Apple’s elevated R&D spending as it seeks to improve the quality of its AI services offerings is an open secret in the industry, and a new report from The Information sheds light on just how much the company has dedicated to cementing itself as a serious AI player amid fears of falling behind competitors.
According to the report, Apple is now spending “millions of dollars a day” on its R&D budget solely dedicated to AI product research. The company has set an ambitious goal that would allow users to simply ask Siri to “automate tasks involving multiple steps,” according to sources who spoke to The Information.
The technology, for instance, could allow someone to tell the Siri voice assistant on their phone to create a GIF using the last five photos they’ve taken and text it to a friend. Today, an iPhone user has to manually program the individual actions.The Information
Apple has apparently been working on Large-Language Models (LLMs) as far as back four years ago after the company’s Senior Vice President of Machine Learning and AI Strategy John Giannandrea “authorized the formation of a team to develop conversational AI.” The team’s formation materialized way before the recent AI craze brought forward by the launch of ChatGPT, but it remains unclear what Apple has achieved with its few years’ lead ahead of OpenAI’s chatbot.
Apple also incorporated a new “Foundational Models” group that focuses on conversational AI research, according to the report. The team employs just 16 people and is led by Ruoming Pang. The report adds that most of the company’s millions of dollars a day budget is dedicated to training its language models, which requires a lot of expensive hardware.
The current goal that Apple is reportedly working towards is to ship an advanced version of the Shortcuts app with next year’s iOS 18 that enables Siri to automate complex multi-step tasks with a simple voice command like the example described above.
Apple’s AI teams – including a “Visual Intelligence” group – are actively developing software for generating 3D videos and images, as well as multimodal AI systems that can effectively process images, videos, and text. Apple’s internal “Apple GPT” chatbot apparently boasts greater capabilities than OpenAI’s GPT-3.5 model and has been trained on a huge dataset of 200 billion parameters. However, the latest GPT-4 model surpasses Apple’s in terms of intelligence.
Bloomberg earlier reported that Apple has been ramping up its AI and machine learning efforts in a bid to announce a consumer-facing AI product as early as next year. The company has internally built a ChatGPT-like bot dubbed “Apple GPT” atop its in-house “Ajax” LLM-powered framework.