返回资讯中心
外部精选
软件工程
#开发者实时流

On Beyond Omega

A mysterious gentleman offers to give you a bag containing as many coins of any denomination of U.S. currency as you want. A million quarters? Easy. A billion double eagles, worth $20 apiece? No problem; the magic bag w…

Hacker News 最新jamespropp2026/10/18 · 2026/10/1830 分钟阅读

以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。

Because most people stop at the Z But not me!

A mysterious gentleman offers to give you a bag containing as many coins of any denomination of U.S. currency as you want. A million quarters? Easy. A billion double eagles, worth $20 apiece? No problem; the magic bag will enlarge itself to make room for them.

“How about infinitely many?” you ask hopefully.

“Oh, I can’t do that,” says the gentleman, shaking his head. “I can only ever put finitely many coins into the bag. But no finite number is too big for me.”

“Oh, it’s such a little thing,” the gentleman says. “But I give away money for the pleasure of seeing people use it, so I must insist that you remove a coin from the bag every day, to spend or invest however you wish. And no fair adding money to the bag yourself; only I can do that.”

“Oh, but then I’ll run out of money eventually,” you say.

“Not to worry,” he replies; “each time you remove a coin larger than a penny, I’ll give you as many coins of lower denomination as you like. Most banks will give you only two nickels or ten pennies in exchange for a dime. I’ll do much better: you get to spend the dime, and I’ll put a million nickels and a million pennies into the bag if you wish.”

“Oh, I won’t let that happen; that would spoil my fun! I’ll keep you alive for as long as you play this little game with me.”

“I might run out of money, say. What happens then?”

The devil sighs (for of course that’s who the gentleman is). “I was afraid you’d ask that tiresome question. If you use up all the coins in the bag and thus spoil my fun, then I will collect your soul and subject you to eternal torment. That’s my other kind of fun. But are you seriously worried about that? Consider the terms of the deal! I’ll give you all the coins you ask for up front, and whenever you remove anything larger than a penny I’ll put as many lower-denomination coins into the bag as you like. Are you really concerned you’ll run out of money, ever? What could possibly go wrong?”

This parable is an intro to today’s topic, but it’s not ideal. For instance, you may be thinking about the far future of the universe, when perhaps all the protons will have decayed, except maybe yours, and there’s nothing to spend money on. But this is a math puzzle, not a physics puzzle, so let’s rephrase it in a more mathematical way. For definiteness, we’ll assume that there are ten denominations of U.S. coins.1 So the state of the bag at any moment is given by a list of ten whole numbers, each zero or greater, recording the number of coins of each denomination from $20 and $10 coins down to quarters, dimes, nickels, and pennies. For instance, the sequence

(call it A) corresponds to a bag containing nothing but 2 dimes, while the sequence

(call it B) corresponds to a bag containing 1 dime, 10 nickels, and 10 pennies. A legal move for you is to reduce one of the positive numbers by 1 and, if it wasn’t the final number in the list (corresponding to the case in which you took a penny from the bag), to increase any or all of the later numbers by as much as you like. For instance, turning sequence A into sequence B corresponds to taking a dime out of the bag and having the devil respond by putting in ten nickels and ten pennies.

In a purely mathematical universe that never ends, can you choose the initial ten numbers so that you can make legal moves for all time? Or must the numbers eventually all become zero, forcing you (back in the parable) to forfeit your soul?

We’ll get to the answer later, but meanwhile, note that I’m not asking whether there’s a limit on how long you can make the game last. Start with a bag containing a googol pennies and you can play for a googol days; ditto for a googolplex. When you commit to how many pennies you want to start with—and remember, it must be a finite number—you’re basically deciding how many days of life you’ll get before your eternal damnation begins. You can make the game last as long as you like in any predefined sense, but that’s not the same as making it last forever. You get to choose when you go to hell, but not whether you eventually go there.

In a similar vein, it’s important to stress that the devil doesn’t allow backsies. If your bag becomes empty, you can’t say “Wait, I didn’t play this game the right way; let’s start over.” Nope! Off to hell you must go. If the devil did allow backsies, that’d be a different game, and you could keep fending off damnation forever with the “Oh, let me try again” gambit. But you’re required to commit to an initial supply of money and once it’s exhausted, you’re damned.

The parable provides an entry point into one of the most revolutionary developments in nineteenth-century mathematics: Georg Cantor’s theory of infinite ordinals, and his curious way of counting past infinity. Yet when Cantor first explored his new kind of counting, he wasn’t trying to revolutionize mathematics; he was just trying to keep track of a whole lot of sets of numbers in an orderly way.

Georg Cantor was born in St. Petersburg in 1845 and moved with his family to Germany when he was eleven. He showed early talent in mathematics, especially trigonometry. After earning his doctorate in number theory in Berlin in 1867, he took a position at the University of Halle, where his senior colleague Eduard Heine urged him to work on the uniqueness problem for trigonometric series.

A trigonometric series is an expression of the form

f(t) = a1 cos t + b1 sin t + a2 cos 2t + b2 sin 2t + a3 cos 3t + b3 sin 3t + …

(it’s customary to include a constant term but let’s not). You can think of it as a musical sound composed of a fundamental pitch and all its overtones. The trigonometric coefficients a1, b1, a2, b2, a3, b3, … describe the mixture of overtones; this sonic recipe largely determines whether we hear a violin, oboe, trumpet, or some other instrument.2

The uniqueness problem asks, could there be two different sonic recipes that give rise to the same sound? Or, to paraphrase the problem more strikingly, could there be a sonic recipe for silence that’s different from the trivial recipe in which all the trigonometric coefficients are equal to 0?3

Since all the sine and cosine functions we’re combining repeat with period 2π (this is advanced trig, so we use radians, not degrees), we don’t lose anything if we restrict to values of t between −π and π.

