Yesterday, Google declared its big artificial intelligence (AI) ambitions after its DeepMind division unveiled the company’s largest AI model Gemini, which is a multimodal foundation model, to the world. However, it wasn’t the only tech company to make a big move towards AI. In order not to be left behind, Apple looks to have joined the AI war that is currently raging in Big Tech by releasing a new machine learning (ML) framework called MLX, but without much fanfare. Many reports are now suggesting that Apple can use this framework to build its own AI foundation models, something that was rumored earlier this year.
According to a report by The Verge, MLX is “a machine learning framework where developers can build models that run efficiently on Apple Silicon and deep learning model library MLX Data”. This is the biggest clue that this framework might be intended to develop its own AI models, although is unlikely to be a generative AI model, the likes of which have been released by Google, OpenAI, and others. Apple is pretty tight-lipped when speaking about the artificial intelligence features on its devices and refers to it as machine learning instead. Despite adding a few features for iPhones, such as Personal Voice, which is essentially based on AI algorithms, the company refrained from using those terms.
Who can use the Apple MLX tool?
According to a report by Computerworld, MLX is not a tool intended for consumers but for its developers to get a powerful environment to train ML models. An interesting part of the report is that Apple has not forced any particular coding language for this framework, allowing developers to choose freely the language they want to, and it apparently invented powerful LLM tools in the process”.
An ML researcher working with Apple, Awni Hannum, posted on X and said, “Just in time for the holidays, we are releasing some new software today from Apple machine learning research. MLX is an efficient machine learning framework specifically designed for Apple silicon (i.e. your laptop!)”.
The framework has been shared on GitHub and is kept open-source for developers to see and freely work with it. The MLX framework works with PyTorch, ArrayFire, or Jax frameworks.
A note accompanying the release mentioned, “The framework is intended to be user-friendly, but still efficient to train and deploy models…. We intend to make it easy for researchers to extend and improve MLX with the goal of quickly exploring new ideas.”
We will only find out in time the kind of foundation models that are created using this framework, which will provide more clarity on the direction of this release.