Java for AI
summary
Java now exposes concrete tooling for AI: the Foreign Function and Memory API plus jextract bind to BLIS for zero‑copy off‑heap matrices; the Vector API provides explicit SIMD kernels; Valhalla value types enable float16-like numerics; and Babylon code reflection turns Java lambdas into Onnx/Onyx protobuff models. A Java Llama2 port jumps from 1.3 → 15 tokens/sec when vectorized, and an MSET‑2 anomaly detector built on the matrix API outperforms MATLAB/Python/CUDA on AMD Zen in these tests.
AI-written summary · about
chapters (8)
- 0:00Java for AI introduction
- 2:13Key features for AI in Java
- 9:03Foreign function and memory
- 13:45Vector API performance
- 22:16Optimizing matrix operations
- 26:21MSET2 anomaly detection
- 32:17Valhalla and numeric types
- 33:44Code reflection and ONNX
related talks
49mJVM in the Age of AI: Babylon, Valhalla, TornadoVM and friends
47mJVM in the Age of AI: Babylon, Valhalla, TornadoVM and friends
02h 23mJava for an AI World - Keynote
49mWriting GPU-Ready AI Models in Pure Java with Babylon
46mThe Era of AAP: Ai Augmented Programming using Modern Java
43mWhy Java and the JVM Will Dominate the Future of Machine Learning, AI, and Big Data
51mJava at Speed: Getting the Most Out of Modern Hardware
43mHow to make your electrons do more in Java?