Qualcomm showcased AI running natively on all kinds of devices at the ongoing MWC 2024 expo in Barcelona, Spain. It also displayed all the AI-related solutions it developed over the past year. It also showcased all the commercially available products that are using Qualcomm's solutions to offer AI features.
Qualcomm AI Hub supports 75+ AI models, making it easier for developers to integrate them into apps
The US-based chip giant unveiled Qualcomm AI Hub. It contains a library of pre-0ptimized AI models for seamless deployment of apps and services on devices that are powered by Snapdragon processors. This huge library consists of 75+ AI and Generative models, including Baichuan 7B, ControlNet, Stable Diffusion, and Whisper, which are optimized to perform well on devices locally. Each model is optimized to take advantage of the AI accelerators inside the Qualcomm AI Engine (CPU, GPU, and NPU), resulting in 4x faster inferencing. Developers can easily integrate these models into their apps, resulting in faster app development, higher performance, improved optimization, and lower cost.
These optimized AI models are available on GitHub, Hugging Face, and Qualcomm AI Hub. The company says support for more AI models will be added to more operating systems and platforms in the future. Developers can sign up today to run AI models on the cloud on devices using Qualcomm processors.
AI running natively on Android phones and Windows laptops
The company also demonstrated the Large Language and Vision Assistant (LLaVA) LLM (Large Language Model) for the first time on an Android smartphone. It has 7+ billion parameters and can accept multiple input types, including images and text. It can generate multi-turn conversations with an AI assistant about images. It runs at a responsive token rate, which means it has enhanced privacy and reliability.
It also showcased the Low Rank Adaptation (LoRA) LLM running on an Android smartphone. The company ran Stable Diffusion with LoRA, which allows users to create high-quality images with text prompts. They can enter their personal or artistic preferences to get desired results. Qualcomm also showcased a 7B+ parameter LLM running on a Windows PC that can accept text and audio inputs. The company showcased AI running natively on the HONOR Magic 6 Pro (AI-powered calendar and video creation), OPPO X7 Ultra (Image Object Eraser), and Xiaomi 14 Pro (AI-Generated Image Expansion).
A laptop with the Snapdragon X Elite processor is used to showcase how the device can create an AI-generated image within 7 seconds. That is 3x faster than solutions from x86 competitors (AMD and Intel). Qualcomm used the free and open source image editor GIMP with the Stable Diffusion plugin.
Qualcomm showcased AI-powered features on cars and consumer IoT
The company also showcased AI and Generative AI capabilities of the Snapdragon Digital Chassis Platform for connected cars. It can offer personalized experiences to drivers and passengers in an efficient, powerful, safe, and reliable manner. At its booth at MWC 2024, Qualcomm also showcased the Humane AI Pin, a connected mobile device that is powered by a Snapdragon processor. It uses an entirely new, screen-less way to experience AI and find information about things.
AI also improves connectivity and telecom infrastructure
Qualcomm also unveiled the Snapdragon X80 Modem-RF System that uses the second-generation 5G AI processor to offer enhanced cellular network coverage, latency, performance, and power efficiency. It also showcased the FastConnect 7900 Mobile Connectivity System, which is the first AI-optimized Wi-Fi 7 system in the world, increasing the bar for adaptable, high-performance, low latency, and low-power wireless connectivity.
To make things easier for cellular network operators, Qualcomm showcased three technologies. Generative AI assistant for Radio Access Network (RAN) makes it easier for engineers to simplify network and slice management tasks. Its AI-based Open RAN application (rApp) reduces power consumption during network management. The company also showcased its AI-based 5G network slice lifecycle management suite.
