"Thank you. Please enjoy the rest of Google I/O."
On May 19 (local time) at the Shoreline Amphitheatre in Mountain View, California, where Google's annual flagship developer event I/O was held, Demis Hassabis, CEO of Google DeepMind and known as "the father of AlphaGo," wrapped up his one-hour-and-50-minute opening keynote. The audience responded with applause — and a touch of puzzlement. Tradition had it that the I/O keynote would open and close with Sundar Pichai, Google's CEO, but this year it opened with Pichai and ended with Hassabis. It was the first time since 2019 (the 2020 event was canceled due to COVID-19) that Pichai did not deliver the finale. In 2019, the closing was handled by Jeff Dean, Google's chief scientist overseeing technology development. Handing the closing to the leader of subsidiary DeepMind rather than to Google's own chief is highly unusual.
At Big Tech keynotes, where major products are unveiled and visions for the future are presented, the choice of speaker and the order of presentations carry significant meaning. Until now, Hassabis's role at I/O had been limited to introducing Google's AI models, such as Gemini, or AlphaFold, the AI model that predicts protein structures. Last year, he laid out a vision to advance Gemini's multimodal capabilities into a "world model." This year, he took the closing stage and focused on showcasing the progress and future direction of Google's AI. What message did Google intend to send by entrusting the keynote's finale to Hassabis? Three takeaways stand out.
AI Embedded Across All Google Services... Hassabis's Commanding Presence
The core message of Google I/O 2026 was the integration of AI across all Google services and the dawn of the agent era for the masses. Leveraging its strength as the owner of diverse platforms — the Android smartphone OS, the Google search tool, the YouTube video service, and AI glasses — Google's strategy is to unify its AI ecosystem around Gemini. Intelligent search now allows users to ask questions on YouTube via voice AI to find the videos they want and to search using a wide range of formats — text, images, files, and videos. Smart glasses were also unveiled, in which a Gemini-powered model on the eyewear summarizes book content and lets users purchase related products via voice commands. Rather than emphasizing how Google's AI performance compares to that of competitors, the company focused on signaling that it is leading an era in which everyone uses an AI agent.
The figure spearheading the development of Gemini, embedded throughout Google's services, is Hassabis. DeepMind, which he co-founded in 2010, was acquired by Google in 2014. He went on to serve as Google's vice president of AI before being named CEO. Founded as a search service company in 1998, Google has now transformed into an AI company leading the global AI market — and Hassabis sits at its center. Last month, marking the 10th anniversary of the AlphaGo matches, he visited Korea and met with President Lee Jae-myung and major business leaders to discuss partnerships, demonstrating his rising stature.
The next-generation AI "Gemini 3.5 Flash," the AI agent "Gemini Spark," and the multimodal AI tool "Gemini Omni" — all unveiled at this year's I/O — were also developed under Hassabis's leadership. Gemini 3.5 Flash, a lightweight model, and Gemini Spark, which individuals can run around the clock, embody Google's message that it is leading the era of AI for the masses. Gemini Omni is an AI model capable of turning any form of input — including video — into the desired output. It is a multimodal service following the success of "Nano Banana Pro," the image-generation model that became a sensation immediately after its November 2024 debut. In his keynote, Hassabis emphasized that while Gemini Omni begins with video editing, it will gradually evolve into a world model capable of input and output across every domain.
DeepMind is Google's AI development subsidiary. Through this year's I/O, DeepMind has emerged as the central organization overseeing the key features of all Google services. Hassabis's standing has been strengthened accordingly. In his keynote, he said, "It's exciting to see Gemini transforming Google products that are used by billions of people every day," noting how encouraging it is that Gemini, which he has led, is now integrated across all of Google's platforms — from search and video to glasses.
"At the Threshold of a New Singularity"... AGI Era Is Not Far Off
In his closing remarks at I/O, Hassabis signaled that the era of artificial general intelligence (AGI) — AI surpassing human intelligence — is imminent. While debate continues over whether AI can advance enough to match humans, he has consistently maintained that AGI could be realized by 2030. Google sought to convey that it is closing in on AI development rivaling human capability, and Hassabis — who shocked the world a decade ago with the debut of AlphaGo — served as the messenger.
Calling AGI the most important invention in history, Hassabis said, "AGI is becoming a reality and will be the most profound and impactful technology ever invented." He stressed, "Google's cutting-edge research and products will help realize AGI's incredible potential for global benefit. Looking back, we will realize that we were standing at the threshold of the singularity." He added, "This could be a deeply meaningful moment for humanity," and "this technology will amplify human ingenuity and usher in a new golden age of scientific discovery and progress, improving the lives of everyone around the world."
The remarks suggest that, while people may not yet realize it, we are now entering the gateway to AGI. Despite the AI bubble concerns raised since last year and the AI pessimism stirred this year by fears of job displacement, Hassabis offered an optimistic outlook, saying that we may already have entered the AGI era. The very reason DeepMind was founded 16 years ago was the belief that realizing AGI would benefit humanity.
In an interview with Mike Allen, co-founder of online media outlet Axios, on the second day of I/O on May 20, Hassabis again predicted that the AGI era would arrive around 2030. "In my view, we're only a few years away [from AGI becoming reality], and people will be able to feel the change through AI agents," he said. "Five or 10 years from now, when we look back at 2026 and 2027, we'll say, 'That was when we entered the AGI era.'" On the timing of AGI's realization, he forecast a window of 2029 to 2031, saying, "It could be 2030, give or take a year."
"AI Is a Tool for the World"... Stressing AI's Benefits
"Looking back, the reason I have devoted my entire life to AI is that I believed AI would be the ultimate tool for advancing science and our understanding of the world. It's truly wonderful to see that dream coming true."
"I have always believed that the most important application of AI should be improving human health."
Throughout the I/O event, Hassabis worked to highlight how AI can enrich human life. In the latter half of his keynote, he explained how Google's AI is shaping advances in science and medicine. Introducing the launch of "Gemini for Science," a platform of AI tools to accelerate scientific research, he stressed that AI is now helping scientists in nearly every field.
He also emphasized that Google's AI models are actually solving complex problems. "AlphaEarth Foundations" is contributing to addressing issues such as deforestation and food security, while "WeatherNext," a state-of-the-art weather forecasting model, is helping predict hurricane paths faster and more accurately than existing systems.
Hassabis also stressed that models developed by DeepMind, such as AlphaFold and AlphaGenome, are tools for advancing human life and medicine. The recipient of the 2024 Nobel Prize in Chemistry for AlphaFold, the AI-powered protein structure prediction technology, said, "Biological models like AlphaFold and AlphaGenome are becoming standard research tool, used by millions of scientists around the world to make important advances in their respective fields," he said. "I'd like to call this 'Science at digital speed.'"
Hassabis said the molecular interaction analysis model from Isomorphic Labs, a DeepMind spinoff, is dramatically accelerating drug development, with projects targeting immune diseases and cancer treatments now in preclinical stages. "What seemed impossible just a few years ago is clearly achievable now," he said, predicting that AI will also make significant contributions to solving medical challenges.
Apparently mindful of AI security concerns that have surfaced since the emergence of "Claude Mitos" from rival Anthropic, Hassabis said Google is preparing for security on multiple fronts. "This is a time of enormous possibility but also tremendous responsibility," he said. "We must clearly identify potential problems and use every tool at our disposal to ensure the safety of agentic systems and ultimately of AGI itself."
"One area of risk that has drawn considerable attention recently is cybersecurity," Hassabis said. "Google has invested in this field for decades and is protecting codebases worldwide using cutting-edge capabilities and deep expertise." He added, "We have tools like CodeMender, a code security agent that automatically finds and fixes critical software vulnerabilities. We're inviting a select group of experts to test the new CodeMender API and will roll it out to more users soon."
