Apple joins AI race with focus on object identification and image generation models

The world is bustling with excitement about AI, and tech giants are actively participating in the race.

However, Apple, another prominent tech player, is quietly making strides in the AI space without much fanfare.

Apple GitHub has recently unveiled two noteworthy GitHub repositories.

One is MLX, an array framework tailored for Apple silicon, and the other one is Ferret.

MLX empowers users to effortlessly run models like LLaMA and Mistral AI, enabling tasks such as speech recognition (akin to OpenAI and Whisper AI) and image generation (similar to Stable Diffusion) directly on your MacBook.

It also facilitates the training of transformer models and fine-tuning LoRA. 

Fine-tuning has become more accessible over time, thanks to the progress of LLaMA factory, making it increasingly user-friendly for Mac users.

Awni Hannun, a machine learning researcher at Apple, described MLX Data as a versatile, efficient, and flexible package for data loading, compatible with MLX, PyTorch, or Jax frameworks. 

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!)

Code: https://t.co/Kbis7IrP80
Docs: https://t.co/CUQb80HGut

— Awni Hannun (@awnihannun) December 5, 2023

Additionally, Apple boasts its very own MLLM (Multimodal Large Language Model), surpassing the GPT model in vision and object identification, according to tech experts.

Despite being relatively modest with 7 billion and 13 billion checkpoints respectively, some speculate that Apple's work behind the scenes might only be fully appreciated when these advancements are integrated into future iPhones.