教 AI 构建科学模型

Teaching AI to build scientific models

Anthropic · 2026-11-06 · Anthropic Research ↗

打开互动全文版(逐段中英对照 + 针对本篇提问)→

摘要 · Abstract

本文探讨 AI 如何改变科学建模的三个核心步骤:构建模型、约束模型与解释模型。以神经细胞电活动的 Hodgkin-Huxley 模型为例,文章指出仅靠数据拟合并不充分,因为许多参数组合或模型结构都能同样好地解释同一批观测结果。作者介绍了基于模拟的推断方法,该方法利用大规模前向模拟与神经网络来恢复参数的完整概率分布,而非单一最优拟合;同时介绍了能够提出并修订候选模型的基于大语言模型的智能体。核心结论是:AI 应当返回全部合理的模型与参数范围,而非仅仅给出一个答案,因为理解要求我们知道哪些替代方案会失败以及为何失败。正如 Richard Hamming 所言,计算的目的在于洞见,而非数字。

This article examines how AI is transforming the three core steps of scientific modeling: building, constraining, and interpreting models. Drawing on the Hodgkin-Huxley model of nerve cell electrical activity, it argues that fitting data alone is insufficient, since many parameter sets or model structures can explain the same observations equally well. The authors describe simulation-based inference, which uses massive forward simulations and neural networks to recover full probability distributions over parameters rather than single best fits, and LLM-based agents that can propose and revise candidate models. The central conclusion is that AI should return the whole range of plausible models and parameters, not just one answer, because understanding requires knowing which alternatives fail and why. As Richard Hamming noted, the purpose of computing is insight, not numbers.

核心贡献 · Key contributions

局限 · Limitations

章节 · Sections(共 6)

全文 · Full text(逐段中英对照)

01 三步计划 01 A three-step program

* 02 分摊账单,或如何约束模型

* 02 Splitting the bill, or how to constrain models

* 03 账单,再升一级,或如何构建模型

* 03 The bill, one level up, or how to build models

* 04 迷失在翻译中,或如何解释模型

* 04 Lost in translation, or how to interpret models

_AI for Science 资助获得者 Jakob Macke、Daniel Gedon 和 Stefan Wahl 是 ELLIS Institute Tübingen 与蒂宾根大学的研究人员,他们在那里构建用于神经科学及其他领域科学建模的 AI 工具。在这篇客座文章中,他们讨论了 AI 如何开始改变科学建模。_

_AI for Science grantees Jakob Macke, Daniel Gedon, and Stefan Wahl are researchers at the ELLIS Institute Tübingen and the University of Tübingen, where they build AI tools for developing scientific models in neuroscience and beyond. In this guest post, they discuss how AI is starting to change scientific modeling._

1951 年,剑桥大学的计算机停摆了六个月。这对生理学家 Andrew Huxley 来说很不方便,因为他与同事 Alan Hodgkin 刚刚构建了一组方程,用来描述某种鱿鱼神经细胞如何放电。科学家们想理解这些神经细胞的“动作电位”:它们用来通信的电脉冲,这是大脑一切活动的基础,从感知到智能。要判断这些方程是否有效,唯一的方法就是求解它们,并将解与数据进行比较。

In 1951, Cambridge University's computer went out of action for six months. This was inconvenient for physiologist Andrew Huxley, who, alongside his colleague Alan Hodgkin, had just built a set of equations to describe how a particular squid nerve cell fired. The scientists wanted to understand the "action potential" of these nerve cells: the electrical pulses they use to communicate, which underlie everything brains do, from perception to intelligence. The only way to find out whether the equations worked was to solve them and compare the solutions to data.

赫胥黎曾希望用新奇的计算机来求解方程。然而,他被迫使用手摇机械计算器来求解。这花了他三个多星期。

Huxley had hoped to use the newfangled computer to solve the equations. Instead, he was forced to solve them using a hand-cranked mechanical calculator. It took him over three weeks.

