March AI News Announcement

# ai # Gemini # Shopping # research # Pixel # Search

For over two decades, we've been diving deep into machine learning and AI, pouring resources into research, tools, and infrastructure to create products that enhance everyday life for more folks. Google teams are exploring ways to harness AI's potential across diverse sectors like healthcare, crisis response, and education. To keep you in the loop, we're rolling out a regular roundup of the latest AI news from Google, spanning products, research, and beyond. Let's take a quick peek at some of our AI highlights from March. [ttpp][yyxx] March was a big month for expanding AI across our products, aiming to make them even more helpful. We upgraded the Gemini app with new features like personalization, Canvas, and Audio Overviews, and made our Deep Research and Gems features free for everyone. We also slipped in our speedy Gemini 2.0 Flash Thinking experimental model behind the scenes — and then topped it off by rolling out our smartest AI model yet, Gemini 2.5 Pro (experimental), to all Gemini users just a few days later. Gemini 2.5 Pro is top-notch on a wide range of benchmarks — give it a whirl in Gemini or Google AI Studio. Keep reading to see how we're making our leading AI even more useful for Pixel users, online shoppers, roboticists, developers, wildfire authorities, and more. [ttpp][yyxx] We broadened access to AI Overviews and rolled out AI Mode. AI Overviews are a hit in Search, now used by over a billion people. We kicked off Gemini 2.0 for AI Overviews in the U.S., tackling tougher questions, starting with coding, advanced math, and multimodal queries for faster, higher-quality answers. We also introduced the AI Mode experiment in Google Search to help users get AI-powered responses and dive deeper with follow-up questions and links to useful web content. We added personalization to Gemini to tailor its responses to you. With personalization, Gemini can use your Search history to provide contextually relevant answers that match your interests. With your okay, Gemini can now fine-tune its responses based on your past searches, saving you time and giving you more precise answers. In the coming months, Gemini will enhance its understanding of you by linking up with other Google apps and services like Photos and YouTube. We rolled out updates for Gemini Live, Scam Detection, and more in our March Pixel Drop. Our first Pixel Drop of the year brought more helpful features and updates to your devices. Gemini Live, which helps you brainstorm, simplify complex topics, and rehearse for big moments, got an upgrade for better performance in multilingual conversations. Now you can switch between over 45 languages seamlessly when chatting with Gemini Live. Plus, enhancements like stronger Scam Detection, better step accuracy, and Auto-bedtime Mode bring even more AI to your Pixel. We launched new AI tools for Google Shopping to help you snag the perfect products. Our new immersive shopping features, like our vision match tool, can help you find clothes and beauty products that suit your style. With vision match, just describe the garment you're thinking of, and get an AI-generated image showing what it might look like, along with similar items you can buy. It's another way we're making shopping easier and more personalized for you. [ttpp][yyxx] We introduced Gemini Robotics to bring AI into the physical world. We unveiled two new AI models based on Gemini 2.0, setting the stage for a new generation of helpful robots. The first, Gemini Robotics, is an advanced vision-language-action (VLA) model built on Gemini 2.0 to control robots directly. The second, Gemini Robotics-ER, is a Gemini model with enhanced spatial understanding, allowing roboticists to run their own programs using Gemini's embodied reasoning (ER) capabilities. [ttpp][yyxx] We released Gemma 3 to help developers build even more helpful applications. Gemma 3, the latest iteration of our lightweight, state-of-the-art open models, can run on a single TPU or GPU. These models are designed to be our most advanced and portable open models for developers, running fast and directly on devices — from phones and laptops to workstations — helping developers create AI applications wherever they're needed. [ttpp][yyxx] The first FireSat satellite for early wildfire detection made contact with Earth. Launched from Vandenberg Space Force Base in California on SpaceX's Transporter-13 mission, this satellite — a joint effort with Muon Space, Earth Fire Alliance, Moore Foundation, wildfire authorities, and others — is the first of over 50 in a pioneering constellation designed to use AI to spot and track wildfires as small as about 5x5 meters. We also kicked off three new initiatives to protect and restore nature using AI. A startup accelerator, set to launch in May 2025, will offer programming, mentoring, and technical support from Google. Google.org is providing $3 million to support AI-enabled solutions for biodiversity, bioeconomy, and agriculture from Brazilian nonprofits. And we released SpeciesNet, a Cloud-based, open-source AI model for identifying animal species from camera trap photos, helping people protect nature and biodiversity.

Related article

Comments (55)