# AI w nauce

EN: **AI for science** · id `aiscience` · aliases: AI4Science, AI for Science, scientific machine learning, nauka wspierana przez AI

Path: [AI Atlas](https://aiatlasgraph.com/llms.txt) › AI w nauce

## Description

- PL: Zastosowania głębokiego uczenia do problemów naukowych w skali kwantowej, atomowej i ciągłej: biocząsteczki, chemia, materiały, klimat i równania fizyki.
- EN: Deep learning applied to scientific problems at quantum, atomistic and continuum scales: biomolecules, chemistry, materials, climate and physical equations.

## In one sentence

- PL: Wykorzystanie uczenia maszynowego do odkryć w biologii, chemii, fizyce i naukach o Ziemi.
- EN: Using machine learning to make discoveries in biology, chemistry, physics and the Earth sciences.

## Problem it solves

- PL: Eksperymenty i symulacje fizyczne trwają miesiącami i dużo kosztują, a możliwych cząsteczek czy materiałów jest więcej, niż da się sprawdzić.
- EN: Experiments and physical simulations take months and cost a lot, and there are more possible molecules or materials than anyone could ever test.

## What it does

- PL: Model uczony na dawnych pomiarach działa jak szybki zwiad: wskazuje najbardziej obiecujących kandydatów, a laboratorium sprawdza tylko ich.
- EN: A model trained on past measurements acts like a fast scout: it points to the most promising candidates, and the lab only tests those.

## Everyday example

- PL: Prognoza pogody na dziesięć dni policzona w minutę albo kształt nowego leku dopasowany na komputerze, zanim ktoś go zsyntetyzuje.
- EN: A ten-day weather forecast computed in a minute, or a new drug shaped on a computer before anyone synthesises it.

## This concerns you if you…

- PL: pracujesz w laboratorium lub przy symulacjach naukowych
- PL: chcesz wiedzieć, skąd biorą się nowe leki i materiały
- PL: śledzisz, jak AI zmienia medycynę i energetykę
- EN: work in a lab or with scientific simulations
- EN: want to know where new drugs and materials come from
- EN: follow how AI is changing medicine and energy

## Keywords

AlphaFold · protein design · molecular property prediction · drug discovery · machine learning interatomic potentials · materials discovery · ML weather prediction · physics-informed neural networks · neural operators · equivariant neural networks · self-driving labs · AI scientist · DeepChem · OpenFold

## How it works

- PL: Modele uczą się na danych eksperymentalnych lub wynikach kosztownych symulacji (np. DFT, numeryczne prognozy pogody) i stają się szybkimi surogatami albo generatorami kandydatów. Kluczowe jest wbudowanie wiedzy fizycznej: symetrii, równań i oceny niepewności, a potem weryfikacja w laboratorium lub dokładną symulacją.
- EN: Models learn from experimental data or the outputs of expensive simulations (e.g. DFT, numerical weather prediction) and become fast surrogates or candidate generators. The key is building in physical knowledge — symmetries, equations and uncertainty estimates — and then verifying results in the lab or with accurate simulation.

## Limits

- PL: Dane naukowe są drogie, nierówne i obciążone metodą pomiaru. Modele słabo ekstrapolują poza zbadane obszary, a trafna predykcja nie zastępuje eksperymentu; walidacja laboratoryjna pozostaje wąskim gardłem.
- EN: Scientific data is expensive, uneven and biased by measurement methods. Models extrapolate poorly beyond explored regions, and an accurate prediction does not replace an experiment; lab validation remains the bottleneck.

## Key terms

- PL: **model surogatowy**: Szybki model zastępujący kosztowną symulację lub pomiar.
- PL: **DFT**: Metoda obliczeń kwantowych dająca energie i siły atomów; częste źródło danych treningowych.
- PL: **inductive bias**: Założenia wbudowane w model, np. symetrie fizyczne, ułatwiające naukę z małej ilości danych.
- PL: **aktywne uczenie**: Model sam wybiera, który kosztowny pomiar lub obliczenie wykonać następne.
- EN: **surrogate model**: A fast model standing in for an expensive simulation or measurement.
- EN: **DFT**: A quantum calculation method giving atomic energies and forces; a common training-data source.
- EN: **inductive bias**: Assumptions built into a model, such as physical symmetries, that ease learning from little data.
- EN: **active learning**: The model chooses which costly measurement or calculation to run next.