他们模型的基本结构由关于细胞的已知事实决定。霍奇金和赫胥黎将细胞膜建模为电路。然而,模型的一部分只是有根据的猜测。为了使曲线拟合他们测量的数据,他们发现必须将一个变量提升到四次方——即将其自乘三次。当时,他们不知道为什么。他们只知道,否则模型无法很好地预测观测数据。

The basic structure of their model was dictated by known facts about the cell. Hodgkin and Huxley modeled its membrane as an electrical circuit. One part of the model, however, was just an educated guess. In order to make the curve fit the data they had measured, they found that they had to raise one variable to the fourth power—that is, to multiply it by itself three times. At the time, they did not know why. They just knew that otherwise the model didn’t predict the observed data as well.

然而,他们确实提出了一个假设。他们谨慎地提出,所讨论的变量(控制细胞膜中特定门如何打开)可能需要四次方,因为膜中的多个带电粒子必须在门打开之前移动到适当位置。

They did offer a hypothesis, however. They cautiously suggested that the variable in question, which governed how particular gates in the cell membrane opened, might require the fourth power because multiple charged particles in the membrane had to move into place before the gate would open.

几十年后,在霍奇金和赫胥黎于 1963 年获得诺贝尔生理学或医学奖很久之后,另一个团队发现了相关钾通道蛋白的结构。他们发现,是的,它有四个相同的部分围绕一个中央孔排列。该模型于 1951 年拟合数据。即使在当时,它也相对准确地预测了脉冲传播的速度。更重要的是,它帮助预测了钾离子通道本身的结构,而这是它从未拟合过的。

Decades later, long after both Hodgkin and Huxley had won the Nobel Prize in Physiology or Medicine in 1963, another team found the structure of a related potassium channel protein. And they discovered that, yes, it had four identical pieces arranged around a central pore. The model was fit to the data in 1951. Even then, it predicted how fast the impulse travels relatively accurately. More than that, it turned out to have helped predict the structure of potassium ion channels themselves, which it had never been fit to.

如今,一个 AI 系统可能在一个下午就能找到像霍奇金和赫胥黎那样的方程,或许它会更紧密地拟合实验数据。但这个故事的要点是,仅仅知道四次方起作用是不够的;重要的是知道模型的更简单版本不起作用。一个仅仅给你模型的 AI 系统可能给你一个答案,但它没有给你理解。用模型重现数据是不够的,因为任何数量的模型都可以做到同样的事情。重要的是能够对现实世界系统本身做出断言,理想情况下对其他类似系统也能做出断言。要做到这一点,你需要知道模型与哪些其他有效的模型共享哪些特征,以及哪些特征使它区别于无效的模型。正如数学家理查德·汉明在 1962 年所写:“计算的目的在于洞察,而非数字。”

Today, an AI system could plausibly find an equation like Hodgkin and Huxley’s in an afternoon, and perhaps it would fit the experimental data even more closely. But the point of the story is that simply knowing the fourth power did the job was not enough; what mattered was knowing that a simpler version of the model did not work. An AI system that merely hands you the model might give you an answer, but it does not give you understanding. Reproducing the data with a model is insufficient, because any number of models could do the same. What matters is being able to make claims about the real-world system itself, and ideally about other, similar systems. To do so, you need to know which features the model shares with other models that work, and which set it apart from models that don’t. As the mathematician Richard Hamming wrote in 1962, “The purpose of computing is insight, not numbers.”

三步计划 A three-step program

70 多年过去了,Hodgkin-Huxley 模型仍然是神经元电活动的标准描述。同样的方程,只是常数不同,就能描述从鱿鱼到人类整个动物界的神经细胞,尽管我们的进化谱系在五亿多年前就已经分道扬镳。

More than 70 years on, the Hodgkin-Huxley model remains the standard description of electrical activity in neurons. The same kinds of equations, with different constants, describe nerve cells across the animal kingdom, from squids to humans, even though our evolutionary lineages separated more than half a billion years ago.

然而,比它的长寿更重要的是,Hodgkin-Huxley 模型告诉我们模型是如何工作的。Hodgkin 和 Huxley 的方法可以分解为三个步骤,而大多数科学建模都涉及它们的某种版本。

Even more important than its longevity, however, is what the Hodgkin-Huxley model tells us about how models work. Hodgkin and Huxley's approach can be broken down into three steps, and most scientific modeling involves some version of them.

