Apple Boosts Budget to $1 Million Daily for Advanced AI, Optimizing Siri!

Recent reports reveal that Apple has been steadily increasing its budget for building the computational infrastructure required for artificial intelligence (AI) and has expanded it to millions of dollars per day. One of the company's goals is to enable iPhone users to perform multi-step tasks with simple voice commands, enhancing Siri's capabilities.

For instance, this technology could allow users to command Siri to create a GIF animation from their five most recent photos and send it to a friend using voice commands. Currently, iPhone users must perform this process manually.

Apple's Early Investment in AI

Compared to global tech giants like Microsoft and Google, Apple's progress in AI technology, including the highly popular generative AI, seemed slower. However, according to reports from sources, Apple is increasing its investment significantly, now allocating millions of dollars daily to AI, with multiple teams concurrently developing various AI models.

One of these teams, responsible for developing conversational AI, is known as the Foundational Models team, comprising several engineers with extensive experience at Google. Currently, it consists of 16 members. In fact, four years ago, Apple's AI head, John Giannandrea, authorized the formation of this team to develop conversational AI based on large language models (LLMs). John Giannandrea joined Apple in 2018 and oversaw the upgrade of Siri.

In addition to this, Apple has at least two other teams working on advancing AI. One focuses on Visual Intelligence, aiming to create software capable of generating images, videos, or 3D scenes. The other team is researching multimodal AI, which can recognize and generate images, videos, or text. One of the leaders of this team is Jon Shlens, who joined Apple at the end of 2021 after working with Google's AI team, DeepMind.

Simultaneously, Apple is aggressively recruiting AI model development personnel. Reports suggest that Apple has recruited several researchers from Meta in the field of large-scale models and is actively hiring engineers specializing in large language models globally. Apple's official website displays over 600 job postings related to large language models, focusing primarily on optimizing AI software for multimedia functions, including image and video generation, visual actions, and interactions.

It is understood that Apple's Foundational Models team has already developed several advanced models and is conducting internal testing. Among these, the most advanced language model, called Ajax GPT, has been trained with over 200 billion parameters and is claimed to outperform OpenAI's previous GPT-3.5 version, ChatGPT.

Different from companies like Google and Microsoft, Apple is committed to integrating generative AI on mobile devices. However, large language models require more storage and computational power, posing challenges for running them on mobile devices like the iPhone. Nevertheless, this is not impossible, as the industry has seen companies use various techniques, including quantization, to reduce model sizes for deployment on terminal devices.

From Apple's recent job postings, it is clear that Apple is also researching in this direction. For example, it seeks candidates who can bring advanced foundational models to our pockets on the iPhone. According to insiders, Apple's chatbot based on large language models may eventually interact with users who use AppleCare. Another application could be enhancing Siri, making it easier to perform complex tasks with voice commands.

Optimizing Siri with Large Language Models

In the past year, the global consumer electronics market has remained sluggish, impacting Apple's performance. According to its financial report for the third quarter of the 2023 fiscal year, Apple's revenue was $81.8 billion, a 1.4% decrease year-on-year, while net profit increased by 2.3% to $19.9 billion. This marked the third consecutive quarter of declining revenue for Apple, with its flagship products, iPhone, iPad, and Mac, all experiencing some decline.

Despite the upcoming release of the latest iPhone 15, the industry's outlook for Apple is not optimistic. Renowned analyst Ming-Chi Kuo believes that demand for the iPhone 15 series will be lower than that for this year's iPhone 14 series.

In this context, Apple seems to have high hopes for large AI models. Recently, Tim Cook stated that Apple has been researching generative AI and other models for many years, and in the coming months, Apple will increase its research spending on generative AI.

Industry experts believe that for Apple, Siri is the quickest place where its new GPT system can be deployed and the fastest way to improve the user experience. Integrating Apple GPT with Siri could likely result in significant efficiency and user interaction improvements, potentially giving Apple a significant advantage in the large model competition.

Siri is Apple's voice assistant used on various products, including the iPhone, iPad, iPod Touch, HomePod, Apple Watch, Apple TV, and Apple CarPlay. Siri enables users to find information, make calls, send messages, get directions, play music, locate Apple devices, and more using voice commands.

Siri supports natural language input and can access built-in applications such as weather forecasts, schedules, and web searches, continually learning new voices and tones to provide conversational responses.

Over the years, Apple has actively improved Siri. However, many users are not satisfied with Siri's current performance, with some even referring to it as "AI with disabilities." Reports suggest that Apple is actively enhancing Siri's intelligence and is expected to unveil these improvements in iOS 18. Apple is expanding Siri's capabilities, enabling it to automatically execute various complex multi-step tasks based on users' natural voice commands. As mentioned at the beginning of the article, Siri can take five consecutive photos with a single voice command, convert them into GIF animations, and send them to designated friends automatically.

However, users have varied opinions about Apple's optimization efforts. Some believe that integrating large models into Siri might be a misdirection and would prefer a voice assistant with robust speech recognition capabilities rather than one that intelligently interprets their spoken content.

Others argue that adding new models to Siri will significantly enhance its usability, and Apple will adjust parameters to ensure that AI-driven Siri is less verbose. Perhaps, by upgrading Siri once again, it will not merely provide a web link but truly address users' needs.

In Summary

Compared to global tech giants like Google and Microsoft, Apple's progress in showcasing AI technology may appear slower, but upon closer examination, it becomes clear that its AI research teams started their work early. Currently, with the continuous decline in the global consumer electronics market, Apple's performance has been affected for several quarters. By integrating generative AI with its existing technologies like Siri, Apple may bring about significant innovations, benefiting both the company and the field of AI applications.