Apple has always kept aloof compared to recent times hype concerning artificial intelligence, the new goose that lays the golden eggs of tech companies, but in recent days without publicizing it it has published a new framework For AI models called MLX.
The main competitors of the house of Cupertino in fact, they invest billions in this technology and continue to release products to attract users and investors, and yesterday’s announcement by Google regarding the model Gemini, which promises to be more powerful than GPT-4is just the latest example (do you know how to access ChatGPT?).
The apple company, on the other hand, even avoids using the word AI in many of his keynote presentations, leading many to wonder why. In September, however, the first ones appeared online rumor about whether Apple was spending Millions of dollars per day just to train the AI, with the aim of working on models to implement in its services.
Now, as we anticipated, the apple research team on machine learning published on GitHub and PyPI the MLX framework, a framework for machine learning upon which developers can build models to run specifically on chips Apple Silicon and the MLX Data deep learning model library.
The news immediately aroused the interest of experts. Awni Hannuna machine learning researcher at Apple, posted on X that MLX Data is a “package agnostic, efficient and flexible for data loading” and which works on the MLX, PyTorch or Jax frameworks.
On GitHub, Apple explains that to create MLX the engineers took inspiration from frameworks such as PyTorch, Jax And ArrayFire, but there’s a difference: shared memory. This means that any activity performed on MLX works on supported devices (right now, CPU and GPU) without requiring data to be moved to the cloud, which is great for privacy.
Second Computerworld, who analyzed the instrument, MLX it’s intended for developers to create new AI apps, possibly for MacBooks, but has enough power to train AI models like Llama by Meta e Stable Diffusion.
At the moment Apple has not yet released one communication official or responded to requests for comment, but there are two considerations to make. On the one hand it must be admitted that Apple has recognized the need to build development environments open And easy to use for machine learning in order to push development in this area, but it’s a bit in delay?
The Apple company has evidently decided to put the best tools for working on artificial intelligence in the hands of developers: a framework on which to create the models and computers with M3 chips on which train them, which in theory shifts the paradigm from AI to human-focused AI.
Whether it’s the right choice (and timing) remains to be seen to impose oneself on the market, but perhaps it is also what Apple has always done: approach a problem from a different point of view.
If you are interested, here you will find the framework’s GitHub page and the MLX site with the documentation.
