Google teases upgraded Gemini Nano AI for Pixel 9 series

Google‘s Pixel 9 series will introduce an upgraded version of Gemini Nano, the company announced at its annual developer conference, aka Google I/O, on Tuesday. The new version will boast multimodal capabilities and can discern context from images, videos, sounds, and spoken language. It requires more powerful hardware, specifically the Neural Processing Units (NPUs), and might not be compatible with older Pixels.

Pixel 9 to bring an improved Gemini Nano

Gemini Nano is a mini version of Google’s generative AI model Gemini (formerly Bard). It is designed to run on smartphones and is the brain behind certain on-device AI features on the Pixel 8 series, including the Pixel 8a. The tool powers the Summarize feature in the Recorder app and Smart Reply in Gboard, among other things. However, its functionality is mostly limited to processing text inputs.

With the Pixel 9 series, Google plans to add new capabilities to Gemini Nano. A teaser posted by the company shows the AI tool explaining the scene in front of the device using its camera. “Hey, what do you think is happening here?” Gemini Nano is asked, pointing to the stage before the Google I/O kicked off. It identified the scene. “It looks like people are setting up for a large event, perhaps a conference or presentation,” it replied.

The tool goes on to converse about the scene with the user, responding to various questions. The demo suggests the upgraded version of Gemini Nano can identify real-world images and discern context accurately. Google plans to use the tool to improve its TalkBack feature to help people with blindness or low vision. It can describe unlabeled photos better than before without requiring an active internet connection.

See more

One more day until #GoogleIO! We’re feeling 🤩. See you tomorrow for the latest news about AI, Search and more. pic.twitter.com/QiS1G8GBf9

— Google (@Google) May 13, 2024

The new version may not roll out to older Pixels

Google didn’t specify whether it plans to roll out the upgraded Gemini Nano to older models following the Pixel 9 launch this fall. However, it dropped a big hint that the new AI tool might not be compatible with older Pixels. The company said it requires new Neural Processing Units (NPUs) and more RAM, and that it will be available “starting with the new Pixels later this year.” This likely means the Pixel 8 and others are out of luck.

That said, Google is working on compact AI models to power on-device AI features on a wider range of devices. The company will share a list of devices compatible with the improved Gemini Nano later this year. It remains to be seen whether the new capabilities trickle down to older Pixel flagships. The Pixel 9 series is still several months away. We might hear more about Google’s plans in the coming months.

May 15, 2024