Peter Greenwood 为 Quanta Magazine
介绍
无穷大的概念可能与数字本身一样古老,可以追溯到人们第一次意识到他们可以永远数下去的时候。但是,即使我们有无穷大的符号并且可以在随意的谈话中提到这个概念,无穷大仍然非常神秘,即使对数学家来说也是如此。在这一集中,Steven Strogatz 与康奈尔大学的数学家同事Justin Moore讨论了一个无穷大如何大于另一个无穷大(以及我们是否可以确定它们之间没有中间无穷大)。他们还讨论了物理学家和数学家如何以不同方式使用无穷大,以及无穷大对于数学基础的重要性。
在Apple Podcasts 、 Spotify 、 Google Podcasts 、 Stitcher 、 TuneIn或您最喜欢的播客应用程序上收听,或者您可以从Quanta 流式传输。
贾斯汀摩尔
Strogatz (01:35):是的,我很高兴能和你交谈。我应该说,也许为了全面披露,贾斯汀是我在康奈尔大学数学系的朋友和同事。好的,那么我们开始思考数学家思考无穷大的问题。实际上,也许在我们深入数学部分之前,让我们先谈谈现实世界,因为我们不会在那里待太久。现在,我是对的,你曾经接受过物理世界的训练吗?
摩尔(02:02):是的,我读本科的时候是物理学和数学双学位。我有点厌倦了物理。我开始喜欢物理,也对数学更感兴趣。然后不知何故,通过它的过程,我对数学和物理更感兴趣了。
Strogatz (02:18):好的。好吧,无限物理学呢?它甚至有意义吗?我们所知道的现实世界中有没有无限的东西?
Moore (02:26):你知道 Charles 和 Ray Eames 制作的视频《10 的幂》吗?基本上每 – 我认为是每 10 秒,你是 10 的次方。好吧,起初,我认为 10 的幂更大。你缩小。然后每隔 10 秒,你就会变小 10 的幂,你会从宇宙的最大尺度下降到最小尺度的亚原子粒子。你知道,这是在我想说的 70 年代末或 80 年代初制作的。而且我认为我们对某些事物的理解从那时起有了一点变化,但不是很大。但我的意思是,关键是,有大约 40 个 10 的幂将最小长度尺度和最大长度尺度分开,也许你可以慷慨地投入几个额外的 10 的幂,只是为了更好的衡量。但可以公平地说,在物理学中没有任何东西可以测量大于 10 100或 10 200或类似的东西。
(03:22) 也许我们认为事物是连续的——连续运动或其他——也许这只是一种幻觉。也许一切都是细粒度和有限的。但事实是,物理学家肯定通过想象事物是平滑和连续的,并且认为无限是有道理的,从而对我们生活的世界有了很多发现。当你进入他们还没有真正形式化事物的物理学部分时,数学家的很多问题都归结为物理学家以各种漫不经心的方式处理无穷大,并从无穷大中减去无穷大,也许不像数学家希望的那样对此负责。我不认为这真的是一个有争议的声明。我认为物理学家会——大多数物理学家可能会——我的意思是,好吧,也许你会知道得更多。但我相信大多数物理学家会说这是一个相当正确的说法。
Strogatz (04:20):所以,就你自己的个人故事而言——我保证我不会讲得太深,以免让你难堪——但是是什么吸引你走向无限?不知何故物理学对你来说太小了?或者你只是喜欢数学的严谨性,或者……?
摩尔(04:33):我的意思是,我认为在我对集合论特别感兴趣之前,我对数学整体感兴趣并远离物理学。具有讽刺意味的是,这是因为我——好吧,如果你上物理课,在某些时候,你最终会很快地和松散地学习数学。你要么同意,要么不同意。我是那些对此不满意的人之一。
Strogatz (04:56):嗯。我是一个很好的人,而且我还在这样做。你知道,我的意思是,那些事情并没有让我太担心,尽管我确实尊重那种关心——纯数学家所拥有的知识完整性,你知道,担心这些事情。
(05:11):好吧,假设我只是,我不知道,就像一个好奇的少年,我什至不知道无限是什么。你会说它是什么?我应该认为它是一个很大的数字吗?是某种符号吗?是财产吗?思考无限是什么的好方法是什么?
摩尔(05:26):是的,我的意思是,我想它是——它可以是直线末端的一个理想化点,好吗?它可以是一个正式的符号。你知道,你可以这样想……一个正式的符号,就像我们引入 -1 一样,对吧?我记得当我还是个小孩的时候,老师们不愿意讲清楚谈论负数是否安全。而且,对,事后看来这听起来很愚蠢,但在某种程度上,-1 是否存在于现实世界中?但是你可以正式地操纵它,你可以在某种程度上正式地操纵无穷大,但你可能必须表现出更多的关注。您还可以使用无穷大来量化某物的数量。这在那里打开了更多的门,因为你可以谈论有无限的集合,其中一些比其他的更大。
Strogatz (06:15):好的。好的。所以你提到了“集合”这个词,我们今天肯定会谈论很多关于集合的内容。我确实说过你的兴趣包括集合论。你想再说一下你所说的集合是什么意思吗?