* 第一步是**构建**模型。这意味着决定模型由什么组成以及方程采取什么形式。在 Hodgkin 和 Huxley 的情况下,这意味着电路的部分——膜和门——以及方程的形式,例如将基础变量提升到四次方。

* The first is **building** the model. This means deciding what the model is made of and what shape its equations take. In Hodgkin and Huxley's case, that meant the parts of the electrical circuit—the membrane and the gates—and the form of the equations, such as raising the underlying variable to the fourth power.

* 第二步是**约束**模型——调整其可调参数以拟合实际测量值。对于 Hodgkin 和 Huxley 来说,这包括诸如在不同电压下门打开的频率等设置。

* The second is **constraining** the model—tuning its adjustable parameters to fit real measurements. For Hodgkin and Huxley, that included settings like how often the gates open under different voltages.

* 第三步是**解释**模型——询问模型对真实系统说明了什么,例如表明四次方意味着门有四个运动部件。

* The third is **interpreting** the model—asking what the model says about the real system, such as suggesting that the fourth power meant the gate had four moving parts.

在实践中,科学家会反复经历这些步骤。如果数值与测量结果不符,或者解释讲不通,他们就会回过头去修改。

In practice, scientists go through these steps repeatedly. If the numbers fail to fit the measurements, or if the interpretation makes no sense, they go back and revise.

这三个步骤过去都由人完成,但 AI 系统正越来越多地参与其中每一步。可以说,机器在约束模型方面取得的进展最大,因此我们就从那里开始。

All three steps were once done by people, but AI systems are increasingly involved in each of them. Machines have arguably made the most progress in constraining models, so we'll start there.

分摊账单,或如何约束模型 Splitting the bill, or how to constrain models

当 Hodgkin 和 Huxley 通过肉眼和直觉来约束他们的模型时,他们最终得到了一组数字。这告诉他们模型*可以*匹配数据。但这并没有告诉他们其他数字是否也同样有效,甚至更好。

When Hodgkin and Huxley constrained their model by eye and intuition, they ended up with a single set of numbers. This told them that the model *could* match the data. It didn't tell them whether other numbers would have worked just as well or even better.

想象一下,你知道一顿团体晚餐总共花了 120 美元,但不知道谁点了什么。许多不同的点餐组合都可能达到这个金额——也许每个人都点了中等价位的汉堡,或者一个人点了牛排和龙虾,而其他人只点了汤。神经细胞模型的工作原理相同:不同的门控参数组合可以产生几乎相同的电活动。仅仅展示一个拟合良好的组合会隐藏所有其他合理的参数设置,而其他组合可能会产生不同的预测。这几乎适用于任何被建模的现象。许多不同的参数集可以同样好地解释观察到的数据。

Imagine you know that a group dinner added up to $120, but not who ordered what. Many different combinations of orders could have added up to that amount—perhaps everyone ordered a mid-priced burger, or perhaps one person had steak and lobster while everyone else had soup. Models of nerve cells work on the same principle: different combinations of gate parameters can produce near-identical electrical activity. Simply showing one combination that fits well hides all the other plausible parameter settings, and other combinations might have led to different predictions. This is true of almost any phenomenon being modeled. Many different sets of parameters can explain the observed data equally well.

约束 Hodgkin-Huxley 模型的示意图。(1) 来自神经细胞的测量。(2) 对于每种钾和钠通道设置组合,该测量的可能性。颜色越深表示拟合越好。蓝点是交互式的。(3) 在三种设置下模型的模拟结果与测量值(灰色)的比较。有些拟合得很好,而另一些则不然。

An illustration of constraining the Hodgkin-Huxley model. (1) A measurement from a nerve cell. (2) How likely the measurement is for each combination of potassium and sodium channel settings. Darker means a better fit. The blue dot is interactive. (3) Simulations of the model at three settings, compared to the measurement (gray). Some fit well, while others do not.

解决方案是一种称为贝叶斯推断的统计方法。它给出所有合理有效的数字集合,以概率分布的形式呈现,该分布考虑了它们对数据的拟合程度以及你对参数的先验知识。分布呈曲线形状,最佳拟合在峰值处,较弱的拟合在边缘。

The solution is a statistical method called Bayesian inference. It gives you every set of numbers that works reasonably well, as a probability distribution that takes into account their fit to the data and your prior knowledge about the parameters. The distribution takes the shape of a curve, with the best fits at the peak and the weaker ones at the fringes.

