跳到主内容
@wquguru
精选90Hacker News Best(web_list)模型发布/更新

Anthropic:Claude解决黑洞物理两大开放难题

Claude solves two open problems in the physics of black holes

原文
发到 X
推荐理由

这是AI在基础科学领域能力的里程碑式展示,不仅展示了推理深度,更体现了多Agent协作解决复杂数学问题的能力,值得所有关注AI前沿的同学关注。

01Introduction

01 引言

Claude has solved two open problems in theoretical physics. Both concern black holes and the search for a theory of quantum gravity. In the first, it completed a calculation that physicists set up more than 20 years ago, but which had never been solved. Claude’s result showed that a sudden transition associated with the formation of black holes persists even where spacetime no longer has a definite shape. In the second, Claude found the exact limit of the simplest version of a technique used to test candidate theories of quantum gravity. This post explains both findings, first in plain terms and then in more technical detail. The preprints are available here and here.

Claude 解决了理论物理学中的两个未解之谜。这两个问题均与黑洞以及寻找量子引力理论有关。在第一个问题中,它完成了一项物理学家在 20 多年前设定、但从未被解决的计算。Claude 的结果表明,与黑洞形成相关的突然转变,即使在时空不再具有确定形状的情况下依然存在。在第二个问题中,Claude 找到了用于测试候选量子引力理论的最简单版本技术的确切极限。本文首先用通俗语言解释这两项发现,随后提供更详细的技术细节。预印本可在此处和此处获取。

Introduction

引言

Modern physics rests on two frameworks: general relativity and quantum mechanics. General relativity describes very large things, like stars; quantum mechanics describes very small things, such as atoms. Both are astonishingly good at predicting how the universe behaves, but they’re mathematically incompatible. Finding a theory that unites them—“quantum gravity”—is one of the field’s biggest goals. Black holes are a natural testing ground, because they’re one of the few areas where both relativity and quantum mechanics matter—they pack immense mass into a very small space.

现代物理学建立在两个框架之上:广义相对论和量子力学。广义相对论描述非常大的物体,如恒星;量子力学描述非常小的物体,如原子。两者在预测宇宙行为方面都表现出色,但它们在数学上是不兼容的。寻找一个能将二者统一的理论——即“量子引力”——是该领域最大的目标之一。黑洞是一个天然的试验场,因为它们是少数同时涉及相对论和量子力学的领域之一——它们将巨大的质量压缩到极小的空间中。

One of the most useful tools for studying quantum gravity is called holographic duality. The real universe has three dimensions of space and one of time, which relativity combines into four-dimensional spacetime. But physicists studying quantum gravity often work with simpler model universes that they can understand in more detail.

研究量子引力最有用的工具之一被称为全息对偶(holographic duality)。现实宇宙拥有三个空间维度和一个时间维度,相对论将它们结合为四维时空。然而,研究量子引力的物理学家通常使用更简单的模型宇宙,以便更深入地理解它们。

The basic idea of holographic duality is that some of these theoretical universes can be described in two ways: as a complete universe with gravity and spacetime, and in quantum terms, with no gravity at all, “written” on the universe’s outer boundary. Just as a real hologram encodes a 3D image on a 2D plane, the quantum model encodes all the information of the universe in one dimension fewer (a universe with four dimensions of space, for instance, can be fully described by a quantum theory with three). This lets physicists turn hard questions about gravity into questions about quantum physics, which are sometimes easier to answer.

全息对偶的基本思想是,某些理论宇宙可以用两种方式描述:一种是有引力和时空的完整宇宙,另一种是完全无引力的量子描述,这些描述“写”在宇宙的外边界上。正如真实的全息图在二维平面上编码三维图像一样,量子模型在一个少一维的空间中编码了宇宙的所有信息(例如,一个拥有四个空间维度的宇宙,完全可以用拥有三个维度的量子理论来描述)。这使得物理学家能够将关于引力的难题转化为关于量子物理的问题,而后者有时更容易解答。

The first problem Claude solved concerns what happens when you heat up one of these theoretical universes. At low temperatures, the universe fills up with a gas of particles. Above a certain temperature, a black hole forms instead. In 1983, Stephen Hawking and Don Page showed that this switch is abrupt: it’s a phase transition, similar to how water suddenly boils at 100°C. But Hawking and Page’s calculation relies on general relativity, which holds only when the universe is vast, compared with the tiny scale at which that description breaks down. As the theoretical universe shrinks toward that smaller scale, spacetime loses any definite shape and becomes blurry, and nobody knows how to do the calculation on the gravity side of the duality. On the quantum side, though, the same situation can be calculated, at least in principle.