摩尔(06:26):我想我……答案是肯定的和否定的。所以我认为凭直觉行事并把它看作是一个未定义的概念并凭直觉使用它是可以的。但它也被用作为数学提供基础的一种机制,当人们意识到我们需要有一些,为数学是什么做一些仔细的基础时。
Strogatz (06:49):嗯嗯。那很有意思。因为我——很喜欢,小时候,我们学会用手指数数,或者我们的父母可能会开始说话,然后他们可能会指着东西说,“1、2、3……”我们学会了声音——孩子们就像他们很小的时候那样,我知道,对吧?我的意思是,如果您自己或亲戚有小孩。所以有事情的那一面。而且我认为大多数人会认为数字是数学的基础。但是你是说,我想大多数数学家都会同意,有比数字更深的东西,就是集合的概念,对吧?
Moore (07:22):我认为“集合”的概念是作为一个基础概念出现的,因为它是如此基础和原始。如果你是,如果你想用某种东西作为数学的结构,你想从它的基本属性看起来非常原始的东西开始,然后从那里开始。然后这个想法是你然后使用集合来编码诸如计数之类的东西,以及诸如有理数和实数之类的东西,等等。然后从那里开始,各种其他更复杂的数学结构,比如流形,或者,或者其他什么。
Strogatz (07:57):所以我记得,在我曾经和孩子们一起看的芝麻街剧集中。那是在电影里;我想是的。有一个角色正在为满屋子饥饿的企鹅点鱼。他让企鹅们大声喊叫,然后他们说:“鱼,鱼,鱼,鱼,鱼,鱼。”然后服务员对着厨房喊道:“鱼,鱼,鱼,鱼,鱼。”然后其他人说,“不,你错了。”还有人说,“嗯,你为什么不直接说他们点了六条鱼呢?”但它指出,这种数字的想法是在鱼的这个集合之后出现的。然后另一个角色很惊讶地说,“它对火花塞有用吗?还有肉桂卷?”
摩尔(08:42):我的意思是,我也认为,如果你有兴趣尝试理解,你能证明这一点吗?或者你能证明吗?并且您正在尝试为如何证明事物或其他任何事物建立规则,您希望基本原则尽可能简单。因此,与其尝试写下算术如何运作的规则,不如先为更简单的事物写下更简单的规则,然后从这些更基本的构建块中构建算术。
Strogatz (09:08):好的。那么,这也让我想起了新数学,在 60 年代小时候,我们曾经学习过交集、维恩图和并集,对吧?那是他们在奥斯汀教授的集合论的开端;我记得是二年级还是三年级。我的父母不知道为什么。但我猜,是你们这种类型的数学家,或者其他认为孩子应该学习集合的人,既不是在他们学习算术之前,也不是在他们学习算术的同时。
摩尔(09:33):是的,大多数人在集合论中学习,我的意思是,现在无限集合实际上是如何工作的。因为我们对无限集的直觉不如我们对有限集的直觉。我认为这就是推动基金会发展的主要原因。部分原因是我们想写下,好吧,我们相当确定无限集合和一般集合的属性是什么,然后尝试从那里发展无限集合的真实性?
Strogatz (10:03):好的,那我们为什么不举几个例子呢?你能告诉我一些无限集合的例子吗?
摩尔(10:08):嗯,就像自然数一样。就像你说的——比如 1、2、3、4、5、6、7、8 等等——还有诸如有理数之类的东西。你知道,分数就像两个相互重叠的自然数,或者可能是一个负分数。但是还有像实数这样的东西,你知道,任何你可以用小数表示的东西,包括 pi 和e之类的东西。
Strogatz (10:28):嗯嗯。所以他们可以在小数点后有无限多位。
摩尔(10:32):是的,是的,无限多的数字。他们不必重复。
Strogatz (10:35):嗯嗯。形状、点或几何事物,而不仅仅是数字事物呢?
Moore (10:41):是的,你也可以谈论几何形状的集合。
Strogatz (10:45):好的,所以这是集合的一个很好的特征,我们可以用集合统一或者至少有一个共同的语言来讨论算术几何,对吧?我想如果我们正在上微积分课程的话,我们可以讨论一组函数。你知道,就像我们在微积分课程中的连续函数集一样。当然。是的。管他呢。所以是的,所以这为我们提供了数学所有不同部分的通用语言,对吗?而且——但就数学的整体历史而言,它作为数学基础是一个相对较新的想法,你说呢?
