Delivering healthcare that meets the needs of a growing and ageing population is a defining scientific and social challenge for the twenty-first century. Breakthroughs in data-gathering technologies have rapidly increased our capacity to collect genetic, epigenetic and phenotypic data at scale, powering large population-level studies of the molecular bases of several important diseases. At the same time, a systematic effort of digitisation has made clinical and health record data increasingly available across life-spans and populations. Biomedicine is a data-rich science, yet we lack a practical framework that robustly and effectively integrates such data to elucidate disease mechanisms across many disparate diseases and data types, to enable the efficient discovery and evaluation of novel strategies for disease prevention and treatment, and to optimise healthcare delivery. Artificial Intelligence (AI) techniques have the potential to contribute significantly to health research and healthcare delivery, due to their ability to effectively capture patterns in large, high-dimensional data sets. Nevertheless, application of current AI techniques to the biomedical domain has lagged behind other areas, because this field faces specific and complex challenges requiring bespoke solutions. Biomedical applications require interpretable models, leveraging data to identify mechanisms and consequently possible interventions. They also need a framework that allows the incorporation of prior knowledge resulting from cutting-edge biomedical research, and quantification of uncertainty to enable rational, risk-averse and informed decision-making based on the data. AI researchers in the biomedical field further need a keen awareness of the social and legal dimensions of their research (including ethical, societal and regulatory issues), and cross-disciplinary communication skills which will enable the creation and application of solutions in collaboration with a broad range of scientific, societal and industrial stakeholders. Additionally, the new generation of researchers in the field will need to exercise leadership in public discourse, engaging with public concerns and expectations in order to facilitate the widespread acceptance of new technologies. These challenges call for a new, distinctive, trans-disciplinary research field: biomedical AI. Our core vision is that AI technologies will be a key driver in furthering our understanding and discovering actionable interventions in biomedicine. Modern data gathering technologies enable us to observe multiple aspects of disease processes at vastly different organisational scales. Large national and international initiatives such as The Cancer Genome Atlas, the Human Cell Atlas and Genomics England are detailing, at unprecedented scale and resolution, the variability of the key molecular drivers at the genetic and epigenetic level of several important diseases. Advanced microscopy and microfluidic tools are increasingly providing a fine-grained portrait of biological dynamics within a disease context, elucidating the stochastic mechanisms underpinning cell physiology. Population health initiatives are gathering increasing information on how such diversity and variability manifests within the broader population. Our scientific focus will be on delivering an AI framework which discovers interactions from molecular and clinical data, by elucidating statistical predictive patterns and by offering novel mechanistic insights. This will generate a unique biological understanding and suggest novel directions in translational research. This article was published on 2024-11-22