但这种方法对于复杂模型通常不切实际,因为它涉及计算在每种可能的数字集合下你的测量值的可能性。在 Hodgkin 和 Huxley 的模型中,这仍然是可行的,但对于任何更复杂的情况,这在计算上是不可能的。例如,如果数千个离子通道中的每一个都随机打开和关闭,你将不得不计算出每一种组合。回到我们的餐厅账单类比:回答“如果 A 点了汉堡,B 点了牛排,账单会是多少?”很容易,但要找出哪些组合可能导致最终账单则难得多,尤其是随着人数和菜单选项的增加。

But this method has usually been impractical for complex models, because it involves calculating how likely your measurements would be under every possible set of numbers. In Hodgkin and Huxley's model, that would still be feasible, but for anything more complicated, it would be computationally impossible. For example, if each of the thousands of ion channels opens and closes at random, you would have to work out every combination. Returning to our restaurant bill analogy: it would be easy to answer, "If person A had the burger and person B had the steak, what would the bill come to?" but much harder to show which combinations could have led to the final bill, especially as the number of people and menu options grows.

一种称为基于模拟的推断(SBI)的不同方法通过纯粹的计算克服了这种逆向工程问题。你随机选择参数,运行模型,并记录输出。你一次又一次地这样做——对于像 Hodgkin-Huxley 这样的模型,可能需要 100,000 次,而对于更大、更复杂的模型,可能需要数百万次。然后,你根据结果训练一个神经网络,使得给定一个输出,它能够返回可能产生该输出的参数范围。

A different approach called simulation-based inference (SBI) overcomes this reverse-engineering problem with sheer computation. You pick random parameters, run the model, and record the output. You do it again and again—perhaps 100,000 times for a model like Hodgkin-Huxley, and millions of times for larger, more complex ones. Then you train a neural network on the results so that, given an output, it can return the range of parameters that could have produced it.

SBI 所需要的只是你能够运行模型:输入一组参数,并得到你期望看到的那种数据。一个可以这样运行的模型——参数输入,数据输出——被称为 _模拟器_,因此得名“基于模拟的推断”。你永远不需要使用公式来计算你的测量值的可能性,而这一步使得标准的贝叶斯推断通常不切实际。如今,模拟器通常是计算机程序,但它不一定是:Huxley 手工计算方程时,就是在运行一个模拟器。许多模拟器易于正向运行,但无法逆向。有些甚至无法写成一个简洁的方程——它们的假设内置于它们遵循的规则中——但它们仍然是模型。对于许多复杂现象,从神经元网络到碰撞的星系,模拟器通常是我们拥有的唯一模型类型。SBI 让我们能够用数据来约束这些模型。

All SBI requires is that you can run the model: put in a set of parameters and get back the kind of data you would expect to see. A model you can run like this—parameters in, data out—is called a _simulator_, hence the name “simulation-based inference.” You never need to use a formula to calculate how likely your measurements are, which is the step that makes standard Bayesian inference often impractical. These days a simulator is usually a computer program, but it doesn’t have to be: Huxley, cranking through the equations by hand, was running one. Many simulators are easy to run forward but impossible to reverse. Some can’t even be written out as a tidy equation—their assumptions are built into the rules they follow—but they are models nonetheless. For many complex phenomena, from networks of neurons to colliding galaxies, simulators are often the only kind of model we have. SBI lets us constrain those models with data.

运行所有这些模拟是昂贵的,可能需要数周时间。但一旦完成,训练好的网络可以在几秒钟内对新数据进行推断,而无需进一步的模拟。这种速度改变了可能性。当两颗中子星碰撞时,SBI 可以大致计算出碰撞发生在天空中的位置,速度足够快,可以在信号仍然可检测时指向望远镜。同样的方法已成功应用于宇宙学、结构生物学、地球物理学和许多其他领域的问题。

Running all those simulations is expensive and can take weeks. But once it is done, a trained network can perform inference on new data in seconds, without requiring further simulations. That speed has changed what’s possible. When two neutron stars collide, SBI can work out roughly where in the sky the collision happened, fast enough to point telescopes at it while the signal is still detectable. The same approach has been successfully applied to problems in cosmology, structural biology, geophysics, and many other domains.

