When I was in fourth grade, my teacher said to us one day: "There are as many even numbers as there are numbers." "Really?", I thought. Well, yeah, there are infinitely many of both, so I suppose there are the same number of them. But even numbers are only part of the whole numbers, all the odd numbers are left over, so there's got to be more whole numbers than even numbers, right? To see what my teacher was getting at, let's first think about what it means for two sets to be the same size. What do I mean when I say I have the same number of fingers on my right hand as I do on left hand? Of course, I have five fingers on each, but it's actually simpler than that. I don't have to count, I only need to see that I can match them up, one to one. In fact, we think that some ancient people who spoke languages that didn't have words for numbers greater than three used this sort of magic. For instance, if you let your sheep out of a pen to graze, you can keep track of how many went out by setting aside a stone for each one, and putting those stones back one by one when the sheep return, so you know if any are missing without really counting. As another example of matching being more fundamental than counting, if I'm speaking to a packed auditorium, where every seat is taken and no one is standing, I know that there are the same number of chairs as people in the audience, even though I don't know how many there are of either. So, what we really mean when we say that two sets are the same size is that the elements in those sets can be matched up one by one in some way. My fourth grade teacher showed us the whole numbers laid out in a row, and below each we have its double. As you can see, the bottom row contains all the even numbers, and we have a one-to-one match. That is, there are as many even numbers as there are numbers. But what still bothers us is our distress over the fact that even numbers seem to be only part of the whole numbers. But does this convince you that I don't have the same number of fingers on my right hand as I do on my left? Of course not. It doesn't matter if you try to match the elements in some way and it doesn't work, that doesn't convince us of anything. If you can find one way in which the elements of two sets do match up, then we say those two sets have the same number of elements. Can you make a list of all the fractions? This might be hard, there are a lot of fractions! And it's not obvious what to put first, or how to be sure all of them are on the list. Nevertheless, there is a very clever way that we can make a list of all the fractions. This was first done by Georg Cantor, in the late eighteen hundreds. First, we put all the fractions into a grid. They're all there. For instance, you can find, say, 117/243, in the 117th row and 243rd column. Now we make a list out of this by starting at the upper left and sweeping back and forth diagonally, skipping over any fraction, like 2/2, that represents the same number as one the we've already picked. We get a list of all the fractions, which means we've created a one-to-one match between the whole numbers and the fractions, despite the fact that we thought maybe there ought to be more fractions. OK, here's where it gets really interesting. You may know that not all real numbers -- that is, not all the numbers on a number line -- are