Google Doubles Down on AI: From Devices to Creativity
Date
May 21st, 2025
Reading Time
10 mins
Tags cloud
What's news
1. The latest Google Gemma AI model can run on phones
May 20, 2025
Source: TechCrunch
At Google I/O 2025, Google expanded its family of open AI models, Gemma, by unveiling Gemma 3n, a lightweight multimodal model capable of running on devices with less than 2GB of RAM — including phones, laptops, and tablets. It processes audio, text, images, and videos offline, promoting privacy and efficiency.
Google also introduced MedGemma, a specialized model for analyzing health-related text and images, and SignGemma, designed to translate sign language (especially American Sign Language) into spoken-language text to aid deaf and hard-of-hearing users.
Despite concerns about Gemma's restrictive licensing for commercial use, the models have been widely downloaded by developers.
2. Google updates the Gemini app with real-time AI video, Deep Research, and more
May 20, 2025
Source: Gemini
At Google I/O 2025, Google announced major updates to its Gemini AI chatbot. One highlight is the expansion of Gemini Live, which now supports camera and screen sharing on iOS and Android. This allows users to have real-time conversations with the AI while showing it what they see through their phone. Deeper integrations with Google Maps, Calendar, and Tasks are also on the way.
Google introduced two new subscription tiers: Google AI Pro ($20/month), a rebrand of Gemini Advanced, and Google AI Ultra ($250/month), which offers high usage limits, exclusive features, and early access to new models. Pro and Ultra users in the U.S. will also be able to use Gemini directly in Chrome for summarizing and interacting with on-screen content.
Gemini’s Deep Research tool is being upgraded to let users upload PDFs and images for personalized, cross-referenced reports. Soon, it will also integrate with Gmail and Google Drive. Free users are getting access to an improved image model, Imagen 4, while Ultra subscribers gain Veo 3, Google’s latest AI video model with native sound generation.
To improve overall performance, Gemini now uses the faster and more efficient Gemini 2.5 Flash as its default model. For students, Gemini will offer personalized quizzes and follow-up plans to help strengthen weak areas.
Gemini now has 400 million monthly active users, and these updates are part of Google’s strategy to compete with ChatGPT, Siri, and other digital assistants.
3. Stitch is Google’s AI-powered tool to help design apps
May 20, 2025
Source: The Verge
Google introduced Stitch, an AI-powered tool designed to help developers and designers quickly generate front-end UI for web and mobile apps. Users can prompt Stitch with a few words or an image, and it outputs HTML and CSS code, powered by either Gemini 2.5 Flash or Gemini 2.5 Pro models.
Stitch supports exporting designs to Figma, allows code refinement in IDEs, and includes customization features. It’s aimed at simplifying the early stages of UI design, not replacing full design platforms like Figma or Adobe XD. A future update will let users modify designs by annotating screenshots.
Google also expanded access to Jules, an AI coding assistant that helps fix bugs, understand code, and manage tasks like GitHub pull requests. Now in public beta, Jules can automate complex upgrades, such as migrating from Node.js 16 to 22, and verify the results. It currently runs on Gemini 2.5 Pro, with support for more models coming soon.
These tools reflect Google's push into the growing “vibe coding” trend and its effort to compete with OpenAI’s Codex and Microsoft’s GitHub Copilot.
4. Google Meet is getting real-time speech translation
May 20, 2025
Source: Google
Google announced real-time speech translation for Google Meet, powered by a large language audio model from DeepMind. The feature translates spoken language into the listener's preferred language in real time, while preserving voice, tone, and expression.
The technology allows for natural, low-latency conversations between speakers of different languages, enabling smoother communication for families and global teams. Users will still hear the original voice faintly, with the translation layered over it.
The feature begins beta rollout this week for consumer AI subscribers, starting with English and Spanish, and expanding soon to Italian, German, and Portuguese. Business testing via Workspace will follow later this year.
5. Google debuts an AI-powered video tool called Flow
May 20, 2025
Source: Google Blog
Google unveiled Flow, a new AI-powered video creation tool designed for filmmakers. It combines three of Google’s models: Veo (video generation), Imagen (image generation), and Gemini (text and prompting).
Flow lets users import or generate characters and scenes, control camera angles, and use a scene builder to direct shot flow. It also includes asset management tools and a feature called Flow TV, showcasing videos with their original prompts to inspire creators.
Initially available in the U.S., Flow is accessible through the Google AI Pro and AI Ultra plans. Pro users can create up to 100 videos per month, while Ultra users get higher limits and early access to Google’s latest video models.
Tags cloud
Newsletter
DISCOVER MORE
![[Recap] UPP Global Technology JSC Establishing Anniversary](/homepage/news-section/new-4.webp)
LET’S TALK...
Content delivered to your inbox
ENTER YOUR EMAIL
YOU WANT TO...
Hanoi, Vietnam
Web3 Tower, No. 15, Alley 4, Duy Tan, Cau Giay, Hanoi, Vietnam