黎曼猜想:数学家与 Claude 声称完成 Lean 证明
4 min Machine translated History & how-to
4 min Machine translated History & how-to
每次登录银行账户,或发送私人消息,你都在依赖质数。
167年来,关于质数最深奥的谜题难倒了地球上所有的数学家。
质数是整数,比如2, 3、5、7和11,它们无法拆成更小的整数相乘得到。
它们是其他所有数的原子。随着数字越来越大,质数越来越少,但它们从不按固定规律出现。
有时两个质数紧挨在一起。有时中间却隔着一大片没有质数的荒漠。
数学家已经找到了一条平滑曲线,可以估算不超过一百万或十亿的范围内大约有多少个质数。
但质数的实际数量呈现为一道锯齿状的阶梯,在这条平滑曲线上下不断起伏。
关键问题是:这种起伏究竟能有多剧烈?回到1859年,一位33岁的德国数学家伯恩哈德·黎曼写了一篇短论文,永远改变了数学。
他研究了一种数学工具,名为ζ函数。把它想象成一片广阔的三维地貌,有山丘也有山谷绵延在一张平坦的数字网格上。
地图上某些特殊的点会一路下沉到海平面,高度为零。
黎曼的突破在于,他意识到这些零点会像波浪一样塑造素数阶梯。
每个零点都会在素数计数中激起一道涟漪,零点在地图上的位置决定了涟漪有多大。
如果所有零点都排列在地图正中央,也就是半处那条细如刀锋的直线上,涟漪就会很微弱,像公平抛硬币产生的噪声。
如果哪怕有一个零点偏到一侧,它就会掀起一股异乎寻常的巨浪,让素数成簇出现,并在剧烈波动中消失。
黎曼猜测,每一个零点都落在那条中心线上,但他在39岁时去世,未能证明这一点。
2000年,克雷数学研究所为此设立了100万美元的奖金。
那么,证明它会带来什么?它不会给你一个能算出下一个素数的神奇公式,也不会破解你的密码。
相反,它证明素数的分布已经尽可能均匀,绝不会出现隐匿的异常波动。
更重要的是,过去一个世纪里,数学家基于黎曼猜想建立了数千条定理,还开发了快速计算机算法,也都假定黎曼是对的。
证明它,就能把所有这些 “如果”变成确凿事实。
计算机已经检验了最前面的10万亿个零点,每一个都恰好落在那条线上。
但零点有无穷多个。检验1万个亿个零点,什么也证明不了,因为上面可能还藏着一个叛逆的零,靠穷举不可能解决无穷问题。
你需要无懈可击的逻辑,一次锁定所有数。如今,人类学数学家 Levent Alpoge 表示,他和一个尚未发布的 Claude 版本找到了一份证明。
Alpoge 制定了开篇策略,把一个更简单的数学世界里的技巧运用到整数上。
但临近终点时,他说,Claude 开始独立行动,想出了远超他的大胆突破。
如果把证明打印成普通论文,篇幅会超过25万页,人类根本无法读完。
所以,他们改为发布了 17.5万个 Lean 代码文件,这是一种严格的数学验证语言,只要有任何一步逻辑不够严密,它就拒绝编译。
MIT 的独立数学家已经编译了整个代码库毫无错误,而且只用了标准数学公理。
所以,电脑判定逻辑成立,但证明实在太庞大,人类可能要花上年才能理解机器究竟是如何证明的。
End of transcript · 630 words · machine translated
Every time you log into your bank or send a private message, you rely on prime numbers.
And for 167 years, the deepest mystery about them defeated every mathematician on Earth.
Primes are whole numbers, like 2, 3, 5, 7, and 11, that can't be split into smaller whole numbers multiplied together.
They are the atoms of every other number.
As you count higher, primes get rarer, but they never arrive on a neat schedule.
Sometimes two sit right next to each other.
Other times, there's a huge desert with none at all.
Mathematicians already have a smooth curve that estimates roughly how many primes exist up to a million or a billion.
But the real count of primes is a jagged staircase that wiggles above and below that smooth curve.
The big question: how wild can that wiggle ever get?
Back in 1859, a 33-year-old German mathematician named Bernhard Riemann wrote a short paper that changed math forever.
He took a mathematical machine called the zeta function.
Picture it as a vast 3D landscape of hills and valleys stretching across a flat grid of numbers.
Certain special spots on that map dip all the way down to sea level, height zero.
Riemann's breakthrough was realizing those zero points act like waves that shape the prime staircase.
Every zero adds a ripple to the prime count, and where a zero sits on the map controls how big its ripple is.
If all the zeros line up down the exact center of the map, one razor-thin straight line at one-half, the ripples stay tiny, like fair coin flip noise.
If even one zero sits off to the side, it creates a rogue wave that makes primes clump and vanish in wild surges.
Riemann guessed every single zero sits on that center line, then died at 39 without proving it.
In 2000, the Clay Mathematics Institute put a 1 million dollar prize on it.
So, what happens if you prove it? It will not give you a magic formula for the next prime, and it won't break your passwords.
Instead, it proves primes are as evenly spread out as math allows.
No hidden rogue waves ever. Even bigger, over the past century, mathematicians built thousands of theorems and fast computer algorithms assuming Riemann was right.
Proving it turns all of those ifs into solid fact.
Computers have checked the first 10 trillion zeros, and every single one landed right on the line.
But there are infinitely many zeros. Checking 10 trillion proves nothing, because a single rebel zero could still be hiding further up.
You can't brute force infinity. You need airtight logic that locks down all of them at once.
Now, anthropic mathematician Levent Alpoge says he and an unreleased version of Claude found a proof.
Alpoge set the opening strategy, adapting a trick from a simpler math world to whole numbers.
But near the finish line, he says Claude took off on its own, inventing leaps far beyond him.
Printed out as a normal paper, the proof would run over 250,000 pages, impossible for any human to read.
So instead, they released 175,000 files of Lean code, a strict math-checking language that refuses to compile unless every single logical step is airtight.
Independent mathematicians at MIT have already compiled the entire code base with zero errors, using only standard math axioms.
So, the computer says the logic holds, but the proof is so massive that it may take humans years just to understand how the machine actually solved it.
TranslateMom transcribes and translates videos from X, TikTok, YouTube and Instagram. Paste a link and you get a page like this one, in your language.