摩尔(11:16):是的,我的意思是,我……好吧,我们所知道的现代数学大约有 100 到 150 年的历史。但我通常把它联系起来——上世纪上半叶,实际上,我们开始看到我们今天所知道的数学的所有主要部分开始发展,并真正成为它们自己的独特学科。大约在同一时间,[伯特兰]罗素发现了他的悖论,这激发了对某种严格的数学基础的需求。
Strogatz (11:49):嗯,嗯。我们应该提到——是的。所以伯特兰·罗素,我们现在谈论的,通常更广为人知的身份是哲学家或和平主义者,但他是一位相当强大的数学家和逻辑学家,对逻辑作为数学的一部分感兴趣。
摩尔:是的,是的。
Strogatz (12:04):正如你所说,他是帮助集合论真正落地的人之一。甚至在他之前,就有这位绅士Georg Cantor ,我们将在 1800 年代后期在德国谈论他。
(12:17):好的,那么在数学中,比方说,数学家如何使用无穷大?你提到它有多大帮助。它在哪里使用?
摩尔(12:27):是的,所以,在微积分课上,它是进行某些计算的有用符号。谈论当输入变得非常大时函数的行为。你可以谈论无穷大的极限,或者当一个数字趋于零或无穷大或类似的东西时的数量比率。这是我提到的第一种意义上的无穷大概念,您将无穷大视为直线末端的理想化点。
(12:53) 但你也可以谈论它——你知道,你可以,你可以谈论计算某个集合或集合的元素数量,并跟踪它有多少个元素,或者,如果它有无限多个元素,试图区分不同大小的无穷大。我的意思是,每个人都理解——或假装理解——有限和无限之间的区别。我认为Cantor 的非凡发现是,对于无限集合,你可以进一步区分。您可以区分它是所谓的可数,然后是所谓的不可数。或者甚至只是在一般情况下,不可数基数比不同不可数基数之间的区别更高。
Strogatz (13:34):好吧,我们去那里吧。因为这是,这真的把我们带入了我们主题的核心。我想第一次听到“可数”这个词的普通人可能会认为它的字面意思是可数的,比如有 10 的东西。你知道,如果桌子上有 10 个火花塞,我可以数出它们——1、2、3 , 最多 10。但是你和其他数学家使用 countable 来表示与此略有不同的东西。
Moore (13:56):这只是意味着您可以为集合中的每个元素分配一个自然数,这样自然数就不会被使用两次。
Strogatz (13:56):所以有些东西可以是可数的和无限的。
摩尔(13:57):无限。所以自然数显然是可数的,因为它们在数自己。但可能不太明显的是,包括自然数负数在内的整数是可数的。
Strogatz (14:18):让我们谈谈这个。因此,如果一个人以前没有考虑过这一点,那就很有趣了。因为就像——所以你说过,你要考虑所有的数字,所有的正整数,所有的负整数和零。
摩尔(14:29):是的。
Strogatz (14:30):你可能做错了。就像如果你从零开始向右数,然后你数到 0、1、2、3,你就再也回不去负数了。那么你将无法计算所有整数。
摩尔(14:41):是的。
Strogatz :但是你应该怎么做呢?
摩尔:你可以做的是,你可以数数,你知道,0、1、-1,然后是 2、-2、3、-3、4、-4、5、-5。如果您以这种方式列出它们,那么您最终会列出所有内容。
Strogatz (14:55):美丽。所以这个曲折的论点,你在积极和消极之间来回跳跃,是一种很好的、有组织的、系统的方式来表明,如果你想到任何整数,最终它会出现在列表上。
摩尔:是的。是的。
Strogatz (15:07):太好了。好吧,所以整数是可数的。 Cantor 还发现了其他一些可数的东西——我不知道他是否感到惊讶,但我们中的很多人在第一次了解它时都会感到惊讶。像,像什么?
摩尔(15:21):是的,我认为两个令人惊讶的好例子是——首先是理性。所以两个整数的所有分数的集合都是可数的。当您考虑时,这实际上很容易看出来,因为您可以列出所有分母为 1 的分数——或者分子和分母的绝对值最多为 1。然后,最多 2,最多 3,最多 4 . 并且在每个阶段,只有有限多个分子和分母的数量级至少为n的分数。然后你可以用这种方式耗尽所有的理性。
Strogatz (15:55):就像,如果我选择数字 n 为 3,你是说我可以得到一个像 1/2 或 2/1,或 0/3 这样的数字,因为分子加分母相加最多 3 个?
摩尔(16:06):是的。另一个有点令人惊讶的是,如果你考虑用拉丁字母表或任何你喜欢的字母表可以写下的单词数。至多有可数个有限的单词,或来自这个字母表的有限符号串。如果你想所有的词或所有的句子,所有的文学作品,如果你愿意——
斯特罗加茨:哦。
摩尔(16:30):- 任何不仅现在存在而且可能在未来某个时间存在的事物。你知道,你把无限多的猴子放在打字机前,看看它们在有限的时间内能产生什么输出。这只是一个可数集。
Strogatz (16:44):哇。那么所有可能的书籍,比方说,用我们知道的所有可能语言的拉丁文?