fractions. The square root of two and pi, for instance. Any number like this is called irrational. Not because it's crazy, or anything, but because the fractions are ratios of whole numbers, and so are called rationals; meaning the rest are non-rational, that is, irrational. Irrationals are represented by infinite, non-repeating decimals. So, can we make a one-to-one match between the whole numbers and the set of all the decimals, both the rationals and the irrationals? That is, can we make a list of all the decimal numbers? Cantor showed that you can't. Not merely that we don't know how, but that it can't be done. Look, suppose you claim you have made a list of all the decimals. I'm going to show you that you didn't succeed, by producing a decimal that is not on your list. I'll construct my decimal one place at a time. For the first decimal place of my number, I'll look at the first decimal place of your first number. If it's a one, I'll make mine a two; otherwise I'll make mine a one. For the second place of my number, I'll look at the second place of your second number. Again, if yours is a one, I'll make mine a two, and otherwise I'll make mine a one. See how this is going? The decimal I've produced can't be on your list. Why? Could it be, say, your 143rd number? No, because the 143rd place of my decimal is different from the 143rd place of your 143rd number. I made it that way. Your list is incomplete. It doesn't contain my decimal number. And, no matter what list you give me, I can do the same thing, and produce a decimal that's not on that list. So we're faced with this astounding conclusion: The decimal numbers cannot be put on a list. They represent a bigger infinity that the infinity of whole numbers. So, even though we're familiar with only a few irrationals, like square root of two and pi, the infinity of irrationals is actually greater than the infinity of fractions. Someone once said that the rationals -- the fractions -- are like the stars in the night sky. The irrationals are like the blackness. Cantor also showed that, for any infinite set, forming a new set made of all the subsets of the original set represents a bigger infinity than that original set. This means that, once you have one infinity, you can always make a bigger one by making the set of all subsets of that first set. And then an even bigger one by making the set of all the subsets of that one. And so on. And so, there are an infinite number of infinities of different sizes. If these ideas make you uncomfortable, you are not alone. Some of the greatest mathematicians of Cantor's day were very upset with this stuff. They tried to make these different infinities irrelevant, to make mathematics work without them somehow. Cantor was even vilified personally, and it got so bad for him that he suffered severe depression, and spent the last half of his life in and out of mental institutions. But eventually, his ideas won out. Today, they're considered fundamental and magnificent. All research mathematicians accept these ideas, every college math major learns them, and I've explained them to you in a few minutes. Some day, perhaps, they'll be common knowledge. There's more. We just pointed out that the set of decimal numbers -- that is, the real numbers -- is a bigger infinity than the set of whole numbers. Cantor wondered whether there are