Qualcomm engineers achieved a groundbreaking feat by deploying the text-to-image AI model, Stable Diffusion, on a smartphone. This achievement, demonstrated on a Sony Xperia 5 II handset with a Qualcomm Snapdragon 865 processor, 8GB RAM, and 30GB+ storage, generated 512 x 512 pixel images in under 15 seconds, offering a significant advancement for AI and smartphones.
Traditionally, such models require substantial cloud-based computing resources, but this innovation brings generative AI to smartphones, offering cost-effectiveness and the potential to revitalize the smartphone market, according to Qualcomm CEO Cristiano Amon.
Generative AI on smartphones could revolutionize how we interact with these devices, paving the way for unprecedented app development possibilities. Language models (LLMs) could even transform smartphone keyboards, as exemplified by Apple’s use of a transformer language model for autocorrect at WWDC 2023.
Trey Connolly, VP, Data & Analysis at Digitas North America weighed in on the matter, “With all the buzz around generative A.I. It is not out of the ordinary to ponder how smartphone developers & large manufacturers will embrace this technology to innovate, and provide a smarter user experience to customers. The article shares groundbreaking developments that could have far-reaching implications for both a new era of personal A.I. and further evolution of the smartphone industry.”
Furthermore, generative AI can personalize user interfaces, app layouts, recommendations, and user experiences, effectively turning your smartphone into a personalized assistant. This local AI processing reduces latency, making real-time applications more efficient, especially in areas with limited connectivity or expensive data plans.
He further added, “The ability to deploy generative AI models on smartphones addresses a fundamental challenge within the industry: cost-effectiveness. It eliminates the need for constant communication with cloud servers, reduces data consumption, network latency, and could improve functional operations of the phone tied to connectivity. Outside of these vast improvements there is room for even more potential.”
Running AI models locally on smartphones is cost-effective, as users have already paid for the hardware upfront, unlike cloud-based solutions, which can be costly. Qualcomm, Google, MediaTek, and Meta are partnering to make on-device generative AI implementations a reality, which could lead to significant cost reductions.
Additionally he said, “Think about the potential for generative AI to revolutionize the smartphone and how you personally interact with it. From enhancing predictive text input to creating personalized user interfaces, it can provide a highly tailored and efficient experience for the end user. The technology has the potential to transform our smartphones into stellar personalized digital assistants, capable of autonomously composing emails and automatically adapting to user preferences on the fly.”
This shift could democratize AI on a global scale, benefiting over 6.5 billion smartphone users worldwide. It also improves data efficiency and responsiveness, enhancing virtual assistants, chatbots, and predictive text input.
For AI companies, running models locally on devices could reduce operational costs significantly. Qualcomm’s efforts to create chips that enable AI processing on smartphones positions them to benefit from this AI revolution much like NVIDIA did with GPUs.
The Qualcomm Summit in October promises new chips that can run foundational generative AI models, offering a cost-effective alternative to cloud-based solutions. As Qualcomm continues to leverage its expertise in smartphone processors, it may become a prominent player in the AI hardware industry. Emad Mostaque even envisions advanced language models on mobile phones in the near future, making this a compelling development to watch.
He concluded by saying that, “The article details collaborations between companies like Qualcomm, Google, Meta, and Mediatek to make this technology available on a wide range of devices and shows its promise & potential. Qualcomm’s dedication to creating chips that enable AI processing on smartphones aligns with the industry’s shift towards edge computing. A smart move for the chip manufacturer, which will position them well in the marketplace for years to come. It is clear that these recent developments are worth watching and could be closer than all of us might think.”