Sundar Pichai says Gemini 3 reads, watches and listens to help learners

Sundar Pichai, CEO of Google, has unveiled the impressive capabilities of the company’s newest AI model, Gemini 3. In a recent post on X, he detailed five innovative features designed to enhance user interaction and simplify everyday tasks. One standout ability of Gemini 3 is its versatility in understanding various input types, including photos, PDFs, sketches, and diagrams. Users can present a simple doodle on a napkin, and Gemini 3 can transform it into a fully functional website. Additionally, it can convert a basic image into a board game or develop an interactive lesson from a diagram. The model also has the capability to analyze lengthy videos, breaking them down into digestible segments. For instance, it can review sports videos to identify areas for improvement and propose specific drills to enhance your skills. Thanks to advancements in visual and spatial reasoning, Gemini 3 is better equipped to assist users in these endeavors. Pichai also highlighted improvements in how Google Search operates with Gemini 3. Beyond providing text-based answers, the model can now generate visual layouts and interactive tools. When queried about complex topics, such as the three-body problem in physics, users can expect simulations that enhance understanding. Search results are now more engaging, featuring photos, interactive modules, and scrollable itineraries for planning trips tailored to personal preferences. In addition, Google is launching Gemini Agent, a smart assistant designed to streamline daily tasks. This tool can manage emails, draft responses, archive old messages, and help users book local services autonomously, suggesting useful actions along the way. It will initially be available to Google AI Ultra subscribers in the United States. Gemini 3 has been crafted to comprehend a multitude of information types simultaneously, including text, images, videos, audio, and even programming code. Its enhanced reasoning capabilities enable it to process information more clearly and manage longer inputs seamlessly. For example, users can input handwritten family recipes in various languages, and Gemini 3 can read, translate, and compile them into a neatly organized family cookbook. Whether it's processing research papers, video lectures, or tutorials, Gemini 3 can generate valuable educational tools like interactive flashcards and visual aids to facilitate learning. Additionally, it can analyze videos of users engaging in activities, such as pickleball, to identify improvement areas and create comprehensive training plans to enhance performance.