## Know first

- [Modele dyfuzyjne (Diffusion models)](https://aiatlasgraph.com/t/diffusion.md)
- [Enkoder, dekoder i enkoder–dekoder (BERT, GPT, T5) (Encoder-only, decoder-only and encoder–decoder models)](https://aiatlasgraph.com/t/transformerfamilies.md)

## Subtopics (7 direct, 36 in total)

- [Automatyzacja badań naukowych (AI-driven scientific discovery and research automation)](https://aiatlasgraph.com/t/discovery.md): Agentic LLM systems and optimisation loops automating the scientific cycle: hypothesis, plan, experiment (computational or robotic), analysis, report.
  - [Agenci generujący hipotezy i eksperymenty obliczeniowe (Hypothesis-generating research agents)](https://aiatlasgraph.com/t/hypothesisagents.md): Multi-agent LLM systems that generate, evaluate and evolve hypotheses and, in computational fields, also write code, run experiments and draft papers.
    - [Ideacja naukowa przez analogie (Analogy-based scientific ideation)](https://aiatlasgraph.com/t/ideascientist.md): An agent system decomposes ideation into gap finding, analogy-based innovation and proposal writing, trained with reinforcement learning.
  - [Laboratoria autonomiczne (self-driving labs) (Self-driving laboratories)](https://aiatlasgraph.com/t/selfdrivinglabs.md): A closed design–make–measure–analyse loop: robotics runs experiments and a model (usually Bayesian optimisation, increasingly an LLM agent) chooses the next ones.
  - [Modele świata badań nad AI (Research world models for AI experiments)](https://aiatlasgraph.com/t/researchworldmodels.md): Models that predict outcomes of AI-model experiments from configuration representations, used to select promising training and inference variants.
- [Białka i biologia molekularna (Proteins and molecular biology)](https://aiatlasgraph.com/t/biomolecules.md): Machine learning for biomolecules: predicting structures and interactions, designing proteins, and language models of protein and genome sequences.
  - [Modele bazowe pojedynczych komórek (Single-cell foundation models)](https://aiatlasgraph.com/t/singlecellfm.md): Transformers pretrained self-supervised on single-cell expression profiles, fine-tuned for annotation, integration and gene-network analysis.
  - [Modele językowe białek i genomów (Protein and genomic language models)](https://aiatlasgraph.com/t/bioseqmodels.md): Transformers and sequence models trained self-supervised on protein or DNA sequences; they yield embeddings, score mutations and generate new sequences.
  - [Projektowanie nowych białek (De novo protein design)](https://aiatlasgraph.com/t/proteindesign.md): Generating new protein backbones and sequences for a given function: target binders, symmetric assemblies and enzyme active sites.
  - [Przewidywanie struktury białek i kompleksów (Protein and biomolecular complex structure prediction)](https://aiatlasgraph.com/t/proteinstructure.md): Predicting the 3D structure of proteins from sequence and, in newer models, of whole complexes with DNA, RNA, small molecules and ions.
  - [Zero-shot ocena skutków mutacji w białkach (Zero-shot protein variant effect prediction with language models)](https://aiatlasgraph.com/t/varianteffect.md): Zero-shot scoring of amino-acid substitutions by comparing variant and wild-type likelihoods under a protein language model.
- [Chemia i odkrywanie leków (Chemistry and drug discovery)](https://aiatlasgraph.com/t/chemistry.md): Models for small molecules: property prediction, generating new structures, docking to proteins and synthesis planning.
  - [Dokowanie cząsteczek do białek (Molecular docking and binding prediction)](https://aiatlasgraph.com/t/docking.md): Predicting a ligand's binding pose in a protein and its binding strength; classically search with a scoring function, now also generative models and whole-complex prediction.
  - [Generowanie nowych cząsteczek (De novo molecule generation)](https://aiatlasgraph.com/t/moleculegen.md): Generative models (SMILES language models, graph models, 3D diffusion) and reward-driven optimisation for proposing new compounds, including directly inside a protein's binding pocket.
  - [Planowanie syntezy wstecznej (retrosynteza) (Retrosynthesis planning)](https://aiatlasgraph.com/t/retrosynthesis.md): Searching synthesis routes from a target molecule back to available starting materials: networks proposing reactions plus tree search (MCTS, A*).
  - [Przewidywanie właściwości cząsteczek (Molecular property prediction)](https://aiatlasgraph.com/t/molproperty.md): Regression and classification of physicochemical, biological and toxicity (ADMET) properties from a molecular representation.
- [Fizyka i astronomia (AI in physics and astronomy)](https://aiatlasgraph.com/t/physastro.md): ML in particle physics, astrophysics, cosmology and plasma physics: signal classification, fast simulation, inference and experiment control.
  - [Odsiewanie fałszywych alarmów w obserwacjach zmiennego nieba (Real–bogus classification of astronomical transients)](https://aiatlasgraph.com/t/transientrealbogus.md): Classifying difference-image cutouts from sky surveys (real source vs. artefact) with a convolutional network in the alert pipeline.
  - [Rozpoznawanie kształtów galaktyk na zdjęciach nieba (Galaxy morphology classification)](https://aiatlasgraph.com/t/galaxymorph.md): Classifying galaxy morphology with convolutional networks (Bayesian ensembles) trained on vote distributions from crowdsourcing volunteers.