Claude 解决的第一个问题涉及当你加热其中一个理论宇宙时会发生什么。在低温下,宇宙中充满了粒子气体。超过某一温度后,则会形成黑洞。1983 年,斯蒂芬·霍金(Stephen Hawking)和唐·佩奇(Don Page)证明这种转变是突然的:它是一种相变,类似于水在 100°C 时突然沸腾。但霍金和佩奇的计算依赖于广义相对论,而广义相对论仅在宇宙与描述失效的微小尺度相比非常巨大时才成立。随着该理论宇宙向更小的尺度收缩,时空失去任何确定的形状并变得模糊,没有人知道如何在对偶性的引力侧进行计算。不过,在量子侧,至少原则上可以计算相同的情况。

In 1999, physicists found that there was a phase transition in that regime as well. But they didn’t know whether it was abrupt, like Hawking and Page’s, or smooth. In 2003, a group of physicists set out the calculation that would settle the question, but it was extremely complex and was never completed. Claude did it by running a small research group of copies of itself, in which it worked out almost every piece of the calculation independently and checked its math against known results. It showed that the phase transition is abrupt. The same sudden switch appears at both extremes, suggesting that black hole formation isn’t just a feature of relativity but is actually built into the underlying quantum theory.

1999 年,物理学家发现该领域也存在相变。但他们不知道它是像霍金和佩奇那样突然的,还是平滑的。2003 年,一组物理学家着手进行能够解决这个问题所需的计算,但该计算极其复杂,从未完成。Claude 通过运行一个由自身副本组成的小型研究组来完成它,在其中它独立完成了计算的几乎每个部分,并将数学结果与已知结果进行核对。它表明该相变是突然的。同样的突然转变出现在两个极端,这表明黑洞的形成不仅仅是相对论的一个特征,而是实际上内置于底层量子理论之中。

The second problem concerns a simplified theoretical universe with two dimensions of space instead of three. This universe is straightforward enough to study in detail, but still has black holes and other mysteries similar to those of our own universe. Even in this simpler universe, physicists don’t know how to create a full theory of quantum gravity, so they approach it from the other direction: rather than building a single, correct theory, they try to rule out wrong ones by showing that they are internally inconsistent. This method is called bootstrapping.

第二个问题涉及一个简化的理论宇宙,它具有两个空间维度而不是三个。这个宇宙足够简单,足以进行详细研究,但仍具有与我们自己的宇宙类似的黑色洞和其他谜团。即使在这个更简单的宇宙中,物理学家也不知道如何建立完整的量子引力理论,因此他们从另一个方向入手:与其构建单一的正确理论,不如通过证明某些理论内部不一致来排除错误的理论。这种方法称为自举法(bootstrapping)。

Since 2009, physicists have used a version of this method to set a ceiling on how heavy the lightest black hole in such a universe could be. Any theory that breaks the ceiling is ruled out, and each lowering of the ceiling narrows the search. Results from numerical calculations hinted at where the best possible ceiling lay, but they couldn’t prove it, and the best proven ceiling was well short of what the numerical results suggested.

自2009年以来,物理学家们一直使用该方法的一个版本来设定上限,以确定在这样的宇宙中最轻的黑洞可能有多重。任何突破该上限的理论都会被排除,而上限的每一次降低都会缩小搜索范围。数值计算的结果暗示了最佳上限的位置,但无法证明它,且已证明的最佳上限远低于数值结果所暗示的水平。

Claude Fable 5.2 noticed a pattern in the numerical results that physicists had missed and borrowed a method from elsewhere in math and physics to calculate the precise ceiling for the simplest version of this technique and prove that it could not be lowered. It doesn’t identify the correct theory, but it may point physicists in the right direction.

Claude Fable 5.2 在数值结果中发现了一个物理学家们忽略的模式,并从数学和物理学的其他领域借用了一种方法,来计算该技术最简单版本的精确上限,并证明其无法进一步降低。它并未识别出正确的理论,但可能为物理学家指明正确的方向。

Neither result describes our own universe directly, and neither has practical applications, but both contribute to a long-running theoretical effort to understand how quantum mechanics and relativity fit together.

