Apple's latest research into running large language models on smartphones more clearly highlights the company's intention to catch up with its Silicon Valley competitors in the field of genetic artificial intelligence.
See also: Aston Martin and Porsche unveiled the next generation of Apple CarPlay coming to their cars in 2024
The paper, titled “LLM in a Flash,” offers a “solution to a current computational impasse,” as its researchers say. Their approach “pave the way for efficient estimation of LLMs on memory-constrained devices,” they say. Estimation refers to how large language models, which power apps like ChatGPT, respond to user queries. Typically, chatbots and LLMs run in massive data centers with far more computing power than an iPhone.
The paper was published on Dec. 12, but gained wider attention after being flagged by Hugging Face, a popular website for AI researchers, late Wednesday. It is Apple’s second paper on generative AI this month and follows previous moves to enable image-generating models, such as Stable Diffusion, to run on its custom displays.
Device makers and processor manufacturers are hoping that new artificial intelligence features will help revive the smartphone market, which had its worst year in a decade, with shipments falling an estimated 5%, according to Counterpoint Research.
Despite Apple releasing one of the first virtual assistants, Siri, in 2011, it is largely seen as lagging behind its competitors in the field of artificial intelligence. Many in the AI community see Apple as lagging behind its big tech rivals, despite hiring Google's top AI executive, John Giannandrea, in 2018.
While Microsoft and Google have focused primarily on delivering chatbots and other generative AI services over the Internet from their massive data centers, Apple's research suggests it will instead focus on AI that can run directly on an iPhone.
See also: United Kingdom: Artificial Intelligence cannot be recognized as a creator
Apple rivals like Samsung are gearing up to launch a new breed of “AI smartphone” next year. Counterpoint estimates that more than 100 million AI-focused smartphones will ship in 2024, with 40% of new devices offering such capabilities by 2027.
The head of the world's largest chip processor company, Qualcomm CEO Cristiano Amon, predicted that the introduction of artificial intelligence into smartphones will create a completely new experience for consumers and reverse the decline in mobile sales.
“You’ll see devices coming out from early 2024 with a number of uses of generative AI,” he told the Financial Times in a recent interview. “As these things scale, they start to cause a significant change in the user experience and enable new innovations that have the potential to create a new upgrade cycle in smartphones.”
See also: OpenAI: Presents “Readiness Framework” for managing risks in artificial intelligence
More sophisticated virtual assistants will be able to predict users' actions, such as automatically sending messages or scheduling a meeting, he said. The devices will also be capable of new kinds of photo editing techniques.
Google this month introduced a version of the new Gemini LLM model that will work "primarily" on its Pixel smartphones.
Running the specialized AI model that powers ChatGPT or Google's Bard on a personal device poses significant technical challenges, as smartphones lack the vast computing resources and energy available in a data center. Solving this problem could mean AI virtual assistants will respond faster than you might think and even work offline.
Ensuring that questions are answered on each person's device without sending data to the cloud is also likely to bring privacy benefits, a key factor for Apple in recent years.
“Our experimentation was designed to best handle estimation efficiency on personal devices,” the Apple researchers said. Apple tested its approach on models including the Falcon 7B, a smaller version of an open LLM model originally developed by the Technology Innovation Institute in Abu Dhabi.
See also: Apple Vision Pro: Will it be released in February 2024?
Optimizing LLMs to run on battery-powered devices has become a growing focus for researchers in the field of artificial intelligence. The scientific papers are not a direct indication of how Apple intends to add new features to its products, but they do offer a rare glimpse into the company's secretive research labs and its latest technical achievements.
“Our work not only offers a solution to a current computational bottleneck, but also paves the way for future research,” the Apple researchers wrote in their conclusion. “We believe that as LLMs continue to grow in size and complexity, approaches like this will be essential to realizing their full potential across a wide range of devices and applications.”
Source: arstechnica.com
