Apple's recent bombshell of canceling its decade-long EV project, Titan, has sent shockwaves through the industry. However, amidst the fallout, a surge in acquisitions, research, and patent applications hints at a bold new focus on expanding AI applications.
Reports from VentureBeat and Computerworld shed light on Apple's upcoming AI ventures, particularly in generative AI. These ventures are set to hone in on three key areas: LLM inference on terminal devices, open-source LLMs, and cutting-edge software development tools.
Apple's AI endeavors
The company's commitment to AI development is palpable with the launch and continuous updates of various AI programming tools. Apple is making significant strides in AI, from Xcode for coding to its Keyframer for animating 2D images, and MGIE for editing images using natural language commands.
While other tech giants may favor cloud-based LLM models, Apple is doubling down on its hardware ecosystem. With a robust lineup including iPhones, iPads, Apple Watches, AirPods, speakers, PCs, and laptops, Apple aims to push the boundaries of LLM and AI applications directly on terminal devices.
LLM prospects
Apple's recent report delves into the feasibility of running LLMs on devices with limited RAM and storage. The paper outlines how enabling model data to switch between flash memory and DRAM can facilitate smooth AI model operation on Apple chips, minimizing memory consumption and inference latency.
In a departure from its closed ecosystem reputation, Apple introduced an open-source multi-module LLM named Ferret in October 2023. Built on Vicuna and LLaVA AI models, Ferret excels in visual image processing and is ideal for future interactive applications on iPhone camera lenses or Vision Pro devices, particularly for scenarios involving small objects or details in images.
Moreover, Apple's machine learning model database MLX and software development tools like Xcode, which generates code using natural language commands, aim to empower developers in exploring efficient hardware and software innovation applications through the potent capabilities of AI technology.
