A team of scientists has created a new AI method that addresses complex problems across science and engineering by reducing them to simpler mathematical equations.

Unlike typical black-box AI models, this approach focuses on interpretable representations that can be expressed in basic symbolic forms, aiding understanding and trust in AI-generated solutions.

The research demonstrates that this symbolic reasoning capability allows AI to uncover underlying structure in tasks such as physics simulations, optimisation challenges and system modelling, potentially boosting both accuracy and generalisation.

Researchers argue that breaking problems down into fundamental components not only enhances performance but also makes AI outputs more understandable to human experts.

By combining machine learning with classical mathematical reasoning, the work points toward a hybrid paradigm in which AI augments human insight rather than merely approximating outcomes. Such methods could accelerate scientific discovery in fields where complexity has traditionally limited the effectiveness of computational approaches.