Apple finally gets serious about gen AI with the release of MLX, an ML framework for devs
Apple has quietly introduced MLX, a new framework that simplifies the process for developers to create machine learning (ML) apps for Apple devices. It is built on top of Apple's pre-existing Core ML framework, but is open source, which should help expedite its development. However, in and of itself, it is not a large language model like ChatGPT, for example, but rather provides developers with the tools they need if they want to build generative AI apps themselves.
In a post announcing MLX on X/Twitter, Apple said that the framework is designed to make it "easier for developers of all skill levels to build ML-powered apps." The company also said that MLX is "optimized for performance and efficiency," so that developers can create apps that run fast and use battery power efficiently by taking advantage of Apple silicon and its unified memory architecture, which shares RAM across the CPU, GPU and NPU. The aim of the MLX framework is to tap into the full potential of Apple silicon to run complex AI models on-device.
As is often the case with Apple developed software tools, the aim is to make it easy to develop AI sophisticated AI models similar to Meta’s Llama and Stable Diffusion. Tools for MLX include a visual interface for creating and training ML models, a library of pre-trained ML models for easy integration by devs into their apps and tools for integrating user interfaces with ML models. It is also compatible with and inspired by more widely used machine learning frameworks such as TensorFlow and PyTorch.
With Apple still well behind the competition in ML/AI, MLX is one way to accelerate the process by encouraging developers to find ways to implement AI features on Apple devices. It remains to be seen whether MLX will be enough to help Apple catch up to the competition, but it is encouraging to see Apple getting more serious about its potential
MLX is available now in the Xcode 13 beta, which developers can also play around with MLX in GitHub now. Apple says that it will be available in the final release of Xcode 13 later this year.