JMIR Publications

ABSTRACT

Artificial intelligence (AI) is rapidly transforming various sectors, with profound implications for medicine. As AI's integration into medicine accelerates, it is crucial for clinicians to develop a foundational understanding of its core concepts and applications. This paper is a tutorial to modern AI, primarily focusing on machine learning and deep learning. Fundamental definitions of terms are defined and explained, the architecture of neural networks is discussed, and an overview of how the training process occurs is given. A brief overview of rising important topics is also covered including safety, explainability and artificial general intelligence. Artificial intelligence is a broad category including machine learning and expert systems. Expert systems was a traditional paradigm where expert knowledge was programmed into a system which has mostly been abandoned. Deep learning is particularly powerful form of machine learning requiring large neural networks. In machine learning, data is given to a algorithm and the algorithm learns relationships for itself. Various different approaches can be taken including supervised learning, unsupervised learning, and reinforcement learning. Supervised learning uses labelled data, unsupervised learning uses data without labels, and reinforcement learning relies on the machine receiving rewards and punishments. The process by which a machine learns involves attempting to minimize the loss function through back propagation and gradient descent. A foundational understanding of AI's principles, methodologies, and inherent challenges is essential for clinicians.