Apple dips toe in the AI pool with new model framework

The company released a framework to build models that run on Apple silicon

MacBook Air 13-inch

Apple quietly released new frameworks and model libraries designed to run on its M-series chips and, possibly, bring generative artificial intelligence (AI) apps to MacBooks.

As detailed by The Verge, Apple’s machine learning team released MLX, a machine learning framework that enabled developers to build models that can run efficiently on Apple chips. There’s also a deep learning library dubbed MLX Data. Apple made both available through open-source repositories like GitHub and PyPI.

Apple notes on its GitHub page that the PyTorch, Jax and ArrayFire frameworks inspired the design of MLX. However, MLX has a notable difference of utilizing shared memory. That means any task running on MLX works on supported devices like CPUs and GPUs without moving data.

Moreover, MLX is intended to be easy for developers to use while offering enough power to train AI models like Meta’s Llama and Stable Diffusion.

MLX Data, on the other hand, is a “framework agnostic, efficient, and flexible package for data loading,” as described by Apple machine learning researcher Awni Hannun on Twitter/X. MLX Data will work with MLX, PyTorch, or Jax frameworks.

While it’s good to see Apple moving forward in this area, it’s worth noting that these new machine learning frameworks are, well, about machine learning. They’re a far cry from the generative AI apps and tools pushed by competitors, such as Microsoft’s OpenAI-powered Copilot or Google, which just unveiled its new Gemini model.

Apple is reportedly working on generative AI, but it remains to be seen if or when it will launch products with it.

Source: The Verge

Related Articles