Apple accelerates "AI on-device" and focuses on tackling "large models on mobile devices"

A series of acquisitions, employee recruitment, and academic paper publications by Apple indicate that it is making efforts to integrate AI into the next generation of the iOS operating system.

According to research by PitchBook, Apple has been more active in acquiring AI startups than other large tech companies, having acquired 21 companies since early 2017. The most recent acquisition was in early 2023, when it acquired WaveOne, a California-based startup that provides AI-based video compression technology.

Media reports quote Daniel Ives from Wedbush Securities as saying:

"They are preparing for some major M&A deals. I would be shocked if they don't make a significant AI transaction this year, as an AI arms race is unfolding and Apple won't stand idly by."

A recent research report by Morgan Stanley also shows that nearly half of Apple's AI job postings include the term "deep learning," which is related to algorithms for generative AI. It is worth mentioning that Apple hired John Giannandrea, a senior AI executive from Alphabet-C, in 2018. Media reports citing industry insiders reveal that Apple is developing its own large language model.

However, Apple's goal seems to be to run generative AI on mobile devices, which would enable AI chatbots and applications to run on the hardware and software of smartphones, rather than being supported by cloud services in data centers.

A previous article by Wall Street News mentioned that Apple researchers published a paper in December of last year, introducing a method for running large language models on devices that exceed available DRAM (memory) capacity.

By significantly improving memory utilization efficiency, this new research supports running large language models twice the size of available memory on devices with limited capacity, and the inference speed of GPUs can be increased by tens of times.

This means that this groundbreaking research expands the applicability and accessibility of large language models, and Apple's plan to integrate generative AI into iOS 18 may be accelerated.

Well-known tech blogger Mark Gurman revealed on January 7th that Apple plans to introduce a series of generative AI tools at the Worldwide Developers Conference (WWDC) in June as part of the iOS 18 system. The underlying work of these AI tools is being done on a large language model called Ajax within Apple, which has been tested since early last year.