Qualcomm presented to the MWC (Mobile World Congress) of Barcelona a series of innovations, which mainly concern the use of on-device AI. Give it to him smartphone and gods laptopcon Snapdragon 8 Gen 3 and the platform Snapdragon X Elite, it becomes possible to use different types of artificial intelligence models. And developers can build applications and services by leveraging the Qualcomm AI Hub. Furthermore, there are also new features for modems 5G e Wi-Fi 7to take advantage of the fast connection to exploit the Ai (and vice versa).
Qualcomm and on-device AI: what’s new at MWC 2024
Today, February 26in Barcelona the chip company announces the birth of Qualcomm AI Huban innovative platform that offers a library of templates AI optimized for solutions Snapdragon. An important choice: Qualcomm has already demonstrated that its platforms, like Snapdragon 8 Gen 3, know how to operate different AI models directly on-device. There are already examples of smartphones that do this successfully. But giving developers access to different templates will allow multiply that potential: we will be able to use more applications and services on our devices.
Con 75 AI models At launch, Qualcomm AI Hub promises to deliver a inferencing faster than 4X. Access for developers seems simple to us: just select the model AIs that want to leverage, the platform (both the Snapdragon chip and the device), and then implement the solutions AI directly on-device.
Qualcomm executives also explained to us that the models will also be available on Hugging Face e GitHub, as well as on the company’s Hub. And to demonstrate to everyone that the possibilities of fully exploiting AI on devices is real, the company also announced several demonstrations in Barcelona.
Large Multimodal Model (LMM) directly on the device
Il MWC it will be the scene of the presentation of the first Large Multimodal Model su sistema Androidwith over seven billion parameters, capable of processing text, images and speech directly on your device. This innovation not only promises faster speeds, without the need to bring computing to the cloud. Furthermore, it also allows for special attention to privacyallowing the processing of private images without sending them to external servers.
Qualcomm will do a similar demonstration with the former LMM su PC Windowsextending use beyond the borders of smartphone. A complete experience that includes the use of text and audio, enabling features such as music recognition and voice-guided conversationsAI.
Qualcomm and on-device AI at MWC: LoRA on an Android phone
The. also comes from Barcelona first test of Stable Diffusion with LoRA (Low Rank Adaptation) on Android phone. Use over a billion parameters Stable Diffusionto allow you to create custom images without having to download entire templates AI on phones. An intelligent method to ensure functionality AI robust with less use of resources.
In Barcelona, Qualcomm will also test sui laptop Snapdragon X Elite plugin integration Stable Diffusion con GIMP, comparing it to an x86 processor. According to the company, it should generate images three times faster.
Fast connections for hybrid AI
Qualcomm executives also told us that the company is a big believer in the concept Hybrid AIwhich sees part of theGenerative AI operate on cloud (for example, during training and aggregation), a part aboutedge cloud and one part directly on-device. In other words, some operations can remain in the cloud, while others are more effective if activated on the device. The important thing is that these different environments talk to each other.
Connectivity becomes the heart of this approach. The presentation of Qualcomm FastConnect 7900a chip for connections AI-enhanced which integrates Wi-Fi 7, Bluetooth e Ultra Widebandopening the doors to new technologies and applications based on proximity sensors, such as Ultra Wideband technology, Wi-Fi Ranging and Bluetooth Channel Sounding.
Qualcomm annuncia Snapdragon X80 5G Modem-RF
Qualcomm also announces the seventh generation of Snapdragon X80 5G Modem-RFcapable of managing frequencies mmWave e Sub-6 GHz. The management of 6 operators and 6 antennas through AI promises greater energy efficiency and increases coverage. With the 60% faster CPE service acquisition and the 10% reduction in consumption, Qualcomm goes a step further in providing connectivity 5G accessible to all.
Qualcomm also explains that this chip integrates the NB-NTN connectivity, to ensure speed even with the satellite network.
Making 5G more accessible
Qualcomm also explained to us that, in the near future, it will provide technologies to make 5G more accessible. In particular, we will look at one Gigabit 5G smartphones under $100with the aim of bringing connectivity 5G beyond 2.8 billion users.
The company also announced the technology Qualcomm 5G Fixed Wireless Access Ultra Gen 3which allows you to offer the mmWave 5G through FWA, ensuring high-speed wireless connections like fiber optic networks. Something that for Qualcomm is a crucial step in reducing the digital divide.
In the end, Qualcomm announces the introduction of processors for Open RAN and 5G cellswhich leverage the company’s experience in developing CPU e NPU. Qualcomm promises a reduction in costs and an improvement in performance performance for telecommunications operators.
These are all the AI and 5G news announced by Qualcomm in Barcelona: you can find more information here.
