GCRI · 人工知能と自律システム

科学における AI

Machine learning now predicts protein structures, emulates weather and climate models, screens materials and helps analyse vast datasets. It can speed discovery, and it can also produce results that are hard to reproduce, overfit or rest on data leakage. This page sets out how AI is used in science, where it fails, how its results are checked, and what evidence serves the researchers, funders and journals involved.

02質問

AI in science: how the risk develops.

Discovery and reproducibility. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.

メカニズム

Learning from data replaces some simulation

Models trained on simulations or observations can emulate expensive calculations quickly. They are reliable within the range of their training data and uncertain outside it.

メカニズム

Leakage inflates results

When information from the test set reaches training, through duplicates, time overlap or preprocessing, reported accuracy overstates real performance. Leakage is a known cause of irreproducible ML science.

メカニズム

Openness enables checking

Results can be verified only when code, data, models and settings are shared. Proprietary models and data limit independent checks.

メカニズム

Scientists stay accountable

AI can suggest hypotheses and draft text, but researchers remain responsible for methods, interpretation and integrity. Journals and funders are setting rules for AI use.

03監視

モデルレジストリ。

このテーマを追跡するためにチームが使う画面:その指標、状態、決定。名前は例であり、値はイメージです。

モデル・レジストリ

AI in science · Model registry · Release gate

1 つのモデルがゲートで阻止されたGCRI・人工知能と自律システム

モデル

M-489 triage assistant公共サービス高20 of 24 suites阻害中
M-530 flood nowcast科学中12 / 12公開済み
M-079 document search内部低8 of 8公開済み
M-881 agent plannerエージェント型高11 of 16評価中

モデルカード

想定用途
トリアージ支援、判断は人が行う
対象外
適格性の決定
既知の限界
まれな事例での精度が低い
オーナー
モデルの所有者

インシデントログ

  1. D-19INC要約に実在しない引用が含まれる
  2. D-8MIT検索拡張による根拠づけを追加
  3. D-3INCアップロードされたファイルによるプロンプトインジェクション
  4. D-1MITツール権限を縮小

Evaluation scores

vs release threshold

Task accuracy81 of 85 required頑健性68 of 75 requiredFairness gap (lower is better)3 of 5 requiredRefusal on misuse85 of 95 required
Model registry: the screen’s structure, terms and states for AI in science.例示の名称と説明用の値です。設計であり、記録でも稼働中のシステムでもありません。
04分野を横断して

AI changes how each domain’s science is done.

05手法とエビデンス

How AI in science is modelled, measured and tested.

モデル、その仮定、データ、不確実性を、科学者やエンジニアが確認する形で示します。構造は公表された手法に従っています。描かれた値は例示であり、知見ではありません。

モデル、方程式、仮定

The model structure used for AI in science, the assumptions it makes and how each is tested.

要素 i のリスクRi = f(Hi, Ei, Vi, Ci)ハザード、エクスポージャー、脆弱性、対処能力
年間期待損失EAL = ∑k pk · Lk年間確率 pk
結合システムxt+1 = A xt + B ut + εtA carries the couplings, u the shock, ε the error
仮定作られる理由検証の方法
一つの事象の中では、ハザードとエクスポージャーは独立している損失の積分を扱いやすく保つ同時記録に当てはめたコピュラと比較
結合は動作点の近くで線形であるショックを A を通じて伝播させる線形範囲を超えて摂動を与えて比較する
脆弱性曲線は場所を越えて適用できる地域固有の曲線はめったに得られない観測された損失がある場合はそれに照らして検証
入力
来歴付きのバージョン管理されたデータセット
成果物
分布であり、単一の数値ではない
コード
モデルカードと試験とともに公開
再現
同じデータ、コード、環境、シード
標準的なリスクおよび状態空間の定式化です。挙げた前提はこの手法に典型的なもので、審査のために示しており、結果ではありません。
FAIR 原則・GO FAIR と FORCE11W3C PROV・World Wide Web Consortium不確実性の一貫した扱いに関するガイダンス・IPCCCEOS 解析即応データ・地球観測衛星委員会
06指標

何を、どの程度うまく測定しているか。

各指標には、その単位、更新頻度、背後にある出典の種類、値が持つことになるエビデンスのラベルが付いています。値は表示しません:プログラムが自らの権限のもとで値を提供します。

指標単位周期出典の種類ラベル
Studies sharing code and datapercent of published ML studiesレビューごとPublication audits観測
Leakage issues identified in reviewscount by typeレビューごとReproducibility studies報告値
Out-of-distribution performanceerror on held-out regimesper evaluationモデル評価観測
Emulator error against reference simulationerror metricper evaluationValidation runs観測
Journals with AI use policies件数年次Publisher policies申告済み
Compute used for scientific model trainingfloating point operationsper modelModel documentation推定
07出典

