With the advancement of technology, hardware has been upgraded to smart hardware, which has added CPUs, operating systems, and also features such as internet connectivity and health monitoring. The application of AI technology has further upgraded ordinary AI in smart hardware to generative AI. At present, we are no strangers to generative AI, which brings more personalized recommendations and more humanized interaction methods.

In the field of wearable devices, hardware manufacturers such as Apple, Samsung, 360, Sony, and XREAL, who produce smartwatches and XR devices, have all incorporated generative AI into their products. Upstream in the industry chain, chip manufacturers involved in processors and sensors have found new opportunities as generative AI becomes increasingly intertwined with wearable devices.

Advertisement

AI Evolution Direction: From Traditional AI to Generative AI and Multimodality

Zhou Hongyi, the founder of 360 Group, believes that there will be two types of hardware in the future: one is hardware + AI, and the other is the evolved version of hardware + AI: AI-Native hardware. The first type is traditional smart hardware, while the second type is AI-Native hardware, where the AI large model is the core function, such as humanoid robots, autonomous driving, AI PCs, and new types of AI hardware.

In a sense, smart hardware is currently transitioning from traditional AI to generative AI, and large models will find more suitable application scenarios on the consumer side.

"2024 is the year of large model application scenarios, with killer apps emerging in the To C market," predicted Zhou Hongyi. Only multimodal large models combined with smart hardware can penetrate deeper into various application scenarios. So, what does multimodality refer to, and what kind of hardware do large models require?

Multimodal capability refers to the various types of information that wearable devices can obtain using health monitoring sensors, such as sound, gestures, heart rate, electrocardiograms, body temperature, positioning, and other data. With the support of large models, AI technology can integrate these individual data points across different dimensions, creating new dimensions and personalized data.

As for hardware, the wearable device sector has already introduced a variety of different types of products, including smartwatches/bands, wireless earphones, AR/VR devices, AI Pins, smart rings, etc.

Based on different application scenarios and needs, smartwatches have become one of the fastest wearable device categories for the implementation of generative AI. This is because they can provide more data required by generative AI. Since 2023, several manufacturers have announced plans to embed cloud-based generative AI in their smartwatches, including Zepp Health, Google (Fitbit), Samsung, Apple, Whoop, 360, and others.

Generative AI has brought two major benefits to the smartwatch market: first, the integration of generative AI has improved the accuracy and relevance of health monitoring data, and product iteration has given manufacturers a competitive edge in the fierce market competition. Second, generative AI has become a remedy for the slow growth of the smartwatch market and has also become a driving force for consumer purchases beyond health monitoring and other smart feature upgrades.If the growth of the smartwatch market in 2023 is attributed to the support of eSIM, satellite communication, and the enhancement of Bluetooth and UWB technologies, then it can be predicted that the addition of generative AI will be one of the keys to the continued growth of the smartwatch market in 2024, especially in the high-end segment. Canalys forecasts that, driven by AI features, the global smartwatch shipment growth will be approximately 20% by 2025.

Unlike traditional AI, generative AI learns and understands vast amounts of data to automatically generate new content or solve new problems, exhibiting stronger creativity and adaptability. In wearable devices, particularly those with health monitoring capabilities, it can provide personalized health advice, tailor-made exercise plans for sports scenarios, and personalized services in daily life.

In the realm of wearable devices, XR manufacturers are also seeking the best integration of AR/VR with AI. Said Bakadir, Senior Director of XR Product Management at Qualcomm Technologies, believes that the best interaction with generative AI is through XR. This represents another application direction for AI functions, distinct from those of Apple and Samsung.

At the AWE XR conference, AR eyewear technology company DigiLens announced that the ARGO glasses will integrate Google's Gemini large model, capable of voice interaction and other functions. It is worth mentioning that the combination of AI with smart glasses sensors can perceive the wearer's environment and intentions. It can also provide professional task visualization effects such as navigation assistance based on application scenarios.

Recently, foreign media revealed details of a new smart glasses product that Apple is planning, mentioning that the product will have built-in speakers, cameras, health sensors, and AI capabilities, allowing users to converse with AI assistants without a display screen.

Hybrid AI is accelerating its implementation, with processors and sensors maintaining their iteration pace.

The progress of generative AI is not limited to smartwatches. In the wearable device market, the most attention is focused on manufacturers Apple and Samsung, who announced their latest advancements in generative AI in June and July of this year, respectively.

Apple introduced the Apple Intelligence personalized intelligent system, which can understand personal contexts. The built-in large model has a deep understanding of natural language, capable of generating language, images, and cross-APP collaboration functions. Apple Intelligence is integrated into the core of iPhone, iPad, and Mac chips and has been empowered to Siri. Siri is an important layout for Apple in the AI field, and in the future, Apple Intelligence will be combined with Siri and integrated into more of Apple's wearable device products.

As a latecomer to the generative AI race, Apple will continue to train Apple Intelligence and introduce more features based on generative AI, bringing a brand-new AI experience to hardware.

Samsung, on the other hand, has deeply integrated AI functions into a new category of wearable devices—the smart ring Galaxy Ring. Unlike Apple's focus on interaction in generative AI, Samsung's generative AI smart ring focuses more on health monitoring features, which is also greatly related to the product's functional attributes. Samsung states that the Galaxy Ring can analyze data through AI to provide personalized health and sleep recommendations, and it can also control Samsung smartphones with gestures.Samsung also announced that the new features of Galaxy AI will be updated to more Samsung Galaxy devices. The built-in AI algorithms will develop a personal knowledge graph for each user and create more personalized services. Samsung believes that multimodal and contextual AI will become an interconnected ecosystem.

From the layout of Apple and Samsung in wearable generative AI, a new development direction can be seen: hybrid AI. For example, Apple chooses to cooperate with OpenAI in the field of AI, while Samsung chooses to join hands with Google. Specifically, Apple uses a self-developed local large model + cloud, some AI functions use Apple Intelligence on the local end, and when the local processing ability is exceeded, it is based on OpenAI's GPT-4o model to achieve more complex task processing in the cloud.

The strategy of hybrid AI not only improves the level of intelligence of the device but also enhances the user experience by introducing advanced generative AI technology. Driven by Apple, the "hybrid AI" of wearable devices in the future will be accelerated. However, the premise is that the AI on the end side is strong enough.

It is not difficult to find that on the one hand, generative AI is landing on more wearable categories, and on the other hand, as the AI functions of wearable devices become more and more abundant, the performance of the main chip, sensors and other hardware also needs to be more powerful to support the realization of more functions, which is also the key to the landing of multimodal large models.

The author believes that the core of wearable device AI includes sensors, storage, etc. In terms of the main chip, as the AI function increases, the processor's performance needs to be more powerful to perform complex algorithms and tasks, and it also needs higher integration. In terms of sensors, the foundation of multimodality is the various data detected by the sensor, and Canalys said that for each additional sensor, AI can create dozens of new application scenarios based on that sensor. Of course, this requires the sensor to have higher accuracy. In terms of storage, the increase in AI applications brings requirements for storage capacity and speed.

In summary:

As Zhou Hongyi mentioned, "Large models are a kind of ability, finding core application scenarios, and combining them with capabilities is very important." Only by finding the pain points and needs of users in specific vertical scenarios. Generative AI aims at the rapidly growing wearable device market, and is realized on devices such as smartwatches, smart rings, and smart glasses, and brings technical competitive advantages to brand manufacturers, and is favored by consumers.

In 2024, which is considered by industry insiders as the first year of application, the penetration of generative AI in wearable devices has just begun. Brand manufacturers are looking for suitable technical routes for themselves, and hybrid AI has become one of the technical routes that everyone sees. The improvement of functional experience also promotes the product iteration of upstream and downstream industry chain enterprises.