Just Determined

Posted on May 21

Google I/O Keynote 2025 Summary: Gemini Deep Think, Search AI Mode, Android XR, Jules

#google #gemini

Google I/O Keynote 2025 Summary: Advancements in AI, Search, and XR.

Ironwood 7th Generation TPU(Tensor Processing Unit): Designed to power thinking and inference at scale with 10X performance over previous generation. Coming to Google Cloud customers later this year.

Update: Gemini Now Has over 400M monthly users.

Gemini 2.5 Pro: Optimized for quality and complex tasks. It includes an enhanced reasoning mode called Deep Think
Gemini 2.5 Flash: optimized for speed and cost-efficiency, particularly for high-volume applications.
Jules: Google's AI Coding Agent. Jules autonomously reads your code performs tasks like writing tests and fixing bugs.
Gemini Live: This feature, leveraging Project Astra, rolled out on Android and iOS already, allowing real-time conversations with Gemini about what's on the screen or through the phone's camera. Details & Demo Here

Imagen 4: Google's next-generation text-to-image model was unveiled, producing visuals with greater detail and improved text rendering within images. It's rolling out in the Gemini app.
Veo 3: The newest text-to-video model can now generate fluid motion graphics. with better visual quality, native audio generation, soundtracks, sound effects that match the visuals, and dialogue.

Flow: A new AI filmmaking tool that combines Imagen 4 and Veo 3, offering features like camera movement and perspective controls, scene extension, and integration of AI-generated video with music from Google Lyria. Details and Demo Here
Music AI Sandbox: For creating music tracks, spark new creative possibilities and help artists explore unique musical ideas.
Lyria 2: Improved Music generation, Can do rich classical music and add quality vocals.
Gemini App Enhancements: The app is getting an Agent Mode to retrieve and summarize web data in real time for multi-step tasks.
* Gemini in Chrome: Your AI Assistant on the web. It will understand the content of your current page and you can ask it anything based on your current page or even summarize for you.

Update: Google Lens now has over 1.5B monthly users.

AI Mode on Google Search: This new AI-powered Search will allow users to ask longer questions and offer more interactive, personalized, and visually rich results, including AI-generated graphs for complex queries and a "try it on" feature for apparel shopping, which will allow users to virtually try clothing before they shop. Details & Demo Here
Personalization in Search: AI Mode will incorporate past searches and connect with other Google services like Gmail for more contextual results.
Deep Research Upgrade: Users will be able to upload files and images for more contextual and detailed responses. It's also being integrated with Canvas for creating dynamic infographics, website, podcast and more. Canvas is just next to Deep Research on Gemini.
Search Live: Utilizing Project Astra capabilities, this feature will allow users to ask questions about what their camera sees. Coming to AI Mode on Google Search, this summer. Details & Demo Here

In patnership with Samsung(Project Moohan) and Qualcomm

Google previewed its evolving Android XR platform for headsets and smart glasses with camera, speaker, microphone and Gemini, demonstrating smart glasses for messaging, navigation, and real-time language translation using built-in lens displays.

Google has partnered with eyewear brands, Gentle Monster and Warby Parker will be the first to develop Android XR-based smart glasses.
A demo showcased live language translation on Android XR glasses.

Through Project Mariner, Google demonstrated agentic AI systems that can perform multi-step tasks like booking tickets or making online purchases autonomously.

Through Project Starline , Google Beam is a new 3D teleconferencing technology using a six-camera array and a custom light field display for realistic 3D representations during calls.

Firesat: A partnership led by Earth Fire Alliance to use AI to create a breakthrough in wildfire detection. FireSat uses high-res multispectral satellite imagery and AI to provide near real-time insights on wildfires. Provides global high resolution imagery that is updates in every 20 minutes.
SynthID Detector: A new portal to help identify AI-generated content in images, audio, video, and text by scanning for the SynthID watermark.
Google AI Ultra: A new subscription plan offering the highest access to Google's most capable AI models and premium features.
Personalized Smart Replies in Gmail: Gemini models can now generate smart replies tailored to your writing style by learning from your emails and Drive documents (with your permission).
Real-time Translation in Google Meet: This feature can translate speech in near real-time, matching the speaker's voice, tone, and expressions. English and Spanish translation is rolling out in beta for Google AI Pro and Ultra subscribers.
Chrome Password Manager Upgrade: It will automatically change passwords on accounts compromised in data breaches.

These were some of the major announcements from yesterday's Google I/O Keynote. The event continues today with more developer-focused sessions.

Follow Tech Psyche on Telegram for more updates.

DEV Community