String Theory Finally Testable with Power of AI — AI 让弦理论首次变得可检验

原文：String theory finally testable with power of AI，King’s College London 新闻稿，发布于 2026-08-05。 本页结构：第一部分为英文原文（Original Article），第二部分为中文深度解读（解析）。 说明：原文无付费墙，全文完整收录。

第一部分：正文（Original Article）

String theory finally testable with power of AI

Overturning decades of popular conjecture that theoretical string theory cannot be tested, scientists have robustly proved that it can be for the first time.

Through the development of artificial intelligence tools that combine statistics and machine learning, physicists at King’s College London and their collaborators have used a popular dark matter candidate to test the extra dimensions of space that sit at the core of string theory.

First gaining prominence in the late 1960s as a potential ‘theory of everything’, string theory is the idea that reality is made up of an incredible number of vibrating strings, all smaller than fundamental particles like electrons.

As these strings vibrate, they are thought to produce effects in up to ten dimensions which helps explain everything, from the effects of quantum mechanics to gravity and general relativity – uniting the two theories that underlie most of modern physics.

However, no experiment has been able to robustly test or prove string theory correct, in part due to the inability of a single machine to measure the four fundamental forces that make up the world – until now.

To establish a test case to compare string theory against, the team asked the hypothetical question: If the axion, a leading dark matter candidate thought to have a frequency like a wave, was discovered tomorrow, what would happen to string theory?

Published in Physical Review D, the study modelled two different possible masses of the axion detectable by two different experiments, DM Radio and ADMX, using a technique called Bayesian inference to see what this would imply for string theory. Comparing the two models, they found if the heavier axion turned out to be detected, then this does very specific things to the extra dimensional space predicted in string theory.

Surveying millions of different models or topologies of string theory, the team discovered that only a subset of theories would fit the heavier axion – meaning that given a future detection of the axion one could verify that some models did not work through rigorous testing; the first time this has been done.

Author Dr David Marsh, Ernest Rutherford Fellow at King’s College London, said “The popular understanding that you can’t experimentally test string theory is no longer true. While previous studies have tested a few different models of string theory, our computational approach improves on this with rigorous testing across millions of different models.

“Rigorous falsification is the core of scientific experiment, and by finally applying this to string theory we finally have the tools to test this vaunted ‘theory of everything’. Finally touching experimentation here is an exciting step forward for theoretical particle physics.”

Building on their rigorous statistical methods, the team also developed a parallel avenue to test string theory using real-world data. Cosmic Microwave Background (CMB) is the cooled radiation that fills the universe, thought to be left over from the Big Bang. However, there exists a small discrepancy between its temperature distribution and the distribution of Hydrogen gas in the universe as measured by the Lyman-alpha forest, an important astrophysical probe. King’s Dr Keir Rogers showed previously this could be solved if dark matter was ultralight and described by a wave function, another prediction of axion models of dark matter.

In a full analysis of the cosmological data from the CMB and the Lyman alpha forest, the group developed a predictive model for the extra dimensions of string theory. They identified just a tiny handful of string theory models capable of explaining the discrepancy in the data.

Using their predictive model, the group then showed that this handful of string theory models also predicted new types of particles that could help make up complex dark matter. If these new predicted particles were found, string theory would once again be falsifiable in a major step forward for theoretical physics.

This research was led by King’s College London postdoc Dr Mudit Jain, funded by a Leverhulme Research Project.

第二部分：解析（深度解读）

核心论点

伦敦国王学院（KCL）团队用「统计 + 机器学习」的 AI 工具，推翻了「弦理论无法被实验检验」这一流行了几十年的成见，证明弦理论首次可以被严格检验。关键是：他们没有直接去测弦理论本身，而是用「如果某个暗物质候选者（轴子，axion）明天被发现，弦理论会怎样」这一反事实问题，构造出一个可证伪的检验场景。

关键概念

1. 弦理论的本质：1960 年代末兴起的「万有理论」候选，认为现实由极小的振动弦构成，振动在多达十个维度上产生效应，试图统一量子力学与引力/广义相对论。但长期以来，没有任何实验能稳健地检验或证实它——部分原因是单台机器无法同时测量构成世界的四种基本力。
2. 轴子（axion）作为探针：轴子是领先的暗物质候选者，被认为具有类似波的频率。团队提出假设：若轴子明天被发现，对弦理论意味着什么？
3. 贝叶斯推断（Bayesian inference）：研究发表在 Physical Review D，用贝叶斯推断为两种不同质量、分别可被 DM Radio 与 ADMX 实验探测的轴子建模，看其对弦理论的额外维度空间意味着什么。结论是：若探测到「较重」的轴子，会对弦理论预言的额外维度空间产生非常具体的约束。
4. 在数百万个模型上做可证伪性测试：团队扫描了数百万个不同的弦理论模型/拓扑，发现只有其中一小部分能容下较重的轴子。这意味着，一旦未来真的探测到轴子，就可以据此严格验证哪些模型不成立——这是史上第一次做到这种规模的可证伪性测试。
5. 用真实宇宙学数据做第二条路径：宇宙微波背景（CMB）与 Lyman-alpha 森林测得的氢气体分布之间存在微小差异；若暗物质是「超轻、由波函数描述」的（轴子模型的另一预言），该差异可被解释。团队据此建立弦理论额外维度的预测模型，只筛出极少数能解释该数据差异的弦理论模型，并进一步发现这些模型还预言了能构成复杂暗物质的新粒子——若被发现，弦理论将再次可证伪。

技术趋势与判断

• 从「几个模型」到「数百万模型」：过去的研究只检验过少数几个弦理论模型；AI/统计方法把严格检验的规模提升到了数百万级，使「证伪」真正成为可行操作。这正是 David Marsh 所说的——严格的可证伪性（falsification）才是科学实验的核心。
• AI 在理论物理中的角色：这里 AI 不是「发现新物理」，而是作为大规模参数空间扫描与贝叶斯推断的加速器——在弦理论庞大的「理论 landscape」里，快速定位哪些模型与观测相容、哪些可被排除。
• 「可检验」≠「已被证实」：需注意，这是「让弦理论变得可检验/可证伪」，而非「证实了弦理论」。它把弦理论从「永远无法判断对错」推进到「未来观测可以排除一大批模型」的状态。

与本站其他文章的连接

• 与本站「AI 用于科学发现」的脉络一致：AI 在这里充当高维假设空间的搜索引擎，与材料/生物/化学中「用 ML 缩小搜索空间」的范式同源。
• 贝叶斯方法与「用数据排除假设」的思路，也可类比到半导体/光通信领域对大量技术路线的筛选与验证。

风险提示

• 该结论依赖「轴子存在且可被 DM Radio/ADMX 探测到」这一前提；轴子至今仍是假设性粒子，尚未被确认。
• 「可证伪性」是理论物理的方法论进步，但距离实验上真正排除/确认具体模型仍有赖于未来探测器的灵敏度与上天时间。