Google moves toward ‘universal AI’: Gemini to gain context awareness and execution capabilities

Google is advancing its vision of bringing AI mode to all aspects of work and daily life. During its developer conference (Google I/O 2025), which kicked off this Tuesday in Mountain View, California, the company presented current and upcoming developments that essentially extend their achievements across all applications — but with greater accuracy, speed, and ease of use.

Demis Hassabis, researcher and CEO of Google DeepMind, sums up the concept, which he calls “universal AI”: “To be useful in your daily life, to be intelligent, to understand the context you’re in, and to be able to plan and act on your behalf on any device — this is our ultimate goal for Gemini [Google’s AI].”

The main advancements will begin to be available in the highest subscription tier (Ultra) at a cost of $249.99 per month. Microsoft has also announced similar advances in agents capable of reasoning and completing complex tasks for users.

Sundar Pichai, CEO of Google, boasts that one of this year’s achievements is the effective integration of its AI into the world’s most used search engine. He adds personalization, coding capabilities, advances in audiovisual content generation, and reduced latency in results retrieval. The executive points out that these capabilities come at a cost but defends that subscription fees for model pricing “are dropping significantly.” “There is a difficult balance between price and performance; however, time and again we have been able to deliver the best models at the most cost-effective price,” he argues.

“Reinvention” of search

Pichai previewed the launch of an AI mode that will be incorporated into the search engine to handle the exponential growth in its usage. “It’s completely new, a total reinvention of search with more advanced reasoning, answers to longer and more complex queries [up to five times the length of traditional searches], and the ability to go further with follow-up questions.” This new tab started rolling out this Tuesday in the United States before expanding globally.

Audiovisual advances

In the video field, Google is introducing Project Starline, a technology that recreates images simulating three dimensions. “The goal is to create the feeling of being in the same room with someone,” explains Pichai.

In this regard, Google Beam was unveiled, a tool that transforms two-dimensional video streams “into a realistic 3D experience” thanks to six cameras that capture and merge different angles in real-time. It can be used for video calls, but the first devices with this technology won’t be available until the end of the year.

Google also presented improved versions of Flash and Astra, AI tools that allow Gemini Live to interact with the device while it observes, memorizes, and analyzes the environment where the interaction takes place. These form the foundation of future Android XR glasses, an augmented reality device to access the agent integrated with the user. Similarly, the new version of VEO, the AI audiovisual creation platform, “combines video with audio for the first time,” Hassabis explained.

Translator

The Starlight feature will be added to Google Meet, enabling simultaneous translation of video calls (initially only available in Spanish and English). The AI adapts to the tone of the speakers and mimics their expressions. It will also be available by the end of the year for subscribers.

Agents

The evolution from chatbots to agents (tools capable of acting on behalf of the user) is based on Project Mariner, an agent that, besides planning, can execute different tasks simultaneously and learn from its actions to be proactive and anticipate user requests. It will be available starting this summer. “We are beginning to incorporate agent capabilities into search in Chrome, and the Gemini app will feature a new agent mode,” announced Pichai.

Work and study tools

Gemini’s advances, featuring personalization capabilities (adapting to the user’s characteristics), will also be integrated — starting this summer — into common workplace utilities such as Gmail, Docs, and Keep. Improvements will also be made for students, who will be able to use AI not only for occasional queries but also, as Hassabis explains, for “exam preparation, understanding materials, taking practice tests, and watching videos.”

Shopping

Vidhya Srinivasan, Vice President of Ads and Commerce, highlighted one of Google’s new shopping-oriented features that aims for AI to handle the entire process — from “inspiration” to payment and ordering. The AI won’t just show options — for example, clothing items — but can also, based on a personal photo, show how the item looks on the user and complete the purchase or hold it until it finds the desired price.

Microsoft and X

Google’s path mirrors that taken by the company founded 50 years ago by Bill Gates and Paul Allen. At Microsoft Build, the company’s annual developer event, Satya Nadella, CEO of Microsoft, announced the “open agent web,” a concept similar to Google’s, which allows AI agents to interact, decide, and act on behalf of people, teams, and organizations.

Microsoft unveiled updates to its programming environment to facilitate the creation of more capable and secure AI agents, boost scientific research, and promote open standards and shared infrastructures and protocols.

In this context, the company presented GitHub Copilot, a programming agent; Windows AI Foundry and Foundry Local, a unified platform for full and customized AI development — from training to inference (reasoning ability in new contexts); and Azure AI Foundry Models, among other new tools for model evaluation.

Microsoft also announced it has incorporated Grok 3 and Grok 3 mini models from xAI, Elon Musk’s company. Musk participated via video during the event and acknowledged past errors, which he said have been quickly corrected thanks to developer collaboration.

Grok has generated responses about the “white genocide” on X, Musk’s social network, even when the questions had no relation to South Africa, the magnate’s home country. For example, Jen Golbeck, a professor at the University of Maryland, received this reply from the AI: “The claim of white genocide is highly controversial. Some argue that white farmers face targeted violence, citing farm attacks and rhetoric such as the song ‘Kill the Boer,’ which they consider incitement.”

Red Hat

Red Hat, a global open-source solutions provider, has also released Enterprise Linux 10 after six months of testing. This platform is designed to meet the dynamic demands of hybrid cloud and artificial intelligence.

“More than just an update, Red Hat Enterprise Linux 10 provides a strategic and intelligent backbone for managing increasing complexity, accelerating innovation, and building a more secure computing foundation for the future,” the company said.

Red Hat credits its platform with the ability to integrate AI workloads with an operating system it describes as intelligent, resilient, and durable, as well as flexible and agile.

“The integration of generative AI directly into the platform helps provide contextualized guidance and actionable recommendations through a natural language interface,” according to the company, which asserts that this feature makes management easier for “both novice and experienced professionals.”

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition