Apple is planning to use its hardware prowess to give Apple’s AI a new twist and rely on media companies to train its AI models. The iPhone maker is aiming to differentiate itself from its Silicon Valley rivals, such as Google and Microsoft, who have been dominating the field of generative artificial intelligence (AI) with their cloud-based platforms and services.
Generative AI is a branch of AI that can create original and realistic content, such as text, images, audio, and video, based on large data sets and user inputs. Examples of generative AI applications include chatbots, image generators, voice assistants, and content creation tools.
While Google and Microsoft have been investing heavily in developing and deploying large language models (LLMs), such as ChatGPT and Turing-NLG, that can run on their cloud servers and offer various generative AI services to their customers, Apple has been taking a different approach. Instead of relying on the cloud, Apple wants to run its AI models directly on its devices, such as iPhones, iPads, Macs, and Apple Vision Pro headsets.
Apple’s advantage in this strategy is its hardware expertise and innovation. The company has been designing its own custom chips, such as the M2 series, that can offer high performance and power efficiency for running AI tasks on devices. Apple claims that its chips can deliver up to 30 times faster AI performance than the previous generation of Intel-based Macs. Moreover, Apple’s chips can support up to 96GB of unified memory, which is crucial for running large AI models that require a lot of data.
Apple’s challenge, however, is how to train its AI models without relying on the cloud. Training AI models requires a lot of data and computing resources, which are usually available in large data centers. Apple has been known for its strict data privacy and security policies, which limit its access to user data and prevent it from using third-party cloud services.
To overcome this challenge, Apple has been exploring ways to leverage its media partnerships and content ecosystem. The company has been working with media companies, such as Disney, Netflix, and Spotify, to use their content libraries and user preferences to train its AI models. For example, Apple could use the movies and shows from Disney+ to train its image-generating models, or the songs and playlists from Spotify to train its voice assistants. Apple could also use its own content platforms, such as Apple Music, Apple TV+, and Apple News, to feed its AI models with data.
By using media content and user feedback to train its AI models, Apple hopes to achieve two goals: first, to ensure the quality and relevance of its generative AI outputs; and second, to respect the data privacy and security of its users. Apple believes that this strategy will give it an edge over its competitors, who may face issues such as data bias, misuse, and regulation.
Apple’s AI strategy is still in its early stages, and the company has not yet revealed many details about its generative AI products and features. However, the company has been publishing some research papers and patents that hint at its direction and ambition. For example, Apple recently published a paper on how to run large language models on smartphones, which could enable chatbots and content creation tools on iPhones. Apple also filed a patent on how to use generative AI to create personalized avatars for its Apple Vision Pro headset, which could enable immersive and interactive experiences in augmented and virtual reality.
Apple’s AI strategy is a bold and innovative move that could reshape the landscape of generative AI and offer new possibilities for its users and partners. However, the company also faces many challenges and uncertainties, such as technical feasibility, user adoption, and market competition. Whether Apple can succeed in its AI strategy remains to be seen, but one thing is clear: the company is not afraid to give AI a new twist.