エビデンスの出所。

名前の明示された公開データセットと機関の記録。それぞれにライセンス区分があります。合意に基づく出典は、その条件のもとでのみ使用し、再公開することはありません。

出典提供するものライセンス区分
Protein Data Bank and AlphaFold Protein Structure DatabaseExperimental and predicted protein structuresOpen, CC0 and CC BY 4.0 respectively
WeatherBench and related benchmarksStandard evaluation of data-driven weather modelsOpen, licence per benchmark (to check)
Princeton reproducibility studies on ML-based scienceSurvey of leakage and reproducibility failuresLicence per publication
Materials ProjectComputed materials properties for screeningオープン、CC BY 4.0
Publisher and funder policies on AI useRules on authorship, disclosure and reviewPublic, terms per publisher
08決定

誰が決め、エビデンスが何に役立つか。

人工知能と自律 システム prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.

誰が決定するか質問このページが提供するエビデンス
Research teamsWhether an ML result is robust enough to publishLeakage checklists; validation protocols
Journals and reviewersWhat evidence to require for ML-based claimsReporting guidance; reproducibility evidence
研究資金提供者How to fund open models, data and compute for scienceOpenness and compute evidence

このトピックが生み出す成果物

  • Leakage and reproducibility checklists for ML studies
  • Validation protocols for scientific emulators
  • Guidance on reporting AI use in research
  • Case studies of AI in specific sciences
09中核となる問い

What scientists ask about AI in research.

  • Does the AI-assisted result hold when tested independently, by experiment or observation?
  • Is the model, data and code available so others can reproduce it?
  • Has data leakage between training and evaluation inflated reported performance?
  • Does the model extrapolate beyond the conditions its training data covered?
10メカニズム

How discovery and error both scale.

Machine learning finds patterns in large datasets faster than people can, and surrogate models can emulate expensive simulations, allowing many more scenarios to be explored. In fields with good data and clear validation, such as structural biology, the gains have been substantial.

The risks are equally systematic. Leakage between training and test data has produced inflated results across many disciplines. Models trained on past data may not respect physical laws when extrapolating. Generative システム can fabricate plausible references or data. And when models and data are proprietary, results cannot be checked.

11手法とデータ

Keeping AI-assisted science reproducible.

  1. 01

    Reporting standards

    Checklists covering data splits, leakage checks, baselines and uncertainty. Standards help only if journals and reviewers enforce them.

  2. 02

    Physics-informed and hybrid models

    Combining learned components with known physical constraints, using 人工知能と機械学習 and high-performance, cloud and edge computing. Constraints reduce but do not remove extrapolation risk.

  3. 03

    Uncertainty estimation

    Ensembles and Bayesian methods from 確率と統計 to flag predictions the model is unsure of. Uncertainty estimates can themselves be miscalibrated.

  4. 04

    Open artefacts

    Publishing code, data and model weights where rights allow, so others can reproduce and challenge results.

13Where this page’s references

Where this page’s references lead.

このページの読解の基準となる機関と文書。状態と版は変わります。各リンクは発行機関へ移動します。

14質問

質問と回答。

What is data leakage in machine learning?
The unintended use of information from the test data during training or model selection, which makes performance look better than it would be on genuinely new data.
Can AI replace scientific simulation?
AI emulators can approximate simulations quickly within their training range, but they need validation and can fail for conditions they have not seen.
How should AI use in research be reported?
By stating which tools were used, for what, with what settings, and how outputs were checked, following the policies of the journal and funder.
Can AI replace experiments?
No. It can prioritise which experiments to run and emulate some simulations, but claims about the world still need testing against observation or experiment.
What is data leakage?
When information from the test set, or from the future, finds its way into training, so the model appears better than it is. It is one of the most common errors in applied machine learning.
Does the platform review scientific papers?
No. It studies methods and publishes guidance and case studies. Peer review remains with journals and the scientific community.
Are AI-generated hypotheses credited?
The platform’s outputs describe the role AI played in any work, so readers can judge the result; authorship and accountability stay with people.
報告書貢献する

このページからエビデンスを提供する。

報告書 is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.

GCRI人工知能と自律システム指数関数的テクノロジーLabs のエビデンス
人工知能と自律システム

手法、データセット、または問いを持ち込んでください。

情報提供と問い合わせは本日から受け付けています。問い合わせには参照番号が付き、次の手順が示されます。申請、支払い、計算資源の割り当て、登録は、担当チーム、規約、審査体制が整ってから開始します。

The Global Centre for Risk and Innovation

Science · evidence · measurement

Not for profit

Our articles, maps and twins are free to read and free to republish. Readers and partners keep it that way.

Partner with us

Who funds us

GCRI is a not-for-profit institution. To ask who funds a piece of work, use the funding questions form.

Independence

Editors and the expert reviewers named on each article decide what we publish. Funders and partners do not review or approve articles before publication.

Licence

Articles: CC BY 4.0. Figures, maps, twins and data: CC BY 4.0. Other uses: ask through the permissions form.