Apple’s AI researchers have made a significant breakthrough, aiming to bring large language models (LLMs) like ChatGPT and Claude to iOS devices, which typically struggle with limited memory capacity.
As reported by MacRumors, the innovative flash memory utilization technique devised by Apple holds the promise of making ‘Apple GPT’ a reality sooner than anticipated.
The challenge lies in the memory-intensive nature of LLM-based chatbots.
They demand considerable memory for smooth functioning, posing a challenge for devices like iPhones. However, Apple’s research introduces a pioneering method that stores AI model data in flash memory.
In their paper titled “LLM in a flash: Efficient Large Language Model Inference with Limited Memory,” Apple’s researchers highlight the advantage of using flash storage, which is more abundant in mobile devices than the RAM traditionally used for running LLMs.
The breakthrough technique revolves around two key strategies:
According to the research, these methods enable AI models to run up to twice the size of the iPhone’s available memory.
This breakthrough opens up new possibilities for future iPhones, such as advanced Siri capabilities, real-time language translation, and the integration of sophisticated AI-driven features in photography and augmented reality.
Apple’s strides in generative AI include the development of “Ajax,” an AI model aimed to rival leading models like GPT-3 and GPT-4.
With 200 billion parameters, Ajax signifies Apple’s ambition to deeply integrate AI across its ecosystem, potentially transforming the user experience.
Reports suggest that Apple is aiming to introduce generative AI features on iPhones and iPads around late 2024, coinciding with the release of iOS 18.
