Papers

The 2024 Physics and Chemistry Nobel Prizes to machine learning (ML) and artificial intelligence (AI) breakthroughs marked “Year 1 of AI for Science,” underscoring their transformative role in physical sciences. Yet data are not the same as understanding—a distinction central to chemistry, which has long relied on concepts such as bond, aromaticity, and reactivity as scaffolds for understanding and explanation. Building on our recent perspectives (ACS Phys. Chem. Au 2024, 4, 135–142; J. Chem. Theory Compt. 2025, DOI: 10.1021/acs.jctc.5c01299), this article explores how ML/AI can become engines of chemical understanding. We introduce a quintet of chemical knowledge—ontology, epistemology, theory, concept, and understanding—and develop the metaphors of the Knowledge Tree and Knowledge Forest to show how diverse epistemologies interact and recursively enrich one another. Case studies on aromaticity, catalysis, orbital-free density functional theory, and protein folding illustrate how ML features, when interpreted as conceptual roots, yield fruits of understanding. Contrasting multiscale modeling with hierarchical modeling, we argue that ML enables emergent, concept-driven integration across levels. Cultivating this plural and hierarchical ecosystem may guide theoretical chemistry toward its next breakthroughs, resolving Dirac’s dilemma not by brute force but by forests of concepts that transform data into enduring understanding.