Scientists in the biophotonics community have long recognized the value of artificial intelligence (AI) and machine learning in diagnostics, ever since the creation of a computer-assisted detection model to highlight areas of concern in images and the FDA’s subsequent approval of this approach for mammography in the 1990s. Software programs have since catapulted radiomics — the extraction of a large number of features from images and the interpretation of that information — into the fore of medical practice. Today, it is used not only for data extraction but also for noise reduction and reconstruction of the images themselves, for both research and clinical use. Simultaneously, this has generated a need to harness this potential in a responsible and productive way. The explosion of AI in imaging is undeniable, which is reflected in literature and the marketplace. Driven by the demand for large data sets as well as government initiatives to promote these technologies, Grand View Research reports that AI in the medical imaging market stood at $753.9 million last year and is projected to grow at a compound annual growth rate of 34.8% through 2030. This reality has academia and industry researchers grappling with a host of questions related to the need to address inherent biases in the quality of the data introduced into imaging algorithms, and to allow for accountability when mistakes are uncovered. Part of the solution may be to program these algorithms in a manner similar to human learning; a study recently published by a team at UCLA showed that a neural network could make accurate diagnostic judgements when it was provided with foundational concepts in physics. Other authors have pointed out that no amount of computer learning will ever replace the need to expand human knowledge, so AI must be integrated alongside other tracks in education and practice. In our last edition, Biopinion authors Karen Drukker and Maryellen Giger articulate that to obtain reliable AI-generated data, information needs to be acquired from as diverse a group of potential patients as possible. And in this issue’s Biopinion here, Lana Feng writes that according to a recent survey, a great number of medical professionals in both research and industry see value in what AI can bring to R&D and diagnostics, particularly when it comes to medical imaging. She notes that this does not detract from the importance of human expertise in proper clinical evaluation, because knowledge of disease and therapeutics is required to provide the right input in the initial formation of algorithms, and to draw the correct conclusions on which to base a treatment regimen. The infrastructure of compatible hardware and software needed for AI will continue to evolve and mature in the years to come. There is no doubt that various forms of AI and machine learning will continue to pervade our personal and professional lives, so the best course for those within the life sciences and biomedicine is to calibrate it properly — and use it wisely. Enjoy the issue!