Moore (16:50):用所有可能的语言。是的。我的意思是,如果你愿意,你可以有一个可数的字母表。那不会使任何东西变大。
Strogatz (16:56):如此可数似乎是一个非常大的无穷大。但是 –
摩尔(16:59):是的。首先令人惊讶的是,那些看起来比自然数大的集合实际上与自然数大小相同。他们是可数的。但还有另一个惊喜,那就是实数,即十进制数的集合,是不可数的。
Strogatz (17:13):所以有一个值得注意的点,你一直在提到可以有不可数的集合。我想,也许最简单的例子是:想象一条在两个方向上都延伸到无穷远的线。就像一条无限长的直线。我们称之为真正的线路。那是不可数的。
摩尔(17:32):对。如果你,如果你递给我一个列表,一个据称包含那条线上所有元素的列表,有一个叫做对角线参数的过程,它允许你产生一个在线上但不在你的列表上的新点。这就是康托尔的著名发现。
Strogatz (17:49):那是一个非常惊人的发现,我当时猜对吧?现在你可以突然谈论两个无限集并比较它们。
摩尔(17:58):是的,是的。可数和不可数之间的区别在数学中非常有用。基本上,可数集,你仍然可以谈论无限长度的总和。这是在标准课程结束时——第二学期微积分课程结束时教授的内容。然而不可数集合的求和意义不大,或者至少你必须以更微妙的方式定义它们。也就是说,更像是积分或类似的东西。
Strogatz (18:30):好的,现在我们有了可数的区别,比如整数——1、2、3、4、5——和不可数的,比如线上的点。还有另一个问题,我认为如果我们能花一些时间解决这个问题会很好。称为连续统假说。你能告诉我们那是什么吗?
摩尔(18:50):是的。所以 Cantor 想知道:是否有,是否有介于两者之间的东西?你可以——你知道,自然数位于实数中,而且自然数是可数的。实数是不可数的,比自然数大。是否有一组实数大于自然数,但小于-
Strogatz (19:10):在这种计数意义上更小。
摩尔(19:12):——比线小?在那条线上,在数轴上,是否有一组点大于自然数,大于有理数,但小于整条线本身?不存在这样的中间集的断言称为连续统假设。这就是希尔伯特的第一个问题,即连续统假设是真命题还是假命题。
Strogatz (19:35):嗯嗯,希尔伯特是这方面的伟大数学家——也许晚了一点,但也不晚了。在这一年——大概是 1900 年左右,我想——他宣布或列出了他认为是未来 20 世纪数学家需要研究的一些最大问题。我认为这是他清单上的第一个问题?
摩尔(19:58):是的,这是第一个问题。
Strogatz (20:00):哇。所以考虑这个很重要。你说康托尔称之为假设。他以为会变成真的。
摩尔:是的。
Strogatz (20:07):他已经知道这两者之间没有无限夹心
摩尔(20:11):是的。事实是,它经受住了寻找反例的考验。我的意思是,如果你开始查看所有的实数集,你可以写下描述或可以通过某种方式构造的线的子集。他试过这个。他证明了,我的意思是,好吧,他证明了没有反例。早期甚至有定理说这种或那种类型的集合不可能是反例。
Strogatz (20:40):太棒了。让我确保我明白了。我从未听过这样的说法:从某种意义上说,仅仅因为其中一些是可以描述的这一事实就使它们不够好。
Moore (20:49):例如,一个封闭的集合有它的所有极限点。 Cantor 证明这不可能是反例。它要么是可数的,要么与实数具有相同的大小。
Strogatz (21:00):所以如果有反例的话,那肯定是不可描述的。
摩尔(21:04):是的,它必须很复杂。
Strogatz (21:06):哇。但当然,有可能存在,只是它会是一些非常奇怪的事情。
摩尔(21:12):是的。所以这让我们回到了这个基本问题上。你知道,大约在那个时候,他们开始尝试将数学公理形式化。后来的某个时候,大约在 1930 年代,[库尔特] 哥德尔证明,实际上任何一种你可能拥有的可理解的公理系统,只要达到对自然数进行形式化算术的适度目标,都必然是不完整的。有些陈述你无法从这个公理系统中证明,你也无法使用标准的有限证明从公理中反驳它们。
(21:52) 我认为这非常令人震惊。因为它告诉你,在某种意义上,以某种算法试图解决你所有的数学问题并产生某种算法基础的目标,某种完整的数学基础是注定的。或者至少必须受一些更高的直觉的支配,而不仅仅是——我不知道——当时可用的东西。
(22:16) 哥德尔证明了什么——他后来证明的一件事是,你无法证明或反驳的陈述之一是你的公理系统首先是一致的。它不会导致任何矛盾。该陈述可以编码为某种关于数论的陈述,关于自然数的算术,但不是以一种特别自然的方式。如果你去和系里的一位数论家交谈,他们不会认为这是一个问题或数论的陈述,尽管从技术上讲是这样。所以它是——哥德尔时代遗留下来的一个问题是连续统假设——或者是否存在其他一些自然数学陈述,它基于我们正在研究的公理系统是不可判定的。
