Apple's latest AI models may power upcoming iPhones, iPads, MacBooks
Apple is making significant strides in enabling Artificial Intelligence (AI) to operate locally on its devices. The company's research team has now launched OpenELM, a collection of four compact language models. These models, now accessible on the Hugging Face model library, are open-source and ready for developers to utilize. According to Apple's Hugging Face model page, OpenELM stands for "Open-source Efficient Language Models" and excels in text-related tasks such as drafting emails.
OpenELM is more compact than many existing lightweight AI models
OpenELM is more compact than many existing lightweight AI models, making it an ideal solution for devices like iPhones, iPads, and MacBooks. It comes in four variants based on the number of parameters: 270 million, 450 million, 1.1 billion, and 3 billion. Parameters are essentially the variables a model comprehends from its training datasets during decision-making. For comparison, Microsoft's newly launched Phi-3 model starts at a minimum of 3.8 billion parameters while Google's Gemma offers a 2 billion parameter version.
Generative AI features are coming to new-gen Apple products
Earlier this year, Apple CEO Tim Cook hinted at the integration of generative AI features into the company's upcoming devices. In February, Cook stated that Apple is dedicating "a tremendous amount of time and effort" in this domain. However, specific details about how Apple plans to leverage AI remain undisclosed. This venture into AI models isn't new for Apple. Last December, they introduced MLX, a machine learning framework aimed at enhancing the performance of AI models on Apple Silicon.
Apple has previously launched AI models for image editing, navigation
Apple has previously launched an image editing model named MGIE, which allows users to edit photos using prompts. They also introduced a model called Ferret-UI designed for smartphone navigation. Rumors suggest that Apple is working on a code completion tool similar to GitHub's Copilot. Interestingly, Apple has reportedly been in discussions with Google and OpenAI about integrating their models into Apple products, indicating a collaborative approach to AI development.