To get a sense of what makes the problem hard, consider the following plot of what mathematicians call a “bump function”:

The bump function can be represented by a perfectly nice trigonometric series with the property that f(t) = 0 on half of the interval from −π to π. Somehow different overtones conspire to cancel out except for half of the values of t. You might say that this bump function goes halfway toward contradicting the uniqueness conjecture. In fact, with other trigonometric coefficients you can create a narrower bump function that takes the value 0 on ninety-nine percent of its domain. But when you require that the function f(t) take the value 0 everywhere—that is, when you require total silence at every instant—then the uniqueness conjecture says that the only choice you’re left with is a1 = 0, b1 = 0, etc.

I like to imagine that Cantor—an excellent violinist as well as a mathematician—was pleased with a topic that combined his boyhood fascination with sines and cosines with his love of music. In any case, Cantor was able to solve this problem in 1870. In our music metaphor, Cantor was able to show, in a purely theoretical context, that there are no other kinds of silence beyond the kind you hear when all the individual harmonics are silent.

Cantor might have then moved on to a different problem. But having reaped the fruits of his labors on the uniqueness problem, he couldn’t resist shaking the tree of knowledge to see what other fruit might fall.

Cantor showed in 1871 that the conclusion “a1 = 0, b1 = 0, and so on” is guaranteed to hold true even if we weaken the hypothesis of his uniqueness theorem a tiny bit. Suppose we have a trigonometric series f(t) that takes the value 0 for all t between −π and π, except for one value of t which might satisfy f(t) = 0 or might not. Then he showed that all the trigonometric coefficients must equal 0 (from which it follows incidentally that f(t) = 0 for that one excluded value after all).

Indeed, Cantor showed that if you exclude two values of t from your f(t) = 0 hypothesis, or three values, or any finite number, then the same conclusion holds: if f(t) = 0 for all t other than possibly the excluded values, then the trigonometric coefficients must all be 0.

In modern terminology, he showed that every finite set is a “set of uniqueness”, or U-set, where we say a set S is a U-set provided that every trigonometric series that equals 0 for all values of t outside the set S has all its coefficients equal to 0.

Cantor wondered, “Are there bigger U-sets?” and friends, were there ever! Cantor found a hierarchy of increasingly complicated sets that were nevertheless U-sets, arranged in orderly ranks. At the bottom we have the empty set, assigned rank 0. Just above the empty set we have the nonempty finite sets, assigned rank 1. Above that we have certain infinite sets that have rank 2, and above that we have more complicated infinite sets that have rank 3, and so on.4 And all of them are U-sets.

But Cantor was just getting started. He found sets that aren’t of any finite rank but were nonetheless U-sets. The simplest new ones he found were assigned the symbolic rank ∞.5 Slightly more complicated U-sets were assigned rank ∞+1. Beyond rank ∞+1 came rank ∞+2, and then rank ∞+3, and so on. Beyond all those ranks came rank ∞+∞, aka ∞⋅2 (though Cantor himself wrote it as 2∞). Then came ∞⋅2 + 1, and ∞⋅2 + 2, and so on; and on beyond those came ∞⋅3.

On beyond ∞ and ∞⋅2 and ∞⋅3 etc. came ∞⋅∞, aka ∞2.

On beyond ∞2 and ∞2 ⋅2 and ∞2 ⋅3 etc. came ∞2 ⋅∞, aka ∞3.

On beyond ∞ and ∞2 and ∞3 etc. came— you guessed it— ∞∞.

There’s much, much more beyond ∞∞ , but that’s as far along that road as I’m going to take you right now. Don’t let the symbols confuse you; the important thing to remember is that Cantor kept applying the same trick—asking “Is there a way to go even farther?”—over and over again.

At some point Cantor stopped using the symbol ∞ and started using the Greek letter ω, the final letter of the Greek alphabet and a longstanding symbol of The Ultimate, as in “I am the Alpha and the Omega” from the Book of Revelations. Except that in Cantor’s new theory, which nowadays is called the theory of transfinite ordinals, ω wasn’t the end of the line.

But what could it mean for an infinite quantity not to be the end of the line?

A good way to get an intuitive handle on the ways of ordinals, or at least the kinds of ordinals that turn up in Cantor’s generalization of his uniqueness theorem, is with pictures.6 These pictures are subsets of the real number line. In the case where n is a natural number, a “picture” of n is just any old set consisting of exactly n real numbers. But when we picture infinite ordinals, things get more interesting.

One obvious way to picture ω, the smallest of Cantor’s infinite ordinals, is with the set { 0, 1, 2, 3, …} containing all the natural numbers, but for the purpose of preparing to move on to larger ordinals, a better picture of ω is the set { 0/1, 1/2, 2/3, 3/4, … }. These fractions are in the same order as the corresponding natural numbers, but they conveniently fit on a finite stretch of the number line. There are lots of other pictures of ω; for instance, the set { 1/2, 2/3, 3/4, … } and the set { 0/1, 2/3, 4/5, … } are also pictures of ω. When we’re picturing ordinals we have plenty of freedom, and that freedom will come in handy soon.

Let’s stick with { 0/1, 1/2, 2/3, 3/4, … } for now as our picture of the ordinal ω. To get a picture of the ordinal ω+1, just adjoin the real number 1 to the set. I’ll write the resulting enlarged set a

正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。

打开原文
去社区讨论 →