AI is transforming science more profoundly than we think
Artificial intelligence is more than just a useful tool for scientists, writes former university president Emile Aarts. AI systems have been contributing to the solution of scientific problems for years. ‘The question is not whether AI will replace scientists, but how science itself is changing.’

In his opinion piece AI Has Not Yet Taken Over Scientific Research (10 November 2026), Hans Van Eyghen argues that the impact of artificial intelligence on scientific research has so far been limited. AI is useful for finding information, summarising literature and improving texts, but creative thinking and scientific reasoning, he maintains, remain the domain of humans. He supports his conclusion, among other things, by pointing to the limited number of AI-generated papers accepted by academic journals.
Van Eyghen is asking the wrong question. The scientific significance of AI should not be measured by the number of papers written by AI and subsequently published. The fundamental question is whether AI contributes to new scientific knowledge, the solution of fundamental problems and discoveries that would have been more difficult, slower or perhaps even impossible to achieve without AI. The answer by now is unequivocal: yes.
Take mathematics, for example. In 2024, the AI systems AlphaProof and AlphaGeometry 2 achieved a level of performance comparable to that of a silver medallist at the International Mathematical Olympiad. In 2025, an advanced version of Gemini Deep Think even reached gold-medal level, solving five of the six problems correctly.
Even more interesting is AlphaEvolve, which has improved existing mathematical solutions and discovered a new algorithm for matrix multiplication. This improved on a result that had stood as the best-known solution for nearly fifty years. This is not text generation, but a contribution to the development of mathematical knowledge and methods.
A similarly fundamental transformation is taking place in biology. AlphaFold has revolutionised the prediction of protein structures and was awarded the Nobel Prize in Chemistry in 2024. In 2025, AlphaGenome was introduced, an AI model that analyses large sections of the human genome and predicts how genetic variations affect the functioning and regulation of genes. This gives biologists new opportunities to investigate the relationship between genetic information, cells and diseases.
Comparable breakthroughs are being reported in economics and econometrics. In 2022, researchers developed The AI Economist, a reinforcement learning-based system capable of independently designing and evaluating tax systems. In simulated dynamic economies, the system identified policy solutions that improved on existing models in terms of social welfare and the trade-off between equality and productivity.
In econometrics, the development of Double Machine Learning has created new opportunities to investigate causal relationships and policy effects in complex datasets containing large numbers of variables. And in 2026, researchers writing in the Annual Review of Economics demonstrate how large language models can fundamentally expand empirical economic research by systematically analysing information from millions of documents and texts.
These examples make it clear that we need to distinguish between two developments: AI as a tool for conducting scientific research and AI as a partner in the process of scientific discovery. It is the second development in particular that is of fundamental significance. AI systems can recognise patterns that are difficult for humans to detect, investigate large numbers of hypotheses, perform complex calculations and find solutions that researchers can subsequently test and develop further.
This does not mean that scientists are becoming redundant. Quite the opposite. Scientific creativity, critical reflection, interpretation, verification and accountability remain essential. AI can make mistakes, misinterpret results and produce convincing-looking nonsense. But that is no reason to downplay the scientific significance of AI. It is a reason to ensure that collaboration between humans and machines is carefully managed.
The question, therefore, is not whether AI will replace scientists. The question is how science will change when human and artificial intelligence complement and reinforce one another. Anyone who assesses the impact of AI solely on the basis of scientific papers written by AI is looking at the wrong indicator. The real revolution is taking place in the laboratory, behind the computer and in the way we think about scientific problems.
AI may not be taking over science, but it is changing the way science is conducted and practised. And that is a far more profound development.
Emile Aarts was Rector Magnificus of Tilburg University and Professor of AI and Data Science. He previously held positions at Eindhoven University of Technology and Philips Research.