SBI 已被应用于神经细胞模型,包括 Hodgkin-Huxley 模型。因为它返回的是一个分布而不是单一拟合,我们可以看到哪些参数被数据很好地约束,哪些地方有多个解释是合理的,以及哪些地方最需要更多数据——从而进行另一个实验。

And SBI has been applied to nerve cell models, including the Hodgkin-Huxley model. Because it returns a distribution rather than a single fit, we can see which parameters are well constrained by the data, where more than one explanation is plausible, and where more data—and thus another experiment—is most needed.

账单之上:如何构建模型 The bill, one level up, or how to build models

在约束一个模型之前,你需要先有一个可供约束的模型。Hodgkin-Huxley 模型的方程并非从第一性原理推导而来。相反,其中底层变量的四次方是**经验性地**从科学家的生物学洞见中浮现出来的,也源于他们发现更简单的版本效果不佳。

Before you can constrain a model, you need a model to constrain. The equations of the Hodgkin-Huxley model were not derived from first principles. Instead, the fourth power of the underlying variable emerged _empirically_ from the scientists' biological insight, and from finding that simpler versions didn't work as well.

在科学史的大部分时间里,模型就是这样构建的。科学家凭借自身的专业知识和假设,写出一个候选模型。然后他们对其进行检验、修正,再试一次。问题在于,可能的模型空间本质上是无限的,而一位科学家凭经验提出少数几个模型,极有可能错过那些本可以同样好地解释观测数据的模型。

For most of the history of science, this is how models have been built. Scientists used their expertise and assumptions to write down a candidate model. Then they tested it, revised it, and tried again. The problem is that the space of possible models is essentially infinite, and a scientist proposing a handful of models from experience will very likely miss models that could have explained the observational data equally well.

如今,AI 系统可以协助、甚至接管这一步。给定一些数据以及对某个系统的自然语言描述,一个基于 LLM 的智能体会为某个候选模型或一组候选模型编写代码。然后它运行该模型、约束其参数、将输出与真实世界数据进行比较,并对其进行修正——也许是通过修改某个方程的形式。这个循环不断重复,直到模型足够准确地复现数据。过去两年中已出现了若干这样的系统。1 细节各有不同,但都用 LLM 来设计模型本身。直到最近,这还被认为是只有人类才能做到的事。

AI systems can now help with, or even take over, this step. Given some data and a natural-language description of a system, an LLM-based agent writes code for a candidate model or a set of candidate models. Then it runs the model, constrains its parameters, compares the output to real-world data, and revises it, perhaps by modifying the form of an equation. This loop repeats until the model replicates the data accurately enough. Several such systems have appeared in the past two years.1 The details vary, but all use an LLM to design the model itself. Until recently, this was thought to be something only humans could do.

但一个模型能够复现数据,并不意味着它能提供洞见。正如餐厅账单或 Hodgkin-Huxley 参数的情形一样,单个模型内许多不同的参数组合都能给出相同的输出。在更高一层、即模型本身这一层,情况同样如此:许多不同的**模型**,其方程形式各异,却能同样好地解释同一批数据。让一个智能体对同一批数据运行两次,它可能返回两个不同但同样优秀的模型。只返回其中一个的智能体,是在隐藏其余的模型,正如单一的最佳拟合数值组合所做的那样。

But the fact that a model replicates data doesn't mean that it provides insight. As with the restaurant bill, or the Hodgkin-Huxley parameters, many different combinations of parameters within a single model can provide the same output. The same is true one level up, for the model itself: many different _models_, with differently shaped equations, can explain the same data equally well. Run an agent twice on the same data, and it might return two different, and equally good, models. An agent that returns only one of them is hiding the rest, just as a single best-fit set of numbers does.

因此,更好的做法是把构建模型与约束模型同等对待:你要求得到所有可能合理地解释数据的模型范围。2 AI 系统提出许多候选模型。然后每个候选模型都按其受数据支持的程度被赋予权重,而那些需要很大灵活性才能拟合的模型,相比那些自然拟合的模型会得到更少的认可。弱候选模型被淘汰,AI 系统则针对更强的模型提出新的变体。

So it is better to treat building a model the same way we treat constraining one: you ask for the whole range of models that could plausibly explain the data.2 The AI system proposes many candidates. Each is then weighted by how much it is supported by the data, and the ones that need a lot of flexibility to fit are given less credit than those that fit naturally. The weak candidates drop out, and the AI system proposes new variations on the stronger ones.

