Google Gemini’s announcement introduces three multimodal models, setting the stage where our focus naturally gravitates towards the most ambitious model, Gemini Ultra, while we pay less attention to the smaller one.

In this case, it makes perfect sense to start paying attention to the smaller model, because even though it may not be as promising as the larger one, it’s likely to have the most impact on us in the future. We’re talking about Gemini Nano.

Gemini Nano

Private, immediate, and independent of our connectivity

Gemini Nano is a model created with a focus on efficiency. It won’t be the end product: if Bard is like ChatGPT, Gemini is like GPT, the backend. The difference lies in the fact that Gemini Nano is designed to be used locally, on the device itself, without the need to connect to Google’s servers to obtain responses to our queries.

This not only means that we won’t need coverage to use it—a minor issue today—but also that the latency will be much lower, and the responses we get will be much more immediate, depending solely on the processing power of our device, not the speed of our connection.

It’s Somewhat similar, keeping a distance, to what Siri introduced in 2021 when iOS 15 debuted its ability to operate locally for everything that didn’t require an online query. However, with an LLM like Gemini, even if it’s Nano, we anticipate uses far beyond setting a timer, making a call, or asking for music from Queen!

Having much more immediate responses to the most essential requests to an LLM, such as asking it to summarize an email, generate a formal response accepting an invitation, or create a script for recording a TikTok explaining the rise of unsettling-flavored nougats, sounds quite appealing.

THE BEST WAY TO SWITCH FROM IPHONE TO ANDROID

Efficiency

And not just for speed, but also for efficiency. Someone once said that using ChatGPT to summarize an email was like driving a Lamborghini to buy bread. Excessive use of resources for a trivial task. If we assume that’s true, and we know that queries to ChatGPT and its generated responses are not free, having a local LLM even on a smartphone translates to many saved queries on external servers.

Deep Integration

This also opens the door to integrating Gemini Nano at any point on our smartphones. It could be used to respond to a WhatsApp message contextually, summarize conversations (given its multimodal nature, which can include voice or text), or even analyze images to understand their content.

TOP AI ART GENERATOR: ARTIFICIAL INTELLIGENCE TO CREATE IMAGES

If the script unfolds as expected, Apple will announce its upcoming operating systems in June 2024 during its annual developer conference. This event is anticipated to unveil the rumored new capabilities surrounding Siri, particularly focusing on generative AI to prevent Microsoft, Google, and OpenAI from gaining too much ground in what is perceived as the major race of this decade.

As of now, Gemini Nano will only be available for the Pixel 8 Pro, apparently because only its Tensor G3 is currently equipped to support it. However, the goal is for this to eventually become integrated into Android.

This assertion isn’t only supported by speculation, but also by the explicit intent expressed by AICore, a system service aimed at incorporating Gemini Nano functions directly into the device. Mediatek, Qualcomm, and Samsung have already been announced as compatible chip manufacturers.