Qualcomm is set to release its new flagship mobile chipset, the Snapdragon 8 Gen 3, which brings generative AI capabilities right to the chipset, eliminating the need for cloud processing. This advancement signifies a significant leap forward in AI technology. The first phones with the Snapdragon 8 Gen 3 chipset are expected to be available in the coming weeks.
The Snapdragon 8 Gen 3 boasts a range of features powered by on-device AI. It supports a chatbot trained on Meta’s Llama 2, capable of accepting text, image, and voice input. Additionally, the chipset can generate images and text, as well as run the AI image generator Stable Diffusion. In a recent demonstration, Qualcomm showcased the chipset’s ability to generate an image in less than one second, significantly faster than previous technology.
The AI engine of the Snapdragon 8 Gen 3 is equipped with Qualcomm’s Hexagon neural processor, while the Sensing Hub utilizes OpenAI’s Whisper for speech recognition. The Sensing Hub provides the AI engine with user information, such as location and personal preferences, to deliver more personalized responses. However, the extent of user information being relayed, including favorite activities and fitness level, has raised some concerns about privacy and potential dystopian scenarios.
Generative AI plays a crucial role in the chipset’s new image processing capabilities. It supports generative fill for image expansion, allowing users to zoom out and re-crop photos directly on the device. The video features are even more impressive, with an object eraser for video and on-device night mode recording at up to 4K/30p. Google has also announced its own Night Sight for video feature, but it will rely on cloud processing.
Another notable feature of the Snapdragon 8 Gen 3 is Vlogger’s View, which combines video from the selfie and rear cameras to create a layered view. This feature uses improved image segmentation to remove the background from the selfie video, giving the impression that the user is standing in front of whatever the rear camera sees.
Recognizing the potential misuse of these AI tools, Qualcomm is collaborating with Truepic to ensure content authenticity. Truepic’s technology complies with the Coalition for Content Provenance and Authenticity’s open standard, providing cryptographic authentication to digital assets, such as photos or videos, to prevent tampering.
In addition to AI capabilities, the Snapdragon 8 Gen 3 offers efficiency gains with a modest 10 percent power savings compared to its predecessor. It supports Dolby’s HDR photo format for high dynamic range image capture. The X75 modem-RF provides enhanced support for 5G carrier aggregation, while the GPU supports hardware-based ray tracing for more realistic light reflections in mobile games. The chipset also supports up to 240Hz refresh rates on compatible external displays.
Qualcomm introduces a new system called Snapdragon Seamless, aiming to facilitate seamless device pairing across manufacturers and OS platforms. This system enables easy switching between devices and is particularly focused on improving connectivity between Android and Windows devices.
While Snapdragon Seamless is an open platform that anyone can join, initially, the technology will be mainly accessible to Qualcomm’s OEM and OS partners. However, Qualcomm envisions expanding the platform to foster greater collaboration and integration among various devices and ecosystems.
In conclusion, the Snapdragon 8 Gen 3 chipset represents Qualcomm’s latest advancement in AI technology, bringing generative AI capabilities directly to the device. With features such as on-device image and text generation, enhanced image processing, and seamless device pairing, the Snapdragon 8 Gen 3 offers a range of innovative functionalities. However, concerns about privacy and data security are essential considerations as AI technology progresses.
