ˈtʃʌŋki/
adj.
89:04
粗块的，粗粒度的
intermediaries
/ˌɪntərˈmiːdieriz/
n.
92:43
中间体，中介
resonance
/ˈrezənəns/
n.
92:43
共振；（化学）共振结构
superimpose
/ˌsuːpərɪmˈpoʊz/
v.
92:43
叠加
hydrophilic
/ˌhaɪdrəˈfɪlɪk/
adj.
93:51
亲水的
hydrophobic
/ˌhaɪdrəˈfoʊbɪk/
adj.
93:51
疏水的
soluble
/ˈsɑːljəbəl/
adj.
93:51
可溶的
grock
/ɡrɑːk/
v.
95:05
透彻领会（标准拼写 grok）
invariance
/ɪnˈveriəns/
n.
96:18
不变性
robustness
/roʊˈbʌstnəs/
n.
96:18
鲁棒性，稳健性
wasted
/ˈweɪstɪd/
adj.
96:18
（俚语）喝得烂醉
continuity
/ˌkɑːntɪˈnuːəti/
n.
98:38
连续性
swap out
phr. v.
99:50
换掉，替换
plug that interest
phr.
100:46
（口语）接上/满足那个兴趣点
Bon Voyage
/ˌbɑːn vɔɪˈɑːʒ/
phr.
101:54
一路顺风
well ordering
phr.
101:54
良序
uncountable
/ʌnˈkaʊntəbəl/
adj.
103:06
不可数的（无穷）
retentive
/rɪˈtentɪv/
adj.
103:06
（此处戏谑）较真的、拘谨的
rigorously
/ˈrɪɡərəsli/
adv.
103:06
严格地
space filling curves
phr.
104:16
空间填充曲线
cardinality
/ˌkɑːrdɪˈnæləti/
n.
104:16
基数（集合大小）
metric
/ˈmetrɪk/
n.
105:22
度量
call it a wrap
idiom
106:29
到此结束，收工
理解自测 · 11 题 · 是真懂了，还是以为自己懂
1. 讲者在开头回顾中，把「判断一个字符串是否是定理」转化成了什么问题？
转化为「判断一个数是否具有 primness（哥德尔可证性）」这一算术性质的问题。依据在第 0–2 段的回顾：先用哥德尔编号把 F 的符号映射到自然数，再把公理和推理规则「算术化」为对数的运算，于是可证明性就像「完全数」「素数」一样，成为一个带自由变量的数论性质。这一转换是哥德尔证明的核心，因为它让系统能以数的语言谈论自身的可证性。
2. Justin 用「有太空服就能走天下」的短信故事想说明什么？
想说明同构（意义层的对应）可以在参与者都不知晓的情况下照常运转。第 9–11 段：Justin 引用海因莱因小说名开玩笑，Curran 没听出典故却仍觉得好笑；后来 Justin 查资料发现书中主角拿到 MIT 全奖，这是双方都不知道的第三层含义。由此引出反事实问题：如果哥德尔没有发现 PM 能谈论自身的同构，数学是否会继续「以为自己完备」而照常前进。
3. Curran 演示的几个程序有什么共同点？他强调「涌现」的定义是什么？
共同点是：每个个体（粒子、小球、车）都服从相同的、局部的简单规则，程序里没有写任何关于整体形状的指令，但整体表现出晶体、液滴、分子链、交通波、三角形/星形、细胞膜等结构。第 56 段给出定义：局部的简单规则导致全局现象，即高层的东西从低层更简单的东西中涌现。第 80 段的「nowhere in my program do I say … form a triangle」是最直接的例证。
4. 马斯洛需求层次在本讲中被用来说明什么？学生用什么反例质疑它？
用来类比「层级」：低层需求（生理、安全）是高层（爱、尊重、自我实现）的前提，正如硬件是软件的前提，人类历史也是逐层向上爬。第 41–45 段还借此说明课堂里的人关心「启蒙」这个塔尖问题，而街上很多人仍在担心安全。反例在第 46–49 段：学生指出有些人在不安全、不被爱的状态下仍达顶峰，Justin 以牛顿为例——他偏执刻薄、在造币厂把剪币者一律送上绞架，却能一晚解出最速降线问题，说明模型「有缺陷」。
5. 为什么讲者说不完备定理保证了「数学家永远不会失业」？推理链是什么？
推理链：（1）哥德尔证明任何足够强的一致系统都有真但不可证的命题 G；（2）把 G 加为公理后，新系统又会产生新的 G'，无穷递归（第 7 段）；（3）因此不可能靠一台机器机械地生成全部数学（第 6 段）；（4）填补空缺需要创造新概念、新定义，探索旧系统未覆盖的推论，这不能机械化，「需要人类智慧」（第 8 段）。所以数学的不完备性反而是对人类创造力的结构性需求。