Strogatz (23:02):所以有公理的概念。我们或许应该试着记住它们的样子。因为如果我们做非常仔细的数学计算,我们必须制定一些定义,还有一些我们接受的东西——我不知道为什么我不想说“我们认为是理所当然的”,但我们接受作为基岩。
摩尔(23:19):是的,是的。所以这是,我的意思是,这是希腊人所做的事情,也就是说,你知道——形式化几何的成就之一——不是试图定义几何是什么,而是将其视为:你是打算写下一些未定义的术语,然后写下支配这些未定义术语行为方式的规则或公理。对他们来说,这就像一个点和一条线。当一个点在一条线上时,这些都是未定义的概念。当一个点位于一条线上的其他两个点之间时,这些都是未定义的概念。然后你写下一组支配这些概念如何运作的公理。如果你做对了,那么每个人都会同意这些属性显然对这些东西是正确的。因此,这些公理是不证自明的真理。
(23:19) 所以对于几何学,你知道,有一个著名的平行假设,你无法从其他假设中推导出来。当发现您实际上可以构建满足所有公理但不满足平行假设的几何模型时,这在某种程度上是革命性的。因此,平行假设不能从其他公理证明。所以从某种意义上说,哥德尔所做的是开发一种方法来做到这一点,但是是在数学模型的层次上,或者至少是我们拥有的数学公理系统的模型。
Strogatz (24:45):啊哈,这么说很有趣。所以,就像我们有欧几里德几何,然后我们也有这些更新奇的非欧几何,爱因斯坦在广义相对论中使用了这些几何,但它们也被用于其他地方。它们在逻辑上与欧几里德几何一样好。但是现在不只是谈论几何,你说它有点像我们可以拥有传统的——好吧,我不确定这些词是什么。什么是欧几里德几何的类比?有传统数学吗?
摩尔(25:16):这是一个悬而未决的问题。我的意思是,我的意思是——我认为这在一定程度上是一个哲学问题。也许这是一个社会学问题,因为这是一个关于什么是数学的问题,对吧?回到那个基本问题。而且我认为我们拥有 100 多年前开发的 ZFC 公理的公理是我们普遍认为这些是真的公理,或者这些公理是“集合”应该具有的属性,但它们’不完整。
Strogatz (25:44):好吧,等等,让我们打开所有这些。听起来很好。那么 ZFC,我们为什么不从它开始呢?那是一些人和事物的名字。
摩尔(25:51):是的,是的。 “ Zermelo-Fraenkel 集合论”,其中包含一种叫做“选择公理”的东西。是的。
Strogatz (25:55):好的。因此,这些是被广泛接受的游戏规则。
摩尔(25:59):是的,这是一个公理列表——它相当长,但没那么长。比如,如果你有两个集合,那么有一个集合将它们都作为它们的元素。配对公理,你可以取一组集合的并集,这就是一个集合。等等。
Strogatz (26:15):好的。所以有 ZFC 方法来做集合论,你说,在某个时间提出,人们喜欢它,但后来你说它不完整?
摩尔(26:26):是的。所以这是你可以写的东西。列出公理的计算机算法。这是一组无限的公理。但是除了两种公理簇之外,它是有限的。如果您不注意,您实际上会认为这些其他公理簇中的每一个都是单个公理。但它们实际上是一个无限的公理家族。你可以生成一个计算机程序来吐出所有的公理。我们倾向于认为 ZFC 是一致的,因为我们没有发现任何矛盾。如果您相信这一点,那么根据哥德尔不完备性定理,ZFC 将无法证明它是一致的。
(27:03) 所以有一些陈述,例如 ZFC 的一致性,ZFC 无法证明。这是一个有趣的观点。因为我们再次相信 ZFC 是一致的。这就是,我的意思是,我的意思是……大多数数学家,他们将要工作的原因之一是基于 CFC 是一致的信念。正确的?但我们认为这是真实的陈述。但这不是ZFC本身就足以证明的东西。
Strogatz (27:27):我只是在想。一路上,我们一直在提到哥德尔。我不知道我们说过他是谁。你想简单地告诉我们吗?
摩尔(27:34) 是的,他是。我的意思是,他是一位革命性的逻辑学家。这一点,不完备性定理是他的主要成就之一。他的另一项主要成就是表明不能使用 ZFC 公理来推翻连续统假设。
Strogatz (27:49):有些人认为他是自亚里士多德以来最伟大的逻辑学家。爱因斯坦是他在高等研究院的朋友和同事,他说他喜欢有幸步行去与库尔特·哥德尔一起工作。我的意思是,他与爱因斯坦处于同一个知识联盟。如果你还没有听说过他,我建议你看一本关于他的书,叫做Journey to the Edge of Reason 。一本关于哥德尔生平的好书。但是好吧,他是,对,所以他是 20 世纪中叶、20 世纪初的逻辑学家。你说他证明了这一点——好吧,再说一遍关于连续统假设?