infinities of different sizes between these two infinities. He didn't believe there were, but couldn't prove it. Cantor's conjecture became known as the continuum hypothesis. In 1900, the great mathematician David Hilbert listed the continuum hypothesis as the most important unsolved problem in mathematics. The 20th century saw a resolution of this problem, but in a completely unexpected, paradigm-shattering way. In the 1920s, Kurt Gödel showed that you can never prove that the continuum hypothesis is false. Then, in the 1960s, Paul J. Cohen showed that you can never prove that the continuum hypothesis is true. Taken together, these results mean that there are unanswerable questions in mathematics. A very stunning conclusion. Mathematics is rightly considered the pinnacle of human reasoning, but we now know that even mathematics has its limitations. Still, mathematics has some truly amazing things for us to think about.
我在四年級的時候, 國小老師有一天跟我們說: 「偶數的個數 和正整數的個數一樣多。」 「真的嗎?」我心想。 噢對!兩個都是無限多個,所以一樣多。 但另一方面,偶數只是正整數的一部份, 而奇數就是剩下的部份, 所以正整數應該要比偶數還多,對吧? 要了解老師那段話的道理, 我們要知道兩個集合一樣大是什麼意思。 當我說我左手的手指 和右手的手指一樣多時,是什麼意思? 當然,兩隻手都是五根手指, 但是可以更簡單一些。 我不用去算,我只要知道 我能夠將它們「一對一」對應起來。 事實上,我們認為古代那些 語言裡數字只到三的人們 就是用這個技倆。 如果你把你的羊從羊圈裡放出去吃草, 你可以隨時知道有幾隻羊跑出去。 你只要在羊出去時將一顆石子放旁邊, 然後在羊回來的時候 再把石子放回來就好。 這樣你就不會亂掉, 儘管你沒有真的去算羊的數目。 另一個「一對一」的例子 比計數更單純一些。 如果在一個擁擠的禮堂裡, 每個位子都有人坐而且沒人站著, 這樣我就知道 人數跟椅子數一樣多, 雖然說我並不知道 這兩者的個數。 所以,我們說兩個集合一樣大時 真正的意思 就是兩集合裡的元素 有辦法「一對一」對應在一起。 所以國小老師將正整數寫成一列, 並將數字的兩倍寫在下面。 你可以看到,底部那列包含了所有的偶數, 這樣就有了「一對一」的對應。 也就是說,偶數和正整數一樣多。 但依舊苦惱我們的是 偶數只是正整數的一部份這件事實。 不過這樣能說服你 我左右手手指數目不一樣嗎? 當然沒有!就算有的方法 配對失敗,那也沒關係, 因為這並沒說服我們什麼。 如果你可以找到一種方法 讓兩邊元素配對起來, 那我們就說這兩個集合個數一樣。 你有辦法將分數像正整數那樣列出來嗎? 可能有點難,分數有很多! 而且不太明顯哪個要放前面、 或是怎樣把它們串起來。 不過,有一個辦法 我們可以把所有分數依序串起來。 這是十九世紀末 數學家康托爾的貢獻。 首先,我們把分數上下左右對好。 全部的分數都在這。比如說,你可以找到 117/243 它在第 117 列第 243 行。 現在我們要把它們串起來, 從左上開始,然後斜對角地串下來、串上去。 其中像 2/2 這類之前已經算過的分數 就把它跳掉。 因此我們就把分數串成一串了, 這意思是分數 和正整數有「一對一」的對應, 雖然我們直覺是分數比較多個。 好,這就是有趣的地方了。 你也許知道用分數沒辦法表示所有的實數 ──也就是那些數線上的數。 像是根號 2、還有圓周率 π 這些。 這類的數字叫作「無理數」。 不只是因為它們很難懂,而是因為分數包含了 所有整數的「比率」,所以被叫「可比的」, 而剩的就被叫作「不可比的」,也就是「無理的」。 無理數可以用無窮小數表示, 而且各位數沒有規律。 那麼,我們可以將正整數和 所有無理、有理的小數 「一對一」對應嗎? 也就是,我們可以將所有小數串起來嗎? 康托爾證明了這行不通。 不只想不到辦法,而是真的沒辦法。 來看看,如果你聲稱你把小數串好了。 我要來告訴你這是不可能的, 因為我要找一個你那串那面沒有的小數。 我要在小數點後一個一個位數決定。 我要用你那串的第 1 個數字的第 1 位數 來決定我的第 1 位數。 如果它是 1,我的就是 2;否則我的就是 1。 再用你的第 2 個數字的第 2 位數 來決定我的第 2 位數。 一樣,如果你的是 1,我的就是 2; 否則我的就是 1。 看出怎麼算下去了嗎? 我找到的這個小數,不可能在你那串裡。 為什麼?比如說,它和你的第 143 個數會一樣嗎? 不可能,因為第 143 位數裡 你的和我的不一樣。 這是我特別挑的。 你沒串成功。 沒有串到所有小數。 而不論你怎麼串,我都可以做同樣的事, 然後找到一個你那串裡沒出現的小數。 所以我們得到了 令人訝異的結論: 所有小數沒辦法串成一串。 它的「無限大」比正整數的「無限大」還大。 所以,儘管你只熟悉幾個無理數, 像是根號 2 和圓周率 π, 無理數的「無限大」實際上也比 分數的「無限大」還要大。 有人曾這樣比喻: 有理數,或者說分數,就像天空中的星星; 而無理數就像是無盡的黑暗。 康托爾同時也證明任何無窮大的集合, 只要把它的所有子集都蒐集起來, 新的集合的「無限大」就比原本的還大。 意思是說,只要你有一種「無限大」 那你就可以用它的所有子集 來做出比它更「無限大」的集合。 接著再用這集合做出更加「無限大」的集合。 不斷做下去。 所以,「無限大」之間也是有分不同的大小。 如果你覺得這令人想吐,並不奇怪。 一些康托爾那年代的偉大數學家 也對這觀念非常反感。 他們試著要把無限這觀念抽離, 讓數學可以 沒有無限也能運作。 康托爾甚至受到人身攻擊, 嚴重到讓他飽受沮喪之苦, 並且在精神療院渡過後半餘生。 不過他的想法最終得到肯定。 今天,這觀念被認為是基礎並重要的。 所有數學研究者都接受這觀念, 每個數學系都也都在教, 而我剛剛已經花了幾分鐘來解釋。 也許有一天,這會變成大家的常識。 還有一點。我們剛剛指出 小數,也就是實數, 比正整數的「無限大」還多。 康托爾在想兩個「無限大」之間 是否還有不同層級的「無限大」。 我們不這麼認為,但也沒辦法證明。 康托爾的猜想變成 有名的「連續統假說」。 在 1900 年,大數學家希爾伯特 把連續統假說列為 數學裡最重要的未解問題。 這問題在 20 世紀露出一些端倪, 但是結果和超乎預期、並跌破大家眼鏡。 在 1920 年代,哥德爾證明了 你不可能證明連續統假說是錯的。 接著在 1960 年代,寇恩證明了 你不可能證明連續統假說是對的。 合在一起,這些結果告訴你 數學裡也有一些不能回答的問題。 這是一個很令人震驚的結論。 數學被公認是人類邏輯的結晶, 但現在我們知道 就算是數學也有它的極限。 還有就是,數學裡有一些值得我們思考、 而且很令人著迷的道理。