Recently, Apple CEO Tim Cook stated that he has always prioritized being the best rather than being the first. Furthermore, there are reports that Apple plans to implement generative AI on the iPhone by late 2024. On the other hand, Samsung’s upcoming Galaxy phones may feature generative AI similar to ChatGPIT, and this could happen as early as 2024.
However, Chinese phone makers have a different approach. Companies like Xiaomi, Oppo, Vivo, and Huawei aim to integrate generative AI into their latest devices faster than Apple and Samsung. They see this as an opportunity to gain an advantage over their competitors in the highly competitive smartphone market.
Chinese phone makers have already introduced products supported by generative AI capabilities after investing several years in developing their artificial intelligence skills. Their agility allows them to move quickly, which aligns with their long-term strategy of attracting high-end users. Generative AI could be the catalyst they need to achieve this.
But can they succeed?
Users already have access to powerful generative AI apps on their phones. So, how much difference will it make if these capabilities are built directly into the devices?
Chinese smartphone manufacturers are focusing on three main areas to incorporate generative AI into their devices: improving their AI assistants, adding multimodal functions, and enabling cross-device functionality.
In August, Xiaomi founder and CEO Lei Jun announced that the company’s digital assistant, Xiao AI, is being upgraded to include generative AI capabilities. Xiaomi claims to have developed a 1.3 billion parameter lightweight AI model that runs locally on the phone, delivering performance comparable to a 6 billion parameter model that runs on its cloud computing service.
Xiao Ai can perform various tasks, including song and object recognition, call blocking, transportation route suggestions, and medication reminders. When integrated with Xiaomi’s smart home appliances, Xiao AI can also control devices such as rice cookers and air conditioners.
This functionality encompasses text, voice, and vision, and Xiaomi plans to expand it to other smart devices in their lineup. Lei Jun expects these functions, many of which already exist, to become even more intuitive and intelligent with the help of generative AI in the future.
The foundation for Xiao AI’s upgrade is Xiaomi’s long-term investment in AI. Since 2016, Xiaomi has built a team of over 3,000 people focused on AI. With over 110 million monthly active users, Xiao AI can rapidly enhance its capabilities.
Xiaomi’s domestic rivals have also made similar announcements. For instance, Huawei stated in August that its AI assistant, Xiao Yi, will be supported by their Huawei Cloud Pan Gu model. Xiao Yi will offer multimodal functions and be integrated across Huawei’s product range, spanning from phones to laptops and potentially smart cars.
In addition, Oppo revealed plans to launch an advanced digital assistant named Xiabu, developed by their Andes Intelligent Cloud team. Xiabu will be built on Oppo’s LLM and GPT configurations, which can support various use cases through models with 100 million, 300 million, and 1 billion parameters. Vivo also announced that their self-developed Blue LM will support their Xiao Vi Assistant, set to be released in November.
Revolutionary or Overhyped?
According to market researcher Canalys, global smartphone shipments declined by 12% to approximately 525 million units in the first half of 2023. China accounted for 124 million smartphone shipments during the same period, experiencing a 7% year-over-year decline. Additionally, since 2021, Xiaomi, Oppo, and Vivo—the three largest Chinese phone manufacturers—have seen a 1-2% decrease in global market share.
The decline in sales is prompting Chinese phone makers to seek new revenue sources, and generative AI appears to be the catalyst capable of reigniting user excitement.
However, one of the major challenges is integrating generative AI into already compact handsets.
Smartphones already have limitations regarding size, performance, and battery life. Implementing generative AI on a handset could impact response times and overall performance significantly. For instance, it might take two seconds to suggest the next word while the user is typing, and running large language models (LLM) on devices may quickly drain the battery.
Deploying LLM on the cloud presents its own set of tradeoffs. Personalizing the AI assistant becomes difficult, and ensuring privacy and security becomes a challenge. Additionally, the cost of running these models on the cloud is exceptionally high.
Chinese phone makers have chosen to use both local and cloud deployments to support different use cases. “Lighter” models are deployed on devices to handle less compute-intensive user needs that require customization, such as email composition and calendar customization. More compute-intensive tasks like vision and audio are handled by larger models stored on the cloud.
These constraints may be alleviated as smartphone chip manufacturers race to provide better solutions. MediaTek aims to develop advanced computing products that accelerate AI applications on phones and other devices. Qualcomm is also strengthening its edge computing solutions.
Ultimately, the success of generative AI in revolutionizing user experiences remains to be seen. Will having more intelligent driving directions or better email writing recommendations be enough to motivate consumers to buy new phones?
It is likely that the race to make smartphones truly “smart” will level the playing field among phone makers rather than driving them apart. As all phones become increasingly intelligent, manufacturers need to find new ways to attract consumers.