Is iOS 18 Apple's AI Answer? You Might Not Truly Understand Mobile AI

[baoshi.rao]

Since the second half of last year, the smartphone industry has been swept by an AI wave, with leading manufacturers rapidly deploying AI strategies. Terms like 'generative AI', 'on-device AI models', and 'AI phones' have since permeated the consumer market.

Currently, major domestic brands have launched their own AI models, such as vivo's Blue Heart model, OPPO's Andes model, and Meizu's Flyme AI model, all of which are currently popular AI products. Clearly, nearly all manufacturers are discussing AI models for smartphones, yet as consumers, we may not have a clear understanding of this trend.

If you ask consumers what mobile AI means, their answers would likely focus on functional-level perceptions rather than comprehensive technical or experiential understanding. Indeed, from a marketing perspective, showcasing practical features as promotional content remains the most effective way to make consumers perceive and accept these innovations. For example, the AIGC removal feature primarily utilizes the unique advantages of AI technology to identify user-selected content, helping users remove unwanted elements from images with a single click and then filling in the gaps to create a picture that better aligns with their preferences. In simple terms, it automates the complex process of Photoshop-like object removal, reducing tedious steps and lowering the barrier to entry.

This feature is likely already familiar to many, so we won't elaborate further here. However, it's evident that whether in live demonstrations at launch events or subsequent official promotional materials, there are numerous tutorials and demo videos available. Users can easily understand and quickly get started, directly experiencing the novelty of AI-enhanced smartphone capabilities.

That said, the AI referred to here mostly pertains to generative AI on smartphones, which is just one part of the broader mobile AI landscape. Currently, the application and adoption of generative AI in smartphones are still insufficient, leading many consumers to develop a narrow perception that mobile AI should be limited to certain frequently highlighted features. This perception isn't entirely wrong, as technology ultimately aims for functional convenience. However, what mobile manufacturers are currently pursuing is on-device generative AI. Compared to cloud-based generative AI, it offers lower costs, enhanced security, greater accuracy, and delivers a comprehensive experience through multi-functional integration. This creates a cognitive gap with consumer expectations.

Market research firm Counterpoint Research has indicated that this year, the market is expected to see over 100 million generative AI smartphones or devices running native AI models that create AI-generated content. In fact, judging by the current market situation, nearly all Android brands have launched corresponding products, with Apple being the only major player not yet participating in the generative AI race.

As a giant in the mobile industry, Apple's moves are closely watched, with iPhone updates drawing particular attention. While all other mobile manufacturers are deploying on-device generative AI, Apple has shown little movement, revealing no visible plans for generative AI features - a situation that naturally arouses curiosity. Recently, Bloomberg revealed that Apple's generative AI might be integrated into iOS 18, set to release this year, potentially making it more noteworthy than the iPhone 16 upgrades. While specific features remain unknown, given Apple's in-house chips and systems, its AI capabilities are certainly worth anticipating.

Meanwhile, Bloomberg also reported that Apple acquired DarwinAI earlier this year to bolster its AI expertise. The company specializes in two core technologies: AI for inspecting components during manufacturing and machine learning for developing smaller, more efficient AI systems—both of which could significantly influence iOS 18's AI features.

As is well known, Apple typically updates its operating systems for phones, tablets, and laptops simultaneously, with increasingly tight integration across devices. If generative AI is indeed introduced in iOS 18, it's reasonable to expect tablets, laptops, watches, and other products to follow suit, potentially creating entirely new interconnected experiences. This exemplifies Apple's ecosystem advantage, where a single update can ripple across all devices, accelerating its AI rollout. In fact, when it comes to AI features in smartphones, almost all manufacturers have implemented them to some extent. For example: copying text from images, one-click background removal, document scanning, and instant photo capture – these functionalities themselves represent applications of AI technology. However, today's on-device generative AI represents a more profound and long-term development direction, which is still in its early stages and naturally faces considerable challenges.

In the mobile internet era, where nearly everyone owns a smartphone, the potential user base for on-device generative AI is enormous. This technology requires greater computational power, more foundational models, and larger model parameters to support it, making overnight success virtually impossible. Therefore, in the early stages of smartphone generative AI development, we can expect numerous shortcomings and occasional failures.

Taking AIGC erasure functionality as an example, the processing of an image typically involves three steps: recognition, removal, and filling. Simple backgrounds pose no problem, but with slightly more complex image elements, noticeable smearing effects may occur, sometimes even filling in irrelevant content, significantly reducing practical usability. For instance, text-to-image generation often produces results that don't align with reality. The command recognition isn't precise enough, failing to render slightly complex instructions and sometimes generating nonsensical outputs. Such issues significantly degrade user experience, eventually leading to loss of interest.

Currently, generative AI on mobile devices is still in its infancy. While some practical functionalities have been implemented, there's considerable room for improvement in areas like feature refinement and privacy protection. However, the various AI concepts proposed by smartphone manufacturers are crucial, both for their own product development and for revitalizing the stagnant mobile industry.

Both manufacturers and consumers eagerly anticipate the true arrival of AI smartphones and a future of more intelligent and digital interactions. However, more usable AI features and higher-frequency computations require more powerful hardware support. This means both Android and Apple ecosystems need to achieve new hardware breakthroughs to sustain on-device AI computations. Consequently, we can expect smartphone manufacturers to soon introduce dedicated hardware for AI functionalities, such as AI chips and AI watches. Let's wait and see.