Like a master chef perfectly executing multiple dishes simultaneously, Google unveiled a massive collection of AI innovations that could reshape the entire industry in just one day.
Google announced revolutionary AI-powered features at its annual developer conference ‘I/O 2025’ held on May 21st (local time) in Mountain View, California, according to international media reports. This announcement, ranging from major upgrades to the Gemini AI model to completely new search experiences, is being evaluated as opening new horizons in AI technology.
Gemini 2.5 Evolves into ‘Thinking AI’
Google’s core AI model, Gemini, has become significantly smarter. The most notable update is the ‘Deep Think’ mode. Like a student trying multiple approaches when solving math problems, this is an advanced reasoning feature where AI examines multiple hypotheses before responding.
Gemini 2.5 Pro with Deep Think mode achieved a 49.4% accuracy rate on 2025 U.S. Mathematical Olympiad problems. This is a significant improvement from the existing Gemini 2.5 Pro’s 34.5%. It also showed top performance on LiveCodeBench, which evaluates complex coding tasks.
The Gemini 2.5 Flash model has also been upgraded. This model, specialized for speed and efficiency, uses 20-30% fewer tokens while significantly improving performance. It’s currently available for free in the Gemini app and is scheduled for official release to developers and enterprise customers in early June.
Search Revolution with AI Mode
Fundamental changes have occurred in Google Search as well. The newly introduced ‘AI Mode’ goes beyond traditional link-based search to provide conversational search experiences. Users can receive detailed answers to complex questions as if directly conversing with an expert.
The core technology of AI Mode is ‘query fan-out.’ This method breaks down a user’s single question into multiple sub-questions and executes hundreds of searches simultaneously. For example, when searching for ‘climate change’s impact on regional economy,’ it provides expert-level reports by synthesizing all information from related papers to news and statistics.
For deeper information needs, users can utilize the ‘Deep Search’ function. This processes hundreds of queries simultaneously to generate expert-level analysis reports with references in just minutes.
All users in the United States can already use AI Mode, accessible through a button at the top of search results pages.
New Experience of Searching with Camera
The ‘Search Live’ feature has changed the paradigm of search. Simply pointing a smartphone camera at an object provides real-time, conversational information. Point at a broken appliance to get repair instructions, or point at a building while traveling to instantly learn about its history and visitor information.
Online shopping has also been revolutionized. When selecting clothes in AI Mode, users can experience ‘virtual try-on’ based on their photos. It precisely renders fabric texture and stretch, allowing users to preview the actual fit.
New Tools for Creative Fields
Image generation AI ‘Imagen 4’ and video generation AI ‘Veo 3’ have also significantly advanced. Imagen 4 reproduces complex fabric textures and animal fur with remarkable precision, with dramatically improved text and typography expression capabilities. It supports up to 2K resolution.
Veo 3 goes beyond simple video generation to create audio as well. It automatically generates background music, sound effects, and even character dialogue to produce high-quality video content.
Google also unveiled ‘Flow,’ a video creation tool that integrates these technologies. This comprehensive creative platform can produce cinema-quality videos from simple text descriptions.
Premium AI Service Launch
Google also introduced paid subscription services for advanced users. Through ‘AI Pro’ and ‘AI Ultra’ plans, it provides more features and higher usage limits. Particularly, the Ultra plan at $249.99 per month (approximately 340,000 won) offers early access to Veo 3 and Deep Think mode.
Ultra subscribers can also use ‘Agent Mode.’ This feature allows AI to handle complex tasks automatically, from apartment hunting to tour booking.
The Future of Voice Calls: Google Beam
The 3D video calling technology previously known as ‘Project Starline’ has evolved into ‘Google Beam.’ Using AI technology, it converts regular 2D video calls into immersive 3D experiences. Through six cameras and AI, it renders the other person’s appearance three-dimensionally and even supports real-time language translation.
Outlook and Implications
I/O 2025 clearly demonstrated Google’s intention to develop AI technology not as a simple tool but as a ‘collaborative partner.’ Google CEO Sundar Pichai described it as “evolution from information to intelligence,” emphasizing that AI will become a true assistant helping users achieve their goals.
Industry experts predict this announcement will significantly strengthen Google’s position in AI competition. Particularly, innovations in search and creative fields are analyzed as presenting new challenges for competitors.
Google’s AI technology is currently used by over 400 million people monthly, with the number of developers exceeding 7 million. With the full implementation of these new features, the popularization of AI technology is expected to accelerate further.
