Honor’s New Phone Should Bring Generative AI to Your Fingertips – CNET [CNET]

View Article on CNET

The Honor Magic 6, which is rumored to launch in 2023, will run on Qualcomm’s new Snapdragon 8 Gen 3 chip, the companies announced Wednesday. The new chip will allow Magic 6 owners to tap into generative AI wherever they are, even offline.

The buzz around generative AI tools started last November when OpenAI released ChatGPT. Since then, companies like Google and Microsoft have been in a rush to bring generative AI tools to computers and phones.

Qualcomm touts the Snapdragon 8 Gen 3 chip for its boosted privacy and faster speeds. But it also runs on-device large language models, which is similar to the technology behind ChatGPT and that can be used for more features built around personalization.

A phone takes a photo of a mountain landscape with the Snapdragon 8 Gen 3 chip logo in the corner.
Qualcomm announced the Snapdragon 8 Gen 3 at its Snapdragon Summit this week. 

Qualcomm

Honor says the Magic 6’s LLM is different from cloud-based LLMs, since those are trained on publicly accessible datasets. The Magic 6 “draws from its understanding of the device’s user to provide personalized services according to their preferences.”

Honor showed using the Magic 6’s AI capabilities via the phone’s assistant, Yoyo, to create a short video featuring photos and footage stored on the phone.

Apart from Honor, upcoming flagship phones from the likes of Xiaomi, Oppo and OnePlus are set to run on the Snapdragon 8 Gen 3 chip. However, those companies will need to work with Qualcomm to integrate AI features into their phones as they will not be natively available.

Earlier this month, Google launched the Pixel 8 and Pixel 8 Pro, both of which are brimming with AI features for improving photos, answering phone calls and even making custom wallpapers.

snapdragonsummit23-supercut-clean

Watch this: Everything Announced at Qualcomm’s Snapdragon Summit

Editors’ note: CNET is using an AI engine to help create some stories. For more, see this post.