Google is out with its quarterlyPixel Feature Drop, giving Pixel owners reasons to think their phones are new again. The spotlight for December, though, is a new on-device generative AI model called Gemini Nano that’s exclusive to thePixel 8 Pro(for now). Here’s everything you need to know about the drop.
What is Gemini from Google?
To know what Gemini Nano is meant to do, we’ve got to explain what Gemini is first. In short, it’s Google’s newest full-fledged generative AI model that is intended to go beyond what large language models (LLMs) — the main target here isOpenAI’s ChatGPT— can do by processing not just text and code, but audio, images, and video, too.
Gemini would remove an interpretation layer between the AI model and other tools picking up text in those audiovisual contexts through methods such as optical character recognition. With the backing of its processing power, Gemini can speed up its own analysis and manipulation of the material with which it is working.
The Gemini Nano model is the most compact of the three versions on offer. Unlike larger iterations that are designed with scale and centralized servers in mind, Nano is intended to be installed onto consumer devices to help with local tasks.
The Pixel 8 Pro with Gemini Nano
That’s where we get to this Pixel Drop’s big exclusive for the Pixel 8 Pro running with its Tensor G3 processor. Some of the new features in this drop will be powered by the Gemini Nano model with all the work done on-device.
There are two big Gemini features thatGoogle is promotingand one of them isSummarize in Recorder.Generating a transcript and labeling speakers within your audio recordings isn’t new, but what is new is that you’ll be also be able to generate a summary of the transcript (much like whatBardcan do on Google Search and elsewhere) and perform all of these features offline through Gemini Nano.
Similarly,Smart Repliesare already well-established across Google services, but the company says the replies Gemini Nano generates are richer and more pertinent. You can try Gemini-powered Smart Replies through a developer preview using Gboard in US English on WhatsApp. You’ll need to follow theSmart Compose tool instructionsto activate and use the feature.
Other features in the drop
Yes, Google has other features that other Pixel phones will also get — it’s not like the people there have forgotten. Here’s a rundown:
Will Gemini Nano make it to other phones?
In a way, yes. Google has created a new system-level service in Android 14 that it callsAndroid AICore(viaThe Verge) where it houses Gemini Nano along with other engines as well as safety checks. The company is also laying out workflows to let app developers interface with AICore (and, thus, Gemini Nano) on users' devices.
That service is backed by the AI-related silicon of your device’s chipset, so in theory, you should be able to use Gemini Nano on an Android 14 device with sufficient AI processing power (think of flagship-level chipsets like Qualcomm’s Snapdragon 8 Gen 3 or MediaTek’s Dimensity 9300).
Still, the notion of a pint-sized generative AI model becoming part of a mobile operating system is nothing to sneeze at. We’ll just have to see how Google refines its approach to deployment as the days wear on.