6. 「箭头只指向一个方向」是什么意思？这与后面「软件能否控制硬件」的争论有何关系？
第 24 段：层级图中的箭头表示「可以用低层搭建出高层」，方向只向上——有物理层可以没有心智（树上不跑 Windows），但没有物理层就没有心智。这是存在上的非对称依赖。而第 63–67 段的争论是因果方向：一位学生坚持「没有更高的东西在控制低层，一切只是电活动」；Curran 回应因果是双向的——软件来自人的心智并决定硬件做什么，晶体管坏了也会「控制」高层。Justin 的调和是：本体上软件确实「只是」晶体管层的事，但人类使用描述层次，把软件当独立实体谈论是合法且必要的。两者并不矛盾：依赖单向，因果描述可以双向。
7. 讲者对「为什么意识问题这么难」提出了怎样的演化解释？
第 36–39 段：演化选择的是解决问题的能力（协作狩猎、农耕），而非「回头审视自己」的能力——在群体末尾琢磨「我是什么」的人会被猎豹吃掉。因此大脑被设计成不给我们访问决策过程的神经层记录，只给结果，「打包塞进一个小老鼠脑袋里」。第 72 段 Curran 补充：我们「被编程为忽略这些问题」。这一解释把内省的困难归为演化上的功能性屏蔽，而不是单纯的复杂度问题。
8. 学生提出「谈论自己就是胡说八道」，他的论证是什么？Curran 如何回应？
第 69–71 段：学生把自我分为「被大脑描述的自我」与「觉知它的觉知」，并主张进行自我谈论的那部分心智处于最抽象的层次，它不了解底层细节，因此对自己的描述只是像在分形上东张西望、靠直觉猜测结构，不能算「构造良好」的思想。Curran 承认这是佛教也在苦苦思索的问题，自己「没开悟、没有答案」，同时把它和大脑的复杂性联系起来：系统复杂到某个程度就无法内省自己的神经元。
9. Curran 提出「每理解自己一部分就多出一个不理解的部分」，并类比哥德尔句。这个类比站得住吗？讲者自己怎么评价？
类比是：把 G 加入数论后，新系统又能被哥德尔化产生 G'；同样，你与自己「达成和解」的那部分，本身又成了新的未被理解的部分（第 75 段）。这在结构上有相似之处——都是自指导致的无限层级。但 Curran 自己立即说这是「对哥德尔不完备定理的危险用法，可能站不住脚」。严格来说，不完备定理只对形式系统成立，人的自我理解并非形式系统，也没有「一致性」和「可证」的精确定义，所以它只是启发性隐喻，不是证明。这种自我克制的态度本身值得学习。
10. 如果彭罗斯在场，他会如何反驳本讲的整体立场？讲者又可能怎样回应？
本讲的隐含立场是：心智是大脑的高层描述，与软件之于硬件同构（第 21 段），因此原则上可由基于主体的系统涌现。彭罗斯（第 14 段）会反驳：人类能看出 G 为真而形式系统不能，说明人类不是图灵机，机器永远受限于停机问题，AI 不可能。讲者可能的回应有三：（1）第 15–16 段暗示的「心智社会」——人脑是大量互相矛盾的子系统投票，未必是单一形式系统；（2）第 86 段的类比——我们「运行自己」就像运行计算机，换掉部件仍能运转；（3）人类「看出 G 为真」也只是在更大的系统里做推理，同样有自己的 G'。Justin 事先也「警告」学生彭罗斯的观点有争议。
11. 把本讲「自我是涌现属性、由高层描述存储」的观点，放到「上传意识/复制大脑」的情境中，还成立吗？会遇到什么问题？
表面上支持：第 85–86 段说自我不是具体神经元，细胞可换、构成「你」的东西存于高层描述，那么只要复制高层结构就能复制自我。但本讲也埋下了三个困难：（1）第 24 段的单向箭头——高层必须有低层支撑，复制品需要等价的物理实现，而第 58 段的「多重可实现性」是否适用于意识并无定论；（2）第 85 段学生的问题——支撑结构消解期间「我在哪」，Curran 的回答是「你暂时不存在，再出现」，这意味着复制与原本之间没有连续性，谁是「我」成了空问题；（3）第 87 段的「关键节点」——某些结构一旦缺失整体崩溃，复制的精度要求可能逼近第 78 段说的「需要和宇宙一样大的计算机」。所以该观点在原则上允许复制，但在「同一性」问题上给不出答案。
下载便签
手机：长按图片保存
← 上一期 · NO.051Frege, Russell, & Modern Logic - A. J. Ayer & Bryan Magee (1987)
下一期 · NO.053 →Douglas Hofstadter • The Deep Interconnectedness of Human Souls
苏菲拉底
THE SOPHIE LAB · ASK THE BEST MINDS THE BIG QUESTIONS
内容仅供学习 · thesophielab.com