AI in science: what does it really contribute?
AI can help find patterns, build models and suggest candidates for experiments.
What is it?
Scientists use AI to find patterns in large datasets, predict protein structures, run simulations and suggest candidates for experiments. AI speeds up research, but hypotheses still have to be tested by experiment.
An everyday example
The AlphaFold model predicted the structures of most known proteins, saving researchers months of laboratory work.
Why does it matter?
AI can speed up discoveries in medicine, climate science and materials. It also brings the risk of errors and results that cannot be reproduced.
A common misconception
That a model’s prediction is already a confirmed discovery. A scientific hypothesis is not a verified fact.
In more detail
Where is it useful?
In research on proteins, materials, astronomical data, climate models and lab automation.
Can AI prove a new theory by itself?
Not usually. Science requires methods, reproducibility, independent checking and explanation.
Is a model the same as an experiment?
No. A simulation or a prediction can suggest a direction, but a real-world test often remains essential.
Why does this matter?
Because a good tool can shorten part of the research, but it can also, very convincingly, steer a hypothesis in the wrong direction.
Remember: A scientific hypothesis is not a confirmed fact.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.