因此,AI 智能体能够给出一个由多个合理模型构成的分布,而非单一的最佳拟合,而这个分布能告诉你单一模型无法揭示的信息。如果所有存活下来的模型都共享某个特征,那么数据很可能支持该特征。如果存活模型之间存在分歧,那就表明证据薄弱之处,并为研究者指明下一次实验的方向。

AI agents can thus provide a distribution of plausible models, rather than a single best fit, and that distribution tells you things a single model can't. If every surviving model shares a feature, the data likely favors it. If the survivors disagree, that shows where the evidence is thin, and points researchers in the direction of the next experiment.

我们不应操之过急。迄今为止,这些系统大多是在重新发现科学家已知的模型这一任务上接受测试。关于它们能在**没有**已知答案的问题上奏效的证据还很稀少。尽管如此,较新的 LLM 在研究任务上的表现优于较旧的模型,我们预计这些系统会持续改进。

We shouldn't get too far ahead of ourselves. So far, these systems have mostly been tested on rediscovering models that scientists already know. Evidence that they work on problems with _no_ known answer is sparse. Still, newer LLMs perform better at research tasks than older ones, and we expect these systems to keep improving.

在翻译中迷失,或如何解释模型 Lost in translation, or how to interpret models

一旦模型被构建并受到约束,它仍然需要被解释。它告诉我们关于所研究的真实系统的什么信息?Hodgkin 和 Huxley 的四粒子假说之所以产生,是因为如果四个独立的事物都必须发生才能打开一个门控,那么涉及四次方的方程正是你所期望的。因此他们提出,这可能就是膜内部正在发生的事情。如上所述,他们最终在很大程度上是正确的。

Once a model has been built and constrained, it still has to be interpreted. What does it tell us about the real system being studied? Hodgkin and Huxley's four-particle hypothesis came about because equations involving the fourth power are what you would expect if four separate things all had to happen before a gate could open. So they suggested that this might be what was happening inside the membrane. As we saw above, they turned out to be largely right.

这种解释之所以有分量,只是因为他们排除了其他可能性。如果一个更简单的模型同样有效,就没有理由认为门控有多个运动部件。正是替代方案的失败才使他们能够提出这一主张。关于确切数字存在一些不确定性:Hodgkin 和 Huxley 指出需要三次方或四次方,并最终确定为四次方。他们写道,更高次方甚至可能拟合得稍好一些,但潜在的“改进被认为不值得增加额外的复杂性”。

That interpretation only carried weight because of what they had ruled out. If a simpler model had worked just as well, there would have been no reason to think the gate had several moving parts. It was the failure of the alternatives that allowed them to make that claim. There was some uncertainty about the exact number: Hodgkin and Huxley noted that a third or fourth power was needed and settled on four. They wrote that higher powers might even have fit slightly better, but that the potential "improvement was not considered to be worth the additional complication."

这些方程还让 Hodgkin 和 Huxley 推导出其他几个推论。例如,方程解释了为什么神经细胞有一个电压阈值,低于它不放电,高于它则放电,以及为什么神经脉冲都具有相同的形状。他们还预测,上升足够缓慢的电流永远不会使细胞放电。科学哲学家称一个理论是“可理解的”,如果它让你无需运行就能推导出它的后果。这个测试适用于变量数量不多的模型,如 Hodgkin-Huxley 模型。对于可能有数千个变量的更大模型,等效的测试可能是我们能否将其归结为几个足以重建它的关键机制思想。模型越大,这就越难。

The equations also let Hodgkin and Huxley reason out several other implications. For instance, the equations explained why nerve cells have a voltage threshold below which they don't fire and above which they do, and why nerve pulses all have the same shape. They also predicted that a current rising slowly enough would never make the cell fire. Philosophers of science call a theory "intelligible" if it lets you work out its consequences without having to run it. That test works for a model with a handful of variables, like Hodgkin-Huxley. For a larger model with perhaps thousands of variables, the equivalent might be whether we can boil it down to a few crucial mechanistic ideas that are enough to rebuild it. The larger the model, the harder that becomes.

