what’s new in machine learning

At WWDC, Apple introduced several new features and enhancements in the realm of machine learning:

1. Create ML Enhancements: The Create ML app now includes an object tracking template for training reference objects on Vision OS, and new components for building custom time series forecasting and classification models. What’s new in Create ML.

2. Core ML Updates: Core ML continues to be the primary framework for deploying models on Apple devices, with new features to help run state-of-the-art generative AI models on-device. This includes a new ML tensor type and performance reports for better model execution insights.

3. BNNS Graph API: The Basic Neural Network Subroutines (BNNS) Graph API has been introduced to optimize machine learning and AI models on the CPU, making them faster and more energy-efficient. Support real-time ML inference on the CPU.

4. Metal 4 Integration: Metal 4 now supports machine learning workloads, allowing for seamless integration and synchronization with compute and render work. This includes the ML command encoder and ShaderML for embedding ML operations directly into shaders. Combine Metal 4 machine learning and graphics.

5. Training on Apple GPUs: Enhancements have been made to support popular transformer models on Apple Silicon GPUs, enabling the use of larger models and batch sizes without the need for CPU-GPU memory copies. Train your machine learning and AI models on Apple GPUs.

6. MLX Framework: MLX is a powerful tool for numerical computing and machine learning, designed to help developers keep up with the rapid advancements in ML research.

These updates reflect Apple's commitment to providing robust tools and frameworks for developers to create innovative machine learning applications across their platforms.