  - [Rozpoznawanie strumieni cząstek ze zderzeń (jet tagging) (Jet tagging in particle physics)](https://aiatlasgraph.com/t/jettagging.md): Classifying hadronic jets (e.g. top, W, b, gluon) with permutation-invariant networks on particle clouds, graphs or transformers.
  - [Sterowanie plazmą w tokamaku przez uczenie ze wzmocnieniem (Reinforcement learning for tokamak plasma control)](https://aiatlasgraph.com/t/plasmacontrol.md): Deep RL trained in a plasma simulator that controls all of a tokamak's coils with one network, holding and shaping the plasma.
  - [Szybka symulacja detektorów modelami generatywnymi (Generative fast simulation of particle detectors)](https://aiatlasgraph.com/t/fastdetsim.md): Generative models (GANs, VAEs, flows, diffusion, flow matching) replacing calorimeter shower simulation, evaluated on shared challenges.
  - [Wykrywanie fal grawitacyjnych przez sieci neuronowe (Deep learning detection of gravitational waves)](https://aiatlasgraph.com/t/gwdetection.md): Convolutional networks on interferometer time series that detect coalescence signals and estimate their parameters faster than matched filtering.
  - [Wyszukiwanie planet pozasłonecznych w zmianach jasności gwiazd (Exoplanet transit detection with deep learning)](https://aiatlasgraph.com/t/exoplanetdetect.md): Classifying transit signals in light curves (planet vs. false positive) with 1D convolutional networks and ranking candidates.
- [Odkrywanie materiałów (Materials discovery)](https://aiatlasgraph.com/t/materials.md): Machine learning for inorganic materials: predicting crystal stability and properties, learned interatomic potentials and structure generation.
  - [Generowanie nowych kryształów (Generative crystal and materials design)](https://aiatlasgraph.com/t/crystalgen.md): Generative models of crystal structures (mainly diffusion-based) conditioned on chemical composition, symmetry and target properties.
  - [Uczone potencjały międzyatomowe (Machine-learned interatomic potentials)](https://aiatlasgraph.com/t/mlip.md): Regressing atomic energies and forces from local atomic environments, trained on DFT data; replaces classical force fields in molecular dynamics.
- [Pogoda i klimat (Weather and climate modelling)](https://aiatlasgraph.com/t/earthsystems.md): Machine learning in the Earth sciences: global medium-range forecasts, ensemble forecasts, nowcasting and climate-model emulation.
  - [Globalne prognozy pogody z sieci neuronowych (Data-driven global medium-range weather forecasting)](https://aiatlasgraph.com/t/mlweatherglobal.md): Deterministic models (GNNs, transformers) trained on reanalysis that autoregressively forecast the global atmospheric state 1–15 days ahead.
  - [Modele hybrydowe i emulatory klimatu (Hybrid and learned climate models (climate emulators))](https://aiatlasgraph.com/t/hybridclimate.md): Learned emulators of general circulation models and hybrids that pair a differentiable dynamics solver with networks parameterising sub-grid processes.
  - [Probabilistyczne prognozy pogody (zespoły) (Probabilistic ensemble weather forecasting)](https://aiatlasgraph.com/t/ensembleweather.md): Generating forecast ensembles with probabilistic (mainly diffusion) models instead of a single deterministic forecast.
  - [Prognozy opadów na najbliższe godziny z radaru (nowcasting) (Precipitation nowcasting from radar)](https://aiatlasgraph.com/t/nowcasting.md): Short-range (5–90 min, up to a few hours) precipitation forecasting from radar image sequences with generative or convolutional models.
  - [Zwiększanie szczegółowości prognoz i projekcji klimatu (downscaling) (Machine-learning downscaling of weather and climate data)](https://aiatlasgraph.com/t/climatedownscaling.md): Mapping coarse fields from global models or reanalysis to high resolution with regression and generative networks (e.g. residual-correcting diffusion).
- [Uczenie maszynowe z fizyką i symulacje (Physics-informed and scientific machine learning)](https://aiatlasgraph.com/t/sciml.md): Surrogate models and solving partial differential equations with neural networks, with physical knowledge built into the loss, architecture or data.
  - [Operatory neuronowe (FNO, DeepONet) (Neural operators (FNO, DeepONet))](https://aiatlasgraph.com/t/neuraloperators.md): Networks learning mappings between function spaces (e.g. initial condition → PDE solution), largely independent of mesh resolution.
  - [Sieci uwzględniające fizykę (PINN) (Physics-informed neural networks (PINN))](https://aiatlasgraph.com/t/pinn.md): A network approximating a PDE solution, trained with a loss combining data fit, the equation residual and boundary conditions.

## Related topics

Untyped, symmetric "see also" links.

- [AI w obserwacji Ziemi i geoinformatyce (Geospatial AI and Earth observation)](https://aiatlasgraph.com/t/geoai.md)

## Sources

- [Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems](<https://arxiv.org/abs/2307.08423>) `ai4science-survey-2023` · by Xuan Zhang, Limei Wang i in. (zespół kilkudziesięciu autorów, m.in. Shuiwang Ji)
- [Computational protein design and protein structure prediction win Nobel Prize in Chemistry (EMBL)](<https://www.embl.org/news/science-technology/alphafold-wins-nobel-prize-chemistry-2024/>) `nobel-chem-2024-embl` · by EMBL

---

AI Atlas revision 122 (2026-10-30) · HTML: https://aiatlasgraph.com/t/aiscience · index: https://aiatlasgraph.com/llms.txt · everything: https://aiatlasgraph.com/llms-full.txt · updated 2026-10-30