在这三个步骤中,解释是最难交给 AI 的。目标更难确定:我们可以衡量模型对数据的拟合程度,但无法衡量它是否导致对系统更好的理解,而且不清楚如何让 AI 系统将“理解”作为其目标。一个提议是通过观察 AI 系统在扰动系统时预测系统行为的能力来判断其理解程度。另一个提议是完全回避这个问题,转而询问智能体的输出是否帮助人类科学家更好地理解问题。一个例子是用于设计量子实验的 AI 系统。它不是提供一个单一的设置,而是被训练编写一个简短的程序,生成一系列相关实验的设置。物理学家然后可以直接从程序中读出底层设计原理,而不是从一个实验设置中逆向工程出来。

Of the three steps, interpretation is the hardest to hand over to AI. The goal is harder to pin down: we can measure how well a model fits the data, but not whether it leads to a better understanding of the system, and it's not clear how to tell an AI system to make "understanding" its objective. One proposal is to judge an AI system's understanding by seeing how well it predicts what a system does when you disturb that system. Another is to sidestep the question altogether, and instead ask whether the agent's outputs help human scientists better understand the problem. One example is an AI system for designing quantum experiments. Rather than providing a single setup, it was trained to write a short program that generates the setups for a whole family of related experiments. A physicist can then read the underlying design principle straight off the program instead of reverse-engineering it from one experimental setup.

在两种方法中,能够访问合适模型的完整分布,而不是单个最佳拟合模型,对于解释至关重要。单个模型无法告诉我们它的哪些部分实际上受到数据的约束,因此很容易将意义解读到不受约束的部分。在 Hodgkin-Huxley 模型中,正如我们所见,关于四个粒子的假说只有在你知道较低次方效果不佳时才有意义。一系列模型也使预测更加稳健。假设你问你的模型当你扰动一个系统时会发生什么——例如,通过阻断细胞膜上全部或部分钠门控。如果所有拟合数据的模型都返回相同的答案,正如当每个门控都被阻断时你所期望的那样,那么你可以相当自信。如果它们不一致,正如当只有部分门控被阻断时可能发生的那样,这可以为下一个实验提供思路。

In both approaches, having access to a full distribution of suitable models, rather than a single best-fitting one, is crucial for interpretation. A single model cannot tell us which of its parts are actually constrained by the data, so it's easy to read meaning into parts that are not. In Hodgkin-Huxley, as we've seen, the hypothesis about four particles only makes sense if you know that lower powers don't work as well. A range of models also makes predictions more robust. Say you ask your models what happens when you disturb a system—for instance, by blocking all, or only some, of the sodium gates on the cell membrane. If all the models that fit the data return the same answer, as you would expect when every gate is blocked, then you can be fairly confident in it. If they disagree, as they might when only some gates are blocked, that can provide ideas for the next experiment.

细胞与尺度 Cells and scales

大多数著名模型,包括 Hodgkin-Huxley 模型,都是紧凑的。它们由少量变量和方程组成,小到足以让人类科学家写下并进行推理。这并非巧合:它们必须紧凑,因为人们需要构建、约束和解释它们。它们的微小性也正是它们易于理解的原因。但随着 AI 承担更多工作,规模限制将开始解除。我们可以开始对过于复杂而无法手写建模的系统进行建模,即使我们不再能仅通过阅读方程来理解它们。

Most famous models, including Hodgkin-Huxley, are compact. They consist of a handful of variables and equations, small enough for a human scientist to write down and reason about. That's no coincidence: they needed to be compact, because people had to build, constrain, and interpret them. Their smallness is also what made them intelligible. But as AI takes over more of the work, the size limit will start to lift. We can begin to model systems that are far too complex to write down by hand, even if we can no longer understand them just by reading the equations.

大脑是一个明显的用例。Hodgkin-Huxley 方程描述单个神经细胞。但果蝇的大脑包含超过 100,000 个神经元,而人类大脑包含数百亿个。正是这些神经元之间的相互作用产生了行为和思维。对于果蝇,我们现在拥有前几代科学家无法想象的东西:其大脑的完整接线图,显示每个神经元以及它如何与其他神经元连接。研究人员已开始在此基础上构建大规模模型,包括一个预测果蝇视觉系统神经元活动的模型。这开辟了以前没有人能合理提出的问题。整个大脑如何将其接线转化为行为?仅从接线可以预测动物行为的多少?哪些连接真正重要?

