Inside Java Newscast #72 // nipafx

Java's reputation for being bad at AI is challenged by breaking AI development into three categories: training ML models (where Python dominates), building AI-centered products (where Java is competitive), and integrating AI features into existing projects (where Java already excels). For the third category—likely the most important long-term—Java's strengths in typing, performance, security, and ecosystem make it a strong choice. Libraries like TornadoVM, ONNX Runtime, DJL, and LangChain4j already provide solid ML execution support. Upcoming OpenJDK projects—Valhalla (value types, potential operator overloading), Panama (vector API, FFM), and Babylon (GPU code generation via code reflection)—will further strengthen Java's AI capabilities. The 'AI in Java is bad' narrative is myopic, applying mainly to model training, while the growing reality is that AI will mostly be a feature integrated into larger applications, a space where Java thrives.