At Google I/O 2024, Google announced integrating its smallest AI model, Gemini Nano, into Chrome desktop starting with version 126.
Google revealed on Tuesday at the Google I/O 2024 developer conference that it is integrating Gemini Nano, the tiniest of its AI models, right into the Chrome desktop client, beginning with Chrome 126.
This will allow developers to fuel their own AI capabilities with the on-device model, according to the company. Google plans to use this expanded capacity to power features such as the Workspace Lab “help me write” function that is now available in Gmail.
These models may run reasonably fast on a variety of hardware thanks to recent work on WebGPU and WASM support in Chrome, according to the business.
In a briefing before Tuesday’s release, Google’s director of product management for Chrome, Jon Dahlke, stated that the company was in discussions with other browser suppliers to enable this—or a feature similar to it—in their browsers as well.
Dahlke stated in the release on Tuesday, “We have begun to engage with other browsers and will be opening up an early preview program for developers.” “We think the web is AI-ready because Chrome has webGPU, WASM, and Gemini built in.”
However, I don’t think the majority of Chrome rivals would want to place all of their money on Google’s AI models. Allowing browsers and developers to use the model of their choosing makes more sense. Although it is obvious that Google would prefer to use Gemini for its apps, these models are sufficiently compact to allow developers to use whichever one they prefer.
However, Google plans to leverage its Gemini models to activate a number of high-level APIs in Chrome, enabling the browser to translate, caption, and transcribe text.
Dahlke stated during today’s I/O developer conference, “To deliver this feature, we fine-tuned our most efficient version of Gemini and optimized Chrome.” We now wish to grant you Chrome access to the Gemini models.
Our goal is to provide you with the most powerful AI models in Chrome so you can reach billions of consumers without worrying about capacity, cost, timely engineering, or fine-tuning. Just make a few high-level API calls to translate, caption, and transcribe the content. We want to do this right because it is a significant change for the web.
Google is now powering some new features in the Chrome DevTools Console for developers utilizing the integrated Gemini Nano model. As a result, issues may now be explained, and debugging solutions can be found directly in the console using the Chrome development tools.