Brains are an obvious use case. The Hodgkin-Huxley equations describe a single nerve cell. But a fruit fly's brain contains more than 100,000, and a human brain tens of billions. And it is the interplay between those neurons that gives rise to behavior and thought. For the fruit fly, we now have something that was impossible for previous generations of scientists: a complete wiring diagram of its brain, showing every neuron and how it connects to the others. Researchers have started building large-scale models on top of it, including one that predicts the activity of neurons across the fly's visual system. That opens up questions nobody could sensibly ask before. How does a whole brain turn its wiring into behavior? How much of an animal's behavior can be predicted from its wiring alone? And which connections actually matter?

但在这些尺度上,不可能找到单一正确的模型。事实上,在大多数情况下不存在这样的模型。正如统计学家 George Box 所写:“所有模型都是错误的,但有些是有用的。”数据总是与许多模型兼容。问题将不再是哪个模型是正确的,而是所有数据兼容模型共享哪些属性,以及可以从它们推导出哪些预测。最有用的实验将是那些排除最多剩余模型的实验,即使它们并不总是人类科学家认为最有趣的实验。

But at these scales it will not be possible to find a single correct model. Indeed, in most cases there is no such thing. As the statistician George Box wrote, "All models are wrong but some are useful." The data will always be compatible with many models. The question will no longer be which model is the right one, but which properties all data-compatible models share, and which predictions can be derived from them. The most useful experiments will be those that rule out the greatest number of remaining models, even if they are not always the ones a human scientist finds the most interesting.

这并非神经科学独有。气候科学家已经这样工作:他们构建许多巨大的数值模拟,每个基于略有不同的假设,并将他们的预测报告为所有模拟的分布范围。大型强子对撞机的粒子物理学家以类似的精神工作:当他们寻找新粒子时,他们不仅宣布某个特定理论是否有效;他们生成一张图,显示哪些版本的理论被数据排除,哪些仍然开放。

This isn't unique to neuroscience. Climate scientists already work like this: they build many gigantic numerical simulations, each based on slightly different assumptions, and report their projections as the spread across all of them. Particle physicists at the Large Hadron Collider work in a similar spirit: when they look for a new particle, they don't just declare whether a particular theory worked or not; they produce a map of which versions of the theory are ruled out by the data and which remain open.

Hodgkin 和 Huxley 所做的艰苦工作——构建模型、目测拟合每条曲线、尝试不合适的幂次,以及 Huxley 花三周时间手摇机械计算器——不是我们想回到的。但这并没有白费。模型提供的许多洞见来自创建它的过程;Hodgkin 和 Huxley 理解模型告诉他们的内容,因为他们自己构建了它,并且知道什么没有起作用。这种理解让他们形成了对门控变量的解释——这是最终为他们赢得诺贝尔奖的发现的一部分。

The painstaking work Hodgkin and Huxley did—building the model, fitting each curve by eye, trying the powers that didn't fit, and, in Huxley's case, spending three weeks hand-cranking the mechanical calculator—is not something we want to go back to. But it was not wasted. Much of the insight that a model provides comes from the process of creating it; Hodgkin and Huxley understood what the model was telling them because they had built it themselves, and they knew what hadn't worked. This understanding let them formulate their interpretation of gating variables—part of the discovery that ultimately earned them the Nobel Prize.

如今,一个 AI 系统有可能在一个下午就完成其中大部分同样的工作。但要提供理解,它必须交付那两位科学家缓慢工作所产生的东西:不仅是拟合数据的模型,还要说明哪些部分已被数据锁定、哪些仍悬而未决、途中哪些替代方案失败了,以及在此过程中发现了哪些新的谜题。一个科学模型或许能给出答案,但真正提供洞见的是构建它的过程。

An AI system could now plausibly do much of the same work in an afternoon. But to offer understanding, it needs to deliver what the two scientists' slow work produced: not just the model that fits, but an account of which parts are pinned down by the data, which are still open, which alternatives failed along the way, and what new puzzles were uncovered in the process. A scientific model may provide an answer, but it is the process of constructing it that provides the insight.

互动版:针对本篇提问 →

接着读 · Read next