Moore (28:23):在任何集合论模型中,他构建了一个较小的集合论模型,它满足连续统假设。所以这表明你不能在集合论的公理内反驳连续统假设。从一个集合论模型,如果你有一个,那么我可以产生一个新的,它满足连续统假设。
Strogatz (28:43):我明白了。所以可能会有集合论的版本,有点小的版本,但仍然足以做算术,我认为。
摩尔:是的。
Strogatz (28:51):但是其中,好吧,连续统假设是正确的,就像 Cantor 猜测的那样。
摩尔:是的。
Strogatz (28:56):然后。但是——这个故事有一个很大的“但是”。
摩尔(28:59):是的。许多年后, [保罗]科恩开发了一种称为强制的技术,使他能够扩大集合论模型。用这个,他证明了你无法证明连续统假设。除了他的技术还可以用来证明你无法反驳。这个,是的,这种叫做强制的技术真的非常强大。强制和在集合论模型中构建较小模型的技术。这些是我们从旧的集合论模型中构建新的集合论模型的两种工具。
Moore (29:32): Going back to the geometry analogy. I mean, even these models of the hyperbolic plane, which were the non-Euclidean models of geometry — those themselves start by taking the Euclidean plane or a subset of it and building the model of geometry like the points and lines there. The points are just ordinary points on this disk. And the lines there are circles in, certain circles in the original geometry. The point that I’m trying to make is that this is a kind of a fruitful thing you do in mathematics. You oftentimes start with some structure that satisfies your axiom system, like a geometry that’s satisfying your axioms of geometry, and you manipulate it somehow and produce a new thing, which maybe satisfies a different set of axioms. That’s what Cohen and Gödel were doing, was that they were taking a model of the axioms of set theory — and therefore, in some sense, a model of mathematics — and manipulating it using various techniques to produce new models, which satisfied either that the continuum hypothesis is true, or that the continuum hypothesis is false.
Strogatz (30:36): So this is really amazing to me, and I’m sure to many people that, you know… Like, Plato has this philosophy that, that there are certain ideal forms out there and truths that — maybe we can’t see them here on Earth, but in some Platonic realm, their truth exists.
Moore : Yeah, yeah.
Strogatz (30:57): And you would feel like the real numbers exist, whether human beings think about them or not, and that the continuum hypothesis is either true of the real numbers, or it’s not. But you’re telling me?
Moore (31:09): Well, I mean, yeah, there are different schools of thought on this. I mean, you couldn’t — you can view it as, there’s this thing that I think goes under the name, that generic multiverse view, that there is nothing more that you can say. There are just all of these models of set theory. And the best that we can do is try to understand what’s true in each of them and move around between them. And that’s a very non-Platonic view of things, a kind of formalist view of things. You might also take the viewpoint that there is some maybe preferred model of set theory. That is, you know, the reality that we live in, and all of these other models, they’re models of the axioms, but they’re not really what we’re trying to describe with the axioms. I think the analogy with geometry is somewhat illustrative there, right? I mean, you can produce many different models of geometry. But we still live in a physical world that has a geometry and maybe that’s the, the geometry that we most care about.
Strogatz (32:03): I see. So in the same way that we could give Euclidean geometry some preferred status because it’s the one we’re used to. It’s the one which has been around long, because it’s sort of the easiest and most obvious, but we still think these others are good, and they have their domains where they’re useful and interesting.
Moore (32:20): But maybe the thing that’s worth pointing out there, too, is that even our understanding of — Well, first, I’m not sure that we live in a Euclidean geometry. But there’s, there’s a question about that. But even our understanding of the physical world is greatly enriched by understanding all of these other geometries, this free exploration of other models of geometry. And the same is true with set theory. I think, even if in the future, we settled on some consensus as to what is a new axiom for set theory, arriving at that destination is something that surely will not have been possible without all of this exploration that occurs beforehand.
Strogatz (33:00): What would proving or disproving the continuum hypothesis mean? For each of these camps? What’s at stake?
Moore (33:08): Yeah, that’s — OK, so I think the camp that takes this sort of “all worlds” viewpoint just would say that this is a meaningless question. That Cohen and Gödel and their techniques for building lots of models of set theory is kind of the end of the discussion. And you know, we’re going to produce lots of new models of set theory, maybe, but we’re never going to have a final answer for saying that the continuum hypothesis is true or false. The people that take the viewpoint that there is some sort of truth or falsity to that statement, would presumably try to come up with some new axiom and presumably some heuristic justification for why this axiom should be true — either a heuristic or maybe a pragmatic justification for why it’s true. And then once you argue that this axiom should be accepted, that it somehow encapsulates some intuition we have about mathematics or sets, then if this axiom also proves or disproves the continuum hypothesis in a sort of formal sense of the word, then you would view that CH is true or false.
Strogatz (34:12): So that’s sort of where we are now. That there really are these two camps at the moment.
