Among the native Java AI frameworks cited by Oracle‘s Smith are Tribuo, LangChain4j, and CoreNLP. Tribuo is a machine learning library written in Java that provides tools for classification, regression, clustering, model development, and other capabilities. LangChain4j is a Java version of the LangChain framework for building applications powered by large language models (LLM); its goal is to simplify the integration of LLMs into Java applications. And CoreNLP offers a suite of tools for doing natural language processing in Java.
Oracle’s own ambitions for AI in Java call for integrating AI services with business logic via Project Panama, which is the OpenJDK project aimed at interconnecting the JVM and native code, and GraalPy, which is an embeddable, high-performance Python 3 runtime for Java. “We expect to see more integration support over time, just as we have seen Java expand into new technologies over the last 30 years,” Smith said. “Note that innovation in Java projects like Valhalla, Babylon, and Panama help Java run even closer to the native compute that has become synonymous with GenAI.”
IDC’s Dayaratna believes it is “eminently possible” that Java will supersede Python for machine learning development. “Java is widely considered more performant and faster than Python,” Dayaratna said. “As organizations begin to leverage generative AI, in particular for more production-grade use cases, Java is likely to increasingly gain traction because of its advantages with respect to resource consumption, application performance, speed of execution, and security.”