这两项结果均未直接描述我们自身的宇宙,也都没有实际应用,但它们都为一项长期进行的理论研究做出了贡献,旨在理解量子力学与相对论如何相互契合。

In the rest of this post, we explain in more detail what each problem was and what Claude did to solve them.

在本文的其余部分,我们将更详细地解释每个问题是什么,以及 Claude 是如何解决它们的。

A note about Claude’s methods

关于 Claude 方法的说明

In August, an Anthropic researcher, Hongbin Chen, was testing Claude’s capabilities as a side project. He asked Claude to draw up a list of interesting open problems in theoretical physics, specifically focusing on black holes and the bootstrap method. Claude came back hours later with 22 candidates, and Hongbin set up separate agents to work on each one; they could read each other’s progress posts and collaborate when working on related problems. The agents ran in the background, with no special setup, for about six weeks.

今年8月,Anthropic 的研究员 Hongbin Chen 在进行一个副项目时测试 Claude 的能力。他要求 Claude 列出理论物理学中一些有趣的开放性问题,特别聚焦于黑洞和自举法(bootstrap method)。几小时后,Claude 返回了22个候选问题,Hongbin 为此设置了独立的智能体分别处理每一个;这些智能体可以阅读彼此的进展帖子,并在处理相关问题时进行协作。这些智能体在后台运行,无需特殊设置,持续了大约六周。

The agents solved the two problems in this post. One was the Hawking-Page phase transition, the question of when a hot gas turns into a black hole. Weeks after one of the agents reached a solution, a physicist, who had independently used Claude to tackle the same problem, shared his result, and a colleague realized that Hongbin’s agent had reached the same answer earlier. The two solutions used different methods, so they were a useful check on each other.

这些智能体解决了本文中的两个问题。其中一个问题是霍金-佩奇相变(Hawking-Page phase transition),即热气体何时转变为黑洞的问题。在其中一个智能体得出解决方案数周后,一位独立使用 Claude 解决同一问题的物理学家分享了他的结果,另一位同事意识到 Hongbin 的智能体更早得出了相同的答案。这两种解法使用了不同的方法,因此彼此构成了有用的验证。

The modular bootstrap result came from several agents. The one that was assigned the problem spotted the pattern in the numerical results when surveying the bootstrap literature, and another one—which was working on a mathematically related problem about how tightly spheres can be packed into a high-dimensional space—turned that pattern into a method. It found that this method reproduced a sphere-packing result recently proved by an OpenAI model. Claude then applied the same method to the bootstrap problem, which an earlier Claude agent had tried, and failed, to solve in 61 parallel attempts. Two other agents also independently arrived at the same answer. One half of the final proof, showing that the ceiling can’t be lowered, follows the strategy of the OpenAI proof. In September, the results were formally verified in Lean, a computerized proof checker. Claude then wrote the paper with minimal guidance, and a physicist checked every main result.

模块化自举(modular bootstrap)的结果来自多个智能体。被分配该问题的智能体在调研自举文献时,在数值结果中发现了规律;另一个正在处理一个数学上相关的问题——即球体能多紧密地堆积在高维空间中——的智能体将该规律转化为一种方法。该方法发现它能复现由 OpenAI 模型最近证明的一个球体堆积结果。随后,Claude 将同一种方法应用于自举问题,而早期的 Claude 智能体曾尝试解决此问题,但在 61 次并行尝试中均告失败。另外两个智能体也独立得出了相同的答案。最终证明的一半部分(证明上限无法降低)遵循了 OpenAI 证明的策略。9 月份,这些结果在 Lean(一种计算机化证明检查器)中得到了形式化验证。随后,Claude 在极少指导下撰写了论文,并由一位物理学家核查了所有主要结果。

The Hawking-Page transition

霍金-佩奇相变(Hawking-Page transition)

To understand the first problem, we need to start with something much more familiar: a walk through New York City.

要理解第一个问题,我们需要从更熟悉的事物开始:漫步纽约市。

Two equally good routes from Penn Station (A) to the High Line entrance (B). In theory, a quantum particle can take both simultaneously.

从宾州车站(A)到高线公园入口(B)的两条同等优质的路线。理论上,量子粒子可以同时走这两条路。

更进一步:量化金融体系

看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力

进入量化体系 →

相似阅读

关联信息,但可能不是同一事件