Moore (34:16): Yeah, to a degree. It’s been so long since the continuum hypothesis was shown to be undecidable based on the axioms, that I think most mathematicians have kind of gotten used to the fact that maybe that’s the most that you can say. And I think it would be amazing at this point if mathematicians as a whole could rally around some new heuristic that, you know, everyone could agree ought to be true. And maybe that will never happen. Maybe, maybe the community has too many different viewpoints in it. To be fair, I think it — I think it’s somewhat of a consensus view, but not a universal view, that ZFC is the set of true axioms for mathematics. There are certainly people that take the view that anything infinite just doesn’t exist. And it doesn’t make any sense to talk about and we shouldn’t be talking about it.
Strogatz (35:05): Well, that’s a time-honored tradition. I mean, that’s — Aristotle was telling us to watch out about infinity. And throughout the history of math, people even as great as [Carl Friedrich] Gauss were very careful about this concept of completed infinity, which is what Cantor opened up this can of worms for us. But I don’t know that it’s worms. It seems like it’s — you know, what’s the harm? It’s that we’re letting our imaginations go and discovering a lot of interesting things.
(35:30) But I do have a question. As someone who’s not a set theorist, I don’t want to ask it in an impolite way. But it might come out sounding a little impolite, which — you know where I’m going, right? Like, how does this affect me? Does the rest of math feel the vibrations that are happening within set theory? Or are we sort of insulated from what you guys are doing?
Moore (35:49): That’s a good question. I think most mathematicians never encounter a statement which is neither provable nor refutable within the usual axiom system for mathematics within ZFC. And the set theorists have to a degree discovered an explanation for that. There’s a model of set theory which is larger than Gödel’s original model but smaller than the universe of all sets called the solid base model, that [Robert] Solovay discovered around the time of Cohen’s work. And the remarkable discovery is that this model — what’s true in it can’t be influenced by forcing. And therefore, essentially, if you can phrase something about what’s true in that model or false in that model, it’s something which is largely immune to independence phenomenon.
(36:35) The catch is that this model of set theory is not — does not satisfy the axiom of choice. So the axiom of choice is — this is another can of worms here. But one of the reasons why the axiom of choice is different from the other axioms is that it’s not constructive. All of the other axioms tell you that some set that you have a description of is, in fact, a set. That’s just how the axioms work. But the axiom of choice tells you that given a collection of sets that are non-empty, you can select something from each one of them — hence choice — but it doesn’t tell you how you’re going to make the selection. This was an axiom that, on the one hand, allowed us to construct all kinds of weird, paradoxical things. You know, I guess, in the ballpark of 100 years ago or so, like non-measurable sets, whatever that is. There’s this famous decomposition of the sphere, that Banach-Tarski paradox , that —
Strogatz (37:29): Oh, this is interesting.
Moore (37:32): — you could cut the sphere into finitely many pieces, and then reassemble them into two spheres that are the same dimensions of the original sphere. And now the reason why that’s absurd is that you ought to be able to assign a mass to each of the — you know, to the original sphere, and then assign a mass to all these pieces that you can cut it up into, and those ought to add up to the original mass. And then when you rearrange them, that process shouldn’t change the mass. But somehow, when you reassemble them, you have twice the mass that you started out with. Now, the point in that argument — where things go wrong is this cutting up of the sphere that the axiom of choice allows you to do is so bad that you can’t assign masses to these pieces that you have.
(38:11) Now, that paradoxical behavior led people to think that the axiom of choice is somehow perhaps problematic. Maybe it’s, it’s gonna lead to some sort of paradox within mathematics itself. And therefore, the axiom of choice shouldn’t be accepted. One of the things that Gödel proved at the same time as he proved that you can’t disprove the continuum hypothesis, is that it’s also safe to assume the axiom of choice. That is, if the axioms of ZFC without the axiom of choice are consistent, then so too is the set of axioms of ZFC with the axiom of choice. It gives you a lot of weird, exotic things, maybe, but from a foundational point of view, it doesn’t pollute the water.
(38:51) Sometime later, there was the discovery of this thing called Zorn’s lemma, which turned out to be equivalent to the axiom of choice. And it’s really very fruitful for developing a lot of different branches of mathematics. It’s something that — you learn about it if you’re an advanced undergraduate, or if you’re a graduate student in math. It’s somehow part of just the required learning for a graduate degree in math. And because of this extreme utility, it’s something that we just accept these days. I think most mathematicians are not comfortable working without the axiom of choice, just because in many cases they might be using it without even knowing it.
(39:31) So I think this is also an instance of how we might settle the continuum hypothesis. It’s that we discover some axiom in the future which is so useful in developing mathematics further, that we just regard this axiom as being true to a degree. That’s what happened with Zorn’s lemma. And with the axiom of choice, it wasn’t something that was initially viewed as true. In fact, it was sort of initially viewed with some skepticism.
Strogatz (39:56): But let me see if I can, since it does… We’ve been talking now a lot about the axiom of choice: Its relation to the continuum hypothesis. Is there a pithy way to say what that is?
Moore (40:06): You know, the axiom of choice and the continuum hypothesis have kind of a curious relationship because they… OK, the continuum hypothesis, from a set theorist’s point of view, it allows you to construct a lot of exotic things. It allows you to do an infinitely long, even uncountably long construction, where you’re doing everything in a very controlled way, an algorithmic way. And building some weird object where you’ve maintained a lot of control along the way. In the absence of the axiom of choice, the continuum hypothesis, as I stated it originally, that there is no set of rules which is intermediate, that’s something which doesn’t have the same bite as if the axiom of choice is true. And the reason for that is that, for instance, in the absence of the axiom of choice, you can talk about even stronger versions of the continuum hypothesis. Like, every subset of this number line, the real number line, is either countable, or there’s a copy of the Cantor set that lives inside of it. Like, there’s kind of a tree of points, a binary tree of points that sits inside of your set. And this is a very concrete way of saying it has the same size as the real numbers.
Strogatz (41:14): So for the rest of us in math outside of set theory, should we be losing any sleep over the — what seems to be — kind of indeterminate status at the moment of the continuum hypothesis? We’re told it’s undecidable in the standard model of set theory. You know, does it matter? Does it affect the rest of math?
Moore (41:35): The answer mostly is no. But it’s not entirely known. The continuum hypothesis. It’s true in the Solovay model , for instance: Every set of reals is either countable or there is a closed set of reals inside it which is uncountable and has no isolated points. But there are statements that show up in mathematics, questions that show up naturally, kind of organically in other fields, where it turns out that they are dependent on either the continuum hypothesis or something else, which is independent of the axioms of ZFC. One example of this is something called a medial limit, which is a device that is useful in probability and some parts of probability for taking limits of things and still maintaining that things are measurable. Medial limits are something that you can construct using the continuum hypothesis, but they’re not something that you can build in ZFC.
Strogatz (42:27): This makes me happy, I have to say. I mean, I want to believe that math is one big web. And that, like there’s an old saying, “No man is an island,” from whoever, I don’t know. But anyway, I don’t want any part of math to be an island. So I would hate to think that set theory is somehow some — I mean, no one would say it is, but even the part that contains the continuum hypothesis, I don’t want that to be divorced from the great continent. And it sounds like it’s not.
Moore (42:52): Right. If you take a Hilbert space, and you look at the bounded operators, and the compact operators, these are well-studied algebras of objects that are studied in mathematics. You can take a quotient of them. Studying what’s called the automorphism group of that is something that a mathematician might ask about. And indeed, Brown, Douglas and Fillmore asked about that in the 1970s. And it’s known that whether the continuum hypothesis is true or false is related to whether there are very complicated automorphisms of that algebra or not. That’s something that is, you know, a standard object in a functional analysis course that you would teach at the graduate level. And these are sort of very, very basic properties of this object.
(43:34) But the point is, this is something that’s, on the face of it — this is not a problem in set theory. Different set theorists have different takes on why the subject is important. But to me, this is why the subject is — what it’s important for. It’s that it plays this unique role of being able to let you know when you’re asking the question that might not be decidable, based on the axioms. Because you don’t want to be studying this problem that you can’t decide without any success for years and years and years. And if someone can tell you that, “Well, you’re never going to actually come up with a solution to that problem, because you can neither prove nor refute that,” right? That’s a good thing to know.
Strogatz (44:13): All right. Well, to me this a very uplifting message you’re giving, Justin, that — John Donne! That’s the name I was looking for, John Donne. And let’s say this in the modern way: No person is an island. And the same with no part of mathematics. There is — even the most esoteric seeming things on the outer reaches of set theory are still linked into very down-to-earth parts of math, in probability, in the functional analysis that underlies quantum theory. So, this is news to me, and I just want to thank you for enlightening us. This was fun. Thanks.
Moore (44:46): Thanks for having me.
Announcer (44:46): Explore more math mysteries in the Quanta book The Prime Number Conspiracy , published by The MIT Press, available now at Amazon.com , Barnesandnoble.com , or your local bookstore. Also, make sure to tell your friends about this podcast and give us a positive review or follow where you listen. It helps people find The Joy of Why .
Strogatz (45:12): The Joy of Why is a podcast from Quanta Magazine , an editorially independent publication supported by the Simons Foundation. Funding decisions by the Simons Foundation have no influence on the selection of topics, guests or other editorial decisions in this podcast or in Quanta Magazine . The Joy of Why is produced by Susan Valot and Polly Stryker. Our editors are John Rennie and Thomas Lin, with supported by Matt Carlstrom, Annie Melcher and Zach Savitsky. Our theme music was composed by Richie Johnson, Julian Lin came up with the podcast name. The episode art is by Peter Greenwood and our logo is by Jaki King. Special thanks to Burt Odom-Reed at the Cornell Broadcast Studios. I’m your host Steve Strogatz. If you have any questions or comments for us, please email us at [email protected] Thanks for listening.
原文: https://www.quantamagazine.org/how-can-some-infinities-be